DESCRIPTION

COMPOSITION PLAN CREATING APPARATUS, METHOD, PROGRAM,
AND RECORDING MEDIUM

Technical Field

[0001] The present invention relates to a composition plan creating apparatus, method, and program creating a composition plan for mixing a plurality of kinds of composition raw materials, and a computer readable recording medium.

Background Art

[0002] In many industries such as steel industry, various kinds of purchased composition raw materials having various properties are mixed, where it is required that properties after mixing fall within a constant range. Further, when creating a composition plan, expenses (costs) are determined as an important index, and it is required to minimize purchase costs and manufacturing costs, as well as shipping costs for shipping raw materials, and so on. Furthermore, for avoiding running out of stocks of the composition raw materials, it is required to plan a composition over plural days while changing composition ratios.

[0003] To achieve a composition plan that satisfies the above requirements, it is necessary to grasp vast amounts of Information such as various properties of various kinds of purchased composition raw materials, stock Information of various kinds of composition raw materials, purchase costs, and so on. Accordingly, to create the composition plan manually, all of these vast amounts of Information should be grasped to determine a composition, which requires a considerable amount of time. Further, accompanying low response to changes of Information, there also occurs a problem such that properties after mixing do not fall within a required range in reality.

[0004] Conventionally, various approaches are proposed as a technique to determine this kind of composition. For example, as disclosed in "Determination of compounding ratio of stock coal" of Patent Document 1, there is an approach to repeat changing a composition ratio or the like by a certain changing width after an appropriate initial value is inputted, thereby allowing to minimize costs while satisfying quality.

[0005] Further, as disclosed in "Method for calculating mixing ratio for raw material for cement clinker firing" of Patent Document 2, there is a method to plan a composition optimized using linear programming.

[0006] Moreover, as disclosed in "Method and device for preparing production and physical distribution plan, method and device for controlling physical distribution, computer readable recording medium, and computer program" of Patent Document 3, there is an approach to divide a plan creation period into

planning target periods, and repeat application of mathematical programming to the divided periods.

[0007] Patent Document 1: Japanese Patent Application Laid-open No. HOl-104688
Patent Document 2: Japanese Patent Application Laid-open No. 2001-146441
Patent Document 3: Japanese Patent Application Laid-open No. 2003-216695

Summary of the Invention

[0008] In the approach to repeat recalculation to create a composition plan as disclosed in "Determination of compounding ratio of stock coal" of aforementioned Patent Document 1, (1) it is necessary to perform calculation while changing the composition ratio based on the set changing width, and repeat evaluation of results many times until a satisfactory result is obtained. Accordingly, there are problems that (2) it takes a long period of time to create a composition plan when there are many kinds of composition raw materials, (3) costs are not always minimized because convergence calculation is performed by a repetitive approach, (4) stocks of composition raw materials are not considered, which makes it impossible to create a composition plan over plural days while avoiding running out of stocks, and so on.

[0009] Moreover, in the approach to create a composition plan using linear programming as disclosed in "Method for calculating mixing ratio for raw material for cement clinker firing" of aforementioned Patent Document 2, there are problems that (1) it is only targeted at satisfying properties, suggesting no approach to minimize costs, (2) it is required to create a composition plan over plural days based on Information of initial stock amounts, reception financial resources, and so on, but it is not considered at all to plan a composition over plural days considering stocks to avoid running out of stocks, and the like.

[0010] Furthermore, in the approach to divide a plan creation period into planning target periods, and repeat application of mathematical programming to the divided periods as disclosed in "Method and device for preparing production and physical distribution plan, method and device for controlling physical distribution, computer readable recording medium, and computer program" described in aforementioned Patent Document 3, there are problems that (1) it aims at planning a manufacturing start time and end time and a stock transition when manufacturing products, or planning a transport start time and end time and a stock transition when transporting products and raw materials, that is, it aims at preparing a production and physical distribution plan, and does not also at planning what ratios are better for compounding raw materials and making qualities and properties after mixing fall within a preset range when mixing them, in other words, it does not aim at making a composition plan,(2) it is possible to optimize a production start time of products and a transport start time and amount of materials, but is not possible to achieve that, when mixing raw materials, the mixed raw materials satisfy required qualities and properties.

[0011] Due to the above reasons, none of Patent Documents 1 to 3 achieves (4) making a composition plan over plural days which (1) avoids running out of stocks, (2) satisfies properties, and (3) minimizes costs as required above.

[0012] Accordingly, there is proposed an approach which enables to build a formula model representing demand-supply balance restrictions of composition raw materials and a formula model representing property restrictions after mixing using mathematical programming, and operate a simulator and an optimization calculating unit in conjunction so as to create a composition plan over plural days while avoiding running out of stocks, satisfying properties, and minimizing costs, when creating a composition plan for mixing a plurality of kinds of composition raw materials.

[0013] To achieve the above-described object, the present invention is structured as follows.
(1) A composition plan creating apparatus creating a composition plan for mixing a plurality of kinds of composition raw materials includes:
a simulator calculating supply-demand conditions of the composition raw materials and properties after mixing;
a data obtaining unit obtaining data containing arrival schedules of the composition raw materials, stock conditions of the composition raw materials, properties of the composition raw materials, and cost Information;
a model building unit building a formula model representing a demand-supply balance restriction of the composition raw materials and a formula model representing a property restriction after mixing, with an optimization period set in advance from a planning start date and time being a target, based on the data obtained by the data obtaining unit; and
an optimization calculating unit performing an optimization calculation based on an objective function built with respect to costs using the formula models built by the model building unit, so as to calculate an instruction to the simulator.

(2) The composition plan creating apparatus as
described in (1) further includes
an output unit outputting a composition plan as a simulation result from the simulator.

(3) In the composition plan creating apparatus as
described in (1),
with data obtained by the data obtaining unit being given, first,
(a) the optimization calculating unit creates a

calculation instruction for an optimization period from a planning start date and time,
(b) with a calculation instruction created by the optimization calculating unit being given, the simulator performs a simulation only for a simulation period set in advance,
(c) a result of the simulation is determined as a composition plan only for a plan determination period set in advance,
(d) a date and time just after the determination is set as a new planning start date and time, and
with an already determined composition plan being given, a series of the processing (a) to (d) to determine a composition plan for a new plan determination period is performed repeatedly until a composition plan for a plan creation period is determined, to thereby create a composition plan of a plan creation period.
(4) In the composition plan creating apparatus as
described in (1),
when creating the composition plan, there is planned a composition approximated to a composition ratio given as a target with respect to a part or all of the composition raw materials.
(5) In the composition plan creating apparatus as
described in {1),
when creating the composition plan, there is planned a composition which does not cause a large separation between the composition ratio of a

previous day and the composition ratio of a following day.
(6) In the composition plan creating apparatus as
described in (1),
when creating the composition plan, if a stock of a composition raw material used on a previous day still exists on a following day, a composition using the composition raw material is planned.
(7) In the composition plan creating apparatus as
described in {1),
when creating the composition plan, it is possible to specify a part of the composition plan in advance.
(8) The composition plan creating apparatus as
described in (1) further includes:
a linearization unit formulating, when a formula model representing the property restriction after mixing includes a nonlinear formula, a formula model by introducing a linear formula instead of the nonlinear formula; and
a judgment unit judging whether or not a solution obtaining result by the optimization calculation unit using the formula model formulated by the linearization unit satisfies a formula model including the nonlinear formula.
(9) In the composition plan creating apparatus as
described in (8),
when the property restriction after mixing has a lower limit value, the linear formula is a formula

forming a lower limit of the nonlinear formula, and
when the property restriction after mixing has an upper limit value, the linear formula is a formula forming an upper limit of the nonlinear formula,
(10) In the composition plan creating apparatus as
described in (8),
in formulating a formula model by introducing the linear formula instead of the nonlinear formula, when the property restriction after mixing has a lower limit value, the linearization unit sets a temporary lower limit value smaller than the lower limit value, and when the property restriction after mixing has an upper limit value, the linearization unit sets a temporary upper limit value larger than the upper limit value.
(11) In the composition plan creating apparatus as
described in (10),
when the solution obtaining result by the optimization calculating unit using a formula model formulated by the linearization unit does not satisfy the formula model including the nonlinear formula, solution obtaining by the optimization calculating unit is repeated while slightly increasing the temporary lower limit value or slightly decreasing the temporary upper limit value.
(12) In the composition plan creating apparatus as
described in (1),
the data obtaining unit obtains purchase cost Information of the composition raw materials and

shipping cost Information when using a ship as the cost Information, and
the optimization calculating unit performs an optimization calculation based on an objective function built with respect to purchase costs and shipping costs of the composition raw materials using the formula model built by the model building unit, so as to calculate an instruction to the simulator.
(13) The composition plan creating apparatus as
described in (12) further includes
an extraction unit extracting a fixed item from predetermined items of a ship allocation plan.
(14) In the composition plan creating apparatus as
described in (13),
the predetermined items of the ship allocation plan are loading port, loaded brand, loaded amount, unloading port, unloaded brand, and unloaded amount.
(15) In the composition plan creating apparatus as
described in (13),
the shipping cost Information obtained by the data obtaining unit includes Information of freights by ship, by loading port, and by unloading port and Information of freights by loaded brand and by unloading port, and
which of the freights by ship, by loading port, and by unloading port and the freights by loaded brand and by unloading port are used in the optimization calculating unit is determined according to the fixed item extracted by the extraction unit.

(16) In the composition plan creating apparatus as
described in (12),
in the optimization calculating unit, the optimization calculation is performed based on an objective function built with respect to prevention of separation from a reference composition plan created in advance, in addition to the objective function built with respect to purchase costs and shipping costs of the composition raw materials.
(17) A composition plan creating method creating a
composition plan for mixing a plurality of kinds of
composition raw materials includes the steps of:
obtaining data containing arrival schedules of the composition raw materials, stock conditions of the composition raw materials, properties of the composition raw materials, and cost Information;
building a formula model representing a demand-supply balance restriction of the composition raw materials and a formula model representing a property restriction after mixing, with an optimization period set in advance from a planning start date and time being a target, based on the obtained data; and
performing an optimization calculation based on an objective function built with respect to costs using the built formula models, so as to calculate an instruction to a simulator which calculates supply-demand conditions of the composition raw materials and properties after mixing.
(18) A program for causing a computer to perform

processing of creating a composition plan for mixing a plurality of kinds of composition raw materials causes a computer to function as:
a simulator calculating supply-demand conditions of the composition raw materials and properties after mixing;
a data obtaining unit obtaining data containing arrival schedules of the composition raw materials, stock conditions of the composition raw materials, properties of the composition raw materials, and cost Information;
a model building unit building a formula model representing a demand-supply balance restriction of the composition raw materials and a formula model representing a property restriction after mixing, with an optimization period set in advance from a planning start date and time being a target, based on the data obtained by the data obtaining unit; and
an optimization calculating unit performing an optimization calculation based on an objective function built with respect to costs using the formula models built by the model building unit, so as to calculate an instruction to the simulator. (19) A computer readable recording medium records the program as described in (18) .
[0014] According to the present invention, it becomes possible to build a formula model representing demand-supply balance restrictions of composition raw materials and a formula model

representing property restrictions after mixing using mathematical programming, and operate a simulator and an optimization calculating unit in conjunction so as to create a composition plan over plural days while avoiding running out of stocks, satisfying properties, and minimizing costs, when creating a composition plan for mixing a plurality of kinds of composition raw materials. Furthermore, when creating a composition plan using mathematical programming, the composition plan can be created even when the formula model representing property restrictions after mixing includes a nonlinear formula.
Brief Description of the Drawings
[0015] Fig. 1 is a diagram illustrating a system structure example including a composition plan creating apparatus;
Fig. 2 is a block diagram illustrating a basic structure of the composition plan creating apparatus;
Fig. 3 is a diagram illustrating a detailed structure of the composition plan creating apparatus according to a first embodiment;
Fig. 4 is a flowchart illustrating steps of a composition plan creating method performed using the composition plan creating apparatus according to the first embodiment;
Fig. 5 is a diagram illustrating an overview of composition plan creation of the first embodiment;
Fig. 6 is a diagram for describing a restriction

that the stock amount of each brand is not less than a safe stock amount;
Fig. 7 is a diagram for describing a procedure of the composition plan creation in the first embodiment;
Fig. 8 is a diagram illustrating a detailed structure of a composition plan creating apparatus according to a second embodiment;
Fig. 9 is a flowchart illustrating steps of a composition plan creating method performed using the composition plan creating apparatus according to the second embodiment;
Fig. 10 is a diagram illustrating an overview of composition plan creation of the second embodiment;
Fig. 11 is a flowchart illustrating processing when a linear formula is introduced instead of a nonlinear formula;
Fig. 12 is a diagram illustrating an example of creating composition plans by beginning, middle, and end of a month;
Fig. 13 is a diagram illustrating past records before applying the present invention and a plan by a composition plan creating unit applying the present invention;
Fig. 14 is a diagram illustrating a detailed structure of a composition plan creating apparatus according to a third embodiment;
Fig. 15 is a flowchart illustrating steps of a composition plan creating method performed using the

composition plan creating apparatus according to the third embodiment;
Fig. 16 is a diagram illustrating an example of a ship allocation plan;
Fig. 17 is a diagram illustrating an example of a ship list;
Fig. 18 is a diagram illustrating an example of a table of freights by ship, loading port, and unloading port included in shipping cost Information when using a ship;
Fig. 19 is a diagram illustrating an example of a table setting freights to be used;
Fig. 20 is a flowchart illustrating processing when a linear formula is introduced instead of a nonlinear formula; and
Fig. 21 is a diagram illustrating a hardware structure example of a computer apparatus capable of functioning as a composition plan creating apparatus of the present invention.
Detailed Description of the Preferred Embodiments [0016] Hereinafter, embodiments to which the present invention is applicable will be described based on the drawings.
-First Embodiment-
Fig. 1 is a diagram illustrating a system structure example including a composition plan creating apparatus according to the present invention. As illustrated in Fig. 1, in a composition plan

creating apparatus 100, when creating a composition plan, data of restriction conditions and prerequisites including a plan creation period, an arrival schedule of composition raw materials, stock conditions of composition raw materials, properties of composition raw materials (characteristics (including qualities), conditions, and so on), cost Information (purchase cost Information of composition raw materials, and so on), which are needed for setting out a composition plan, are set by an operator or obtained from a process computer 105 or a business computer 106. For example, it may be configured that a part of the composition plan can be specified in advance.
[0017] The composition plan creating apparatus 100 creates a mixing plan for mixing various kinds (plural brands) of composition raw materials by performing a simulation, and obtains a composition plan so as to satisfy demand-supply balance restrictions of composition raw materials and property restrictions after mixing. The composition plan creating apparatus 100 attempts to optimize the composition plan by building a formula model representing the demand-supply balance restrictions of composition raw materials (also referred to as a "demand-supply balance model") and a formula model representing the property restrictions after mixing (also referred to as a "property model") using mathematical programming such as LP (linear

programming), MIP (mixed integer programming), QP (quadratic programming), and the like, details of which will be described later.
[0018] A display unit 103 displays use amounts of respective brands (composition ratios) obtained by the composition plan creating apparatus 100, incoming amounts, a stock transition graph, and various forms. [0019] An operator evaluation unit 104 enables the operator to evaluate an obtained composition plan from various aspects (for example, stock transition, property, and so on), and modify a composition ratio and so on as necessary if the result is unsatisfactory. When this happens, the operator changes the weight of an objective function or an index of evaluation, and/or changes a target period and a plan determination period for building the formula models as necessary. Further, the unit is able to reflect the operator's intention such as fixing all use amounts or a use amount of specified processing, and the like. Then the composition plan is created again in the composition plan creating apparatus 100.
[0020] Fig. 2 is a block diagram illustrating a basic structure of the composition plan creating apparatus 100 according to this embodiment. As illustrated in Fig. 2, the composition plan creating apparatus 100 is structured including a simulator (a stock transition simulator 201 and a property simulator 202), a model building unit (a demand-

supply balance model building unit 203 and a property model building unit 204), and a planning unit 205 functioning as an optimization calculating unit, and further has input and output units. [0021] The stock transition simulator 201 is a simulator which calculates the demand-supply condition (stock transition) of each composition raw material. The property simulator 202 is a simulator which calculates the property after mixing a composition raw material. By the stock transition simulator 201 and the property simulator 202 operating in conjunction with each other, the stock transition of a composition raw material and the property after mixing are calculated.
[0022] In this embodiment, based on input data 206 of a plan creation period, an arrival schedule of composition raw materials, stock conditions of composition raw materials, properties of composition raw materials, cost Information, and so on, which are needed for setting out a composition plan, the formula model representing the demand-supply balance restrictions (stock restrictions) is built by the demand-supply balance model building unit 203 and the formula model representing the property restrictions is built by the property model building unit 204, with an optimization period set in advance from a planning start date and time of the composition plan being a target, based on a time accuracy set in advance, and conforming to mathematical programming

such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like .
[0023] Using the formula models built by the demand-supply balance model building unit 203 and the property model building unit 204, the planning unit 205 performs an optimization calculation so as to create a composition plan while avoiding running out of stocks, satisfying required properties, and minimizing costs, and calculates a calculation instruction to the stock transition simulator 201 and the property simulator 202. Upon receipt of this calculation instruction, the stock transition simulator 201 simulates the stock transition, and the property simulator 202 simulates the property. [0024] With the composition plan creating apparatus according to this embodiment, the calculation instruction is not performed based on predetermined rules as is conventional. The calculation instruction based on a result of the optimization calculation performed by the planning unit 205 is outputted to the stock transition simulator 201 and the property simulator 202. Thus, it becomes possible to securely perform an optimal calculation instruction corresponding to an event on a certain occasion.
[0025] Further, for example, when a simulation for a plan determination period set in advance as illustrated in Fig. 7 by the stock transition

simulator 201 and the property simulator 202 is completed, the formula model representing the stock restrictions is built by the demand-supply balance model building unit 203 based on Information of stock transition and property in a final state of the plan determination period, and the formula model representing the property restrictions is built by the property model building unit 204, which are given to the planning unit 205. When given Information of the stock transition and the property, the planning unit 205 performs the optimization calculation. [0026] As described above, an optimal composition plan can be created by carrying out a detailed simulation with the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the property model building unit 204), and the planning unit 205 operating in conjunction. That is, the simulation performed in this embodiment is not based on predetermined rules as is conventional but is performed based on results of the optimization calculation, and thus a logical optimal solution can be obtained securely by just carrying out one time of simulation. Accordingly, it is not necessary to perform simulations again and again repeatedly after evaluating a simulation result as is conventional, and a simulation result 207 can be created quickly with high accuracy. Therefore, it is sufficiently possible to create a plan within a

practical time period even when the target of creating a composition plan is large in scale.
[0027] Further, when the plan creation period becomes long, the period to be considered becomes long, and there occurs a problem that it becomes not possible to obtain a solution because the scale of problem becomes large by a conventional method. However, with this approach, it is possible to reduce the scale of problem by dividing into optimization periods, and thus it is possible to solve the problem even when the plan creation period becomes long. The simulation result 207 obtained as described above is outputted as a composition plan,
[0028] Further, when the scale of a model built by the demand-supply balance model building unit 203 or the property model building unit 204 is quite large or when there are great many complicated restriction conditions, the demand-supply balance model building unit 203 and the property model building unit 204 are configured to obtain only an important part having a large influence on creation of composition plan out of the demand-supply balance restrictions and the property restrictions described in the stock transition simulator 201 and the property simulator 202. Thus, the scales of formula models of the demand-supply balance model building unit 203 and the property model building unit 204 come within an appropriate range, and the optimization calculation can be performed within a practical time period. The

stock transition simulator 201 and the property simulator 202 are capable of describing all demand-supply balance restrictions and property restrictions which should be considered, and thus it is assured that the composition plan created by performing one time of simulation is practicable in reality. [0029] As described above, in this embodiment, the composition plan is created by operating the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the property model building unit 204), and the planning unit 205 in conjunction. Thus, (1) the composition plan can be created without performing the simulation repeatedly. (2) The calculation time can be reduced by allowing the planning unit 205 to obtain only an important part having a large influence on creation of composition plan, and (3) it becomes possible to solve a large-scale problem.
[0030] Hereinafter, with reference to Figs. 3 to 7, a structure of the composition plan creating apparatus 100 according to this embodiment and steps of a composition plan creating method performed using this apparatus 100 will be described in detail. Fig. 3 is a diagram illustrating a detailed structure of the composition plan creating apparatus 100 with respect to the basic structure of the composition plan creating apparatus 100 described using Fig. 2. Further, Fig. 4 is a flowchart illustrating the steps

of the composition plan creating method performed using this apparatus 100.
[0031] Fig. 5 illustrates a schematic diagram of a raw material composition as a target of implementing the composition plan creating apparatus according to this embodiment. Using Fig. 5, creation of a composition plan for mixing composition raw materials to make the raw materials after mixing satisfy required properties will be described. However, this is merely an example, and the composition plan creating apparatus according to this embodiment can be applied to and particularly effective for creating a composition plan for mixing various kinds of composition raw materials to satisfy required properties with minimum costs.
[0032] In the composition plan creating apparatus according to this embodiment, it is required to satisfy the demand-supply balance rest rietions. That is, the sum of daily use amounts of the composition raw material of each brand is less than the amount of adding an incoming amount and an initial stock amount. Further, it is required that raw materials after mixing have various properties within the ranges of thresholds set in advance. Moreover, the first alm of a composition plan is to minimize purchase costs of composition raw materials.
[0033] (1) Obtaining input data (input data obtaining unit 301 in Fig. 3, Step S401 in Fig. 4) Information needed in this processing (arrival

schedules of composition raw materials, stock conditions of composition raw materials, properties of composition raw materials, cost Information, and so on) is read online, and the operator makes a modification to them as necessary.
[0034] The input data obtaining unit 301 and Step S401 described above are examples of a data obtaining unit and processing thereof mentioned in the present invention.
[0035] (2) Setting a composition plan creation period (plan creation period setting unit 302 in Fig. 3, Step S402 in Fig, 4)
A period to create a composition plan is set. This creation period can be set to an arbitrary period as needed by the planner. Here, a plan for ten days is made as an example.
[0036] (3) Setting a composition plan creation time accuracy (time accuracy setting unit 303 in Fig. 3, Step S403 in Fig. 4)
A time accuracy and a simulation accuracy for creating the composition plan are set. These time accuracy and simulation accuracy can be set to arbitrary accuracies separately as needed by the planner. For example, the accuracies can be increased in the first half of the plan creation period in which high accuracies are required in planning, and then the accuracies can be decreased in the second half of the plan creation period in which a rough plan will suffice, thereby enabling efficiënt

plan creation with sufficiënt accuracies within a short time.
[0037] (4) Setting an optimizat ion period (optimization period setting unit 304 in Fig. 3, Step S404 in Fig. 4)
An optimization period for creating the composition plan is set. This optimization period can be set to an arbitrary target period separately as needed by the planner. Here, the optimization period is set to three days through the plan creation period as an example.
[0038] (5) Setting a plan determination period (plan determination period setting unit 305 in Fig. 3, step S405 in Fig. 4)
A plan determination period for determining a composition plan is set. This plan determination period can be set to an arbitrary period separately as needed by the planner. For example, the plan determination period can be shortened in the first half of the plan creation period in which high accuracies are required in planning, and then the plan determination period can be lengthened in the second half of the plan creation period in which a rough plan will suffice, thereby enabling efficiënt plan creation with sufficiënt accuracies within a short time. Here the plan determination period is set to one day as an example. In this case, it is determined for the first day through the plan creation period with respect to the composition plan

obtained as a result of simulation based on a solution for a formula model.
[0039] (6) Formulating demand-supply balance restrictions of the composition plan into a formula model (demand-supply balance model building unit 306 in Fig. 3 (corresponding to the demand-supply balance model building unit 203 in Fig. 2), step S406 in Fig. 4)
Based on all or part of data obtained by the input data obtaining unit 301, demand-supply balance restrictions are formulated into formula models for the set optimization period by the set time accuracy. A variable representing a use amount of each brand is defined as shown by following Formula (1). Further, a variable representing a stock amount of each brand is defined as shown by following Formula (2). [0040]
Xi,d : use amount of brand i on day d ... Formula (1) [0041]
Si,d : stock amount of brand i on day d ... Formula (2)
[0042] A formula model built based on demand-supply Information, that is, a demand-supply balance restriction model is shown below. It is required that the stock amount of each brand is not less than a constant value called a safe stock amount (see Fig. 6) . The restriction in this case is represented by following Formula (3).

[0043]
Si,d ^ SafeStock;i,d : SafeStock;i,d is a safe stock amount (constant) of brand i on day d ... Formula (3) [0044] Further, the stock amount of each brand is determined from the stock amount on a previous day, the incoming amount on the previous day, and the use amount on the previous day. A restriction formula representing the relation in this case is represented by following Formula (4). [0045]
Si,d = Si,d-1 + yi,d-1 - Xi,d-1 : yi,d is incoming amount (constant) of brand i on day d ... Formula (4) [0046] Further, it is required that the sum of use amounts of each brand on a certain day matches a scheduled use amount. A restriction formula representing a relation in this case is represented by following Formula (5) [0047]
: Ud is use amount (constant) on day d ... Formula (5)
[0048] Further, the operator sets a target composition ratio from factors with respect to purchase of various kinds of raw materials and the like, and demands that a composition plan with a composition ratio close to the given target composition ratio is created. That is, when the composition ratio becomes largely different from the operator's assumption, it is expected that the

assumed purchase amount is not satisfied, the purchase amount is exceeded, and/or it causes an unreasonable operation in the operation equipment. Thus, it is necessary to output a composition ratio close to the composition ratio given as a target. Restrictions to achieve the aforementioned function are shown below. [0049]
Xi,d - xTargeti,d ≤ OverUi,d ••• Formula (6)
xTargeti,d is a target use amount (constant) of brand i on day d.
OverUi,d is an overuse amount from a target use amount of brand i on day d. [0050]
xTargeti,d - Xi,d ≤ CoverUi,d ••• Formula (7) xTargeti,d is the target use amount (constant) of brand i on day d.
CoverUi,d is a deficiënt amount from the ,>^ target use amount of brand i on day d.
[0051] Further, when the composition ratio on a previous day and the composition ratio on a following day separate largely, it causes a difficulty in operation. Specifically, it causes an increase in setup time for using a different raw material and a failure in equipment. Accordingly, it is required to create a composition plan which does not cause a large separation between the composition ratio on a previous day and the composition ratio on a following day. A restriction to achieve this function is

illustrated below. [0052]
Xi,d - Xi,d < Diffi,d
Xi,d-i - Xi,d ^ Diffi,d ... Formula (8) Diffi,d is a difference between a use amount of brand i on day d and a use amount of brand i on day d-1.
[0053] In addition, if a composition is not performed on a following day when there is a stock even though the composition is performed on a previous day, this results in a number of small stocks of brands being left and causes a difficulty in operation. Accordingly, when a composition is performed on a previous day but there is a stock on a following day, it is required to create a composition plan to perform the composition on the following day. A restriction to achieve this function is illustrated below. [0054]
x5i,d + (2 - x5i,d-1 - s5i,d) ≥1 ... Formula (9)
xδi,d is a variable which is O when the use amount of brand i on day d is O, and is otherwise 1.
sδi,d is a variable which is O when the stock amount of brand i on day d is O, and is otherwise 1. [0055] Note that the above-described demand-supply balance restrictions are just an example and may be replaced with other restrictions, or other restrictions may be added. [0056] (7) Formulating property restrictions of the

composition plan into a formula model (property model building unit 307 in Fig. 3 (corresponding to the property model building unit 204 in Fig. 2), step S407 in Fig. 4)
Based on all or part of data obtained by the input data obtaining unit 301, property restrictions are formulated into a formula model for the set optimization period by the set time accuracy.
[0057] A property restriction model built based on property Information is illustrated below. Here, the case where a property fu(Xi,d) of u component can be calculated by following Formula (10) is considered.
[0058]
Formula (10)
Wu,i is a property with respect to u component contained in the brand i.
Vu,i is a constant of the influence on u property in proportion to a secondary term of the composition ratio in the brand i.
[0059] The property in the above-described Formula (10) is formed of the composition ratio and a secondary term of the composition ratio. However, normally the influence here of the secondary term on the property is quite small, and thus in the formulation in the optimization the property is considered with following Formula (11) in which the secondary term is omitted.

[0060]

[0061] Properties of a raw material after mixing has to satisfy required property restrietions. A restriction formula in this case is represented by following Formula (12) . [0062]

LwWu,d is a property lower limit value
(constant) with respect to u component on day d. UpWu,d is a property upper limit value
(constant) with respect to u component on day d.
[0063] Note that the above-described property restrictions are just an example and may be replaced with other restrictions, or other restrictions may be added.
[0064] The above-described demand-supply balance model building unit 306 (demand-supply balance model building unit 203) and Step S406, and the property model building unit 307 (property model building unit 204) and Step S4G7 are examples of a model building unit and processing thereof mentioned in the present invention.
[0065] (8) Optimizing a composition plan formula model based on an objective function (composition

plan solution obtaining unit 309 in Fig. 3 (corresponding to the planning unit 205 in Fig. 2), Step S409 in Fig. 4)
The demand-supply balance model and the property model formed of the linear and integer restriction formulas built above make up a composition plan formula model. A problem is solved as an optimization problem by mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like based on an objective function set in advance, so as to calculate an optimal use amount and incoming amount.
[0066] Here, an example using a linear formula for the objective function is shown. An object of this embodiment is to minimize costs, and an example of objective function Z is shown by Formula (13). [0067]
, . . Formula (13) Vi is a purchase cost (constant) of brand i.
[0068] Further, when it is required to approximate a composition plan to a composition ratio close to the given target composition ratio and to further create a composition plan by which the composition ratio on a previous day and the composition ratio on a following day does not separate largely, the objective function is following Formula (14).
[0069]

... Formula (14)
[0070] An optimal solution for the composition plan formula model combining the demand-supply balance model and the property model is obtained by solving the above formulated formulas (formula models) by mixed integer programming.
[0071] The above-described composition plan solution obtaining unit 309 (the planning unit 205) and Step S409 are examples of an optimization calculating unit and processing thereof mentioned in the present invention.
[0072] (9) Simulating a stock transition based on the obtained solution (stock transition simulator 311 in Fig. 3 (corresponding to the stock transition simulator 201 in Fig. 2), Step S412 in Fig. 4)
Based on the solution for the composition plan formula model and all or part of data obtained by the 'input data obtaining unit 301, a simulation for all or part of a targeted composition for the set plan determination period is performed by a set plan creation accuracy. In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in

actual operation without any problem. Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation.
[0073] Further, as an example of a restriction which is difficult to handle by a formula model, a setup time needed for setting up equipment when the composition ratio is changed or the like is taken into the simulation so as to perform the simulation precisely, and thus it is possible to set out a composition plan considering detailed restrictions required in actual operation.
[0074] (10) Simulating a property based on the obtained solution (property simulator 312 in Fig. 3 (corresponding to the property simulator 202 in Fig. 2), Step S413 in Fig. 4)
Based on the solution for the composition plan formula model, the stock transition simulated by the stock transition simulator 311, and all or part of data obtained by the input data obtaining unit 301, a property is simulated for all or part of a targeted composition for the set plan determination period by a set plan creation accuracy, so as to obtain a property result of a composition raw material after mixing.
[0075] In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model

are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in actual operation without any problem. For example, the influence of the secondary term of the composition ratio is ignored in calculation in the optimization because its influence on quality and property is quite small, but even the secondary term is considered in the simulation to calculate the quality and property by the aforementioned Formula
(10). Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation.
[0076] The above-described stock transition simulator 311 (stock transition simulator 201) and Step S412, and the property simulator 312 (property simulator 202) and Step S413 are examples of a simulator and processing thereof mentioned in the present invention.
[0077] (11) Determining the composition plan
(determination unit 313 in Fig. 3, Step S414 in Fig. 4)
The composition plan derived by the stock transition simulation and the property simulation is determined for the set plan determination period. As shown in Fig. 7, since the plan determination period is set to one day in this embodiment, first one day

of the created composition plan is determined. The plan for the portion of the created composition plan that is not within the plan datermination period is discarded without being determined.
[0078] (12) Judging whether the plan for the plan creation period or the plan determination period is determined (judgment unit 314 in Fig. 3, Step S415 in Fig. 4)
It is judged whether the plan determination period determined at this point is determined for the plan creation period set in advance. In this embodiment, since the plan creation period is ten days, the plan for the plan determination period is determined at the point when the plan is determined in a tenth loop. Accordingly, the composition plan for ten days is created at the point when determination of the plan is finished in the tenth loop, and the process is finished. [0079] (13) Updating the planning start date (updating unit 315 in Fig. 3, Step S416 in Fig. 4)
When the determined plan determination period is not determined for the plan creation period set in advance, the date and time just after the determined composition plan period in the composition plan is set as a new planning start date. In this embodiment, as illustrated in Fig. 7, the planning start date which is initially first day and zero o'clock in the first loop is updated to second day and zero o'clock, and the planning start date which is initially second

day and zero o'clock in the second loop is updated to third day and zero o'clock.
[0080] (14) Outputting the composition plan (output unit 316 of Fig. 3, Step S417 in Fig. 4)
The composition plan created as described above is displayed on the screen of the display unit 103 and/or is transmitted as data to a not-illustrated external apparatus by the output unit 316. [0081] The output unit 316 and Step S417 described above are examples of an output unit and processing thereof mentioned in the present invention. [0082] As described above, corresponding to the current stock transition state, formula models for the demand-supply balance restrictions and the property restrictions are built first by the plan creation time accuracy for the predetermined optimization period, a solution for the built composition plan formula model is obtained based on the objective function, and the stock transition and the property after mixing are simulated based on the obtained solution. The composition plan obtained from the simulation result is determined for the set plan determination period, and the date and time just after the plan determination period are set to the new planning start date and time. Thus, a series of processing for determining a composition plan for a new plan target period can be performed sequentially and repeatedly by a predetermined number of times, so as to create a composition plan for a desired plan

creation period. Accordingly, a composition plan requiring an arbitrary time accuracy can be optimized quickly in detail, and can further be applied as it is in actual operation.
[0083] -Second Embodiment-
Incidentally, the formula model representing the property restrictions after mixing may include a nonlinear formula, This case cannot be solved by linear programming or mixed integer programming, and a composition plan cannot be created. In a second embodiment, an example in which a composition plan can be created even when the formula model representing the property restrictions after mixing includes a nonlinear formula will be described.
[0084] A system structure example including a composition plan creating apparatus according to this embodiment and a basic structure of a composition plan creating apparatus 100 are the same as those illustrated in Figs. 1, 2, and will be described here also with reference to Figs. 1, 2.
[0085] As illustrated in Fig. 1, in the composition plan creating apparatus 100, when creating a composition plan, data of restriction conditions and prerequisites including a plan creation period, an arrival schedule of composition raw materials, stock conditions of composition raw materials, properties of composition raw materials { characteristics
(including qualities), conditions, and so on), cost Information (purchase cost Information of composition

raw materials, and so on), which are needed for setting out a composition plan, are set by an operator or obtained from a process computer 105 or a business computer 106.
[0086] The composition plan creating apparatus 100 creates a mixing plan for mixing various kinds of composition raw materials by performing a simulation, and obtains use amounts (composition ratios) of respective brands as a composition plan so as to satisfy demand-supply balance restrictions of composition raw materials and property restrictions after mixing. The composition plan creating apparatus 100 attempts to optimize the composition plan by building a formula model representing the demand-supply balance restrictions of composition raw materials (also referred to as a "demand-supply balance model") and a formula model representing the property restrictions after mixing (also referred to as a "property model") using mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like, details of which will be described later. [0087] A display unit 103 displays use amounts of respective brands (ratios) obtained by a composition plan creating unit 102, a stock transition graph, and various forms.
[0088] An operator evaluation unit 104 enables the operator to evaluate an obtained composition plan from various aspects (for example, stock transition,

property, and so on), and modify a composition ratio and so on as necessary if the result is unsatisfactory. When this happens, the operator changes the weight of an objective function or an index of evaluation, and/or changes a target period and a plan determination period for building the formula models as necessary. Further, the unit is able to reflect the operator's intention such as fixing all use amounts or a use amount of specified processing, and the like. Then the composition plan is created again in the composition plan creating apparatus 100 .
[0089] As illustrated in Fig. 2, the composition plan creating apparatus 100 is structured including a simulator (a stock transition simulator 201 and a property simulator 202), a model building unit (a demand-supply balance model building unit 203 and a property model building unit 204), and a planning unit 205 functioning as an optimization calculating •unit, and further has input and output units, [0090] The stock transition simulator 201 is a simulator which calculates the demand-supply condition (stock transition) of each composition raw material. The property simulator 202 is a simulator which calculates the property after mixing a composition raw material. By the stock transition simulator 201 and the property simulator 202 operating in conjunction with each other, the stock transition of a composition raw material and the

property after mixing are calculated.
[0091] In this embodiment, based on input data 206 of a plan creation period, an arrival Schedule of composition raw materials, stock conditions of composition raw materials, properties
(characteristics (including qualities), conditions, and so on) of composition raw materials, cost Information, and so on, which are needed for setting out a composition plan, the formula model representing the demand-supply balance restrictions (stock restrictions) is built by the demand-supply balance model building unit 203 and the formula model representing the property restrictions is built by the property model building unit 204, with an optimization period set in advance from a planning start date and time of the composition plan being a target, based on a time accuracy set in advance, and conforming to mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like. [0092] Using the formula models built by the demand-supply balance model building unit 203 and the property model building unit 204, the planning unit 205 performs an optimization calculation so as to create a composition plan while avoiding running out of stocks, satisfying required properties, and minimizing costs, and calculates a calculation instruction to the stock transition simulator 201 and the property simulator 202. Upon receipt of this

calculation instruction, the stock transition simulator 201 simulates the stock transition, and the property simulator 202 simulates the property.
[0093] With the composition plan creating apparatus according to this embodiment, the calculation instruction is not performed based on predetermined rules as is conventional. The calculation instruction based on a result of the optimization calculation performed by the planning unit 205 is outputted to the stock transition simulator 201 and the property simulator 202. Thus, it becomes possible to securely perform an optimal calculation instruction corresponding to an event on a certain occasion.
[0094] Further, for example, when a simulation for a plan determination period set in advance as illustrated in Fig. 7 by the stock transition simulator 201 and the property simulator 202 is completed, the formula model representing the stock restrictions is built by the demand-supply balance model building unit 203 based on Information of stock transition and property in a final state of the plan determination period, and the formula model representing the property restrictions is built by the property model building unit 204, which are given to the planning unit 205. When given Information of the stock transition and the property, the planning unit 205 performs the optimization calculation.
[0095] As described above, an optimal composition


plan can be created by carrying out a detailed simulation with the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the property model building unit 204), and the planning unit 205 operating in conjunction. That is, the simulation performed in this embodiment is not based on predetermined rules as is conventional but is performed based on results of the optimization calculation, and thus a logical optimal solution can be obtained securely by just carrying out one time of simulation. Accordingly, it is not necessary to perform simulations again and again repeatedly after evaluating a simulation result as is conventional, and a simulation result 207 can be created quickly with high accuracy. Therefore, it is sufficiently possible to create a plan within a practical time period even when the target of creating a composition plan is large in scale. [0096] Further, when the plan creation period becomes long, the period to be considered becomes long, and there occurs a problem that it becomes not possible to obtain a solution because the scale of problem becomes large by a conventional method. However, with this approach, it is possible to reduce the scale of problem by dividing into optimization periods, and thus it is possible to solve the problem even when the plan creation period becomes long. The simulation result 207 obtained as described above is

outputted as a composition plan.
[0097] Further, when the scale of a model built by the demand-supply balance model building unit 203 or the property model building unit 204 is quite large or when there are great many complicated restriction conditions, the demand-supply balance model building unit 203 and the property model building unit 204 are configured to obtain only an important part having a large influence on creation of composition plan out of the demand-supply balance restrictions and the property restrictions described in the stock transition simulator 201 and the property simulator 202. Thus, the scales of formula models of the demand-supply balance model building unit 203 and the property model building unit 204 come within an appropriate range, and the optimization calculation can be performed within a practical time period. The stock transition simulator 201 and the property simulator 202 are capable of describing all demand-supply balance restrictions and property restrictions which should be considered, and thus it is assured that the composition plan created by performing one time of simulation is practicable in reality. [0098] As described above, in this embodiment, the composition plan is created by operating the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the property model building unit 204), and the planning

unit 205 in conjunction. Thus, (1) the composition plan can be created without performing the simulation repeatedly. (2) The calculation time can be reduced by allowing the planning unit 205 to obtain only an important part having a large influence on creation of composition plan, and (3) it becomes possible to solve a large-scale problem.
[0099] Hereinafter, with reference to Figs. 8 to 13, a structure of the composition plan creating apparatus 100 according to this embodiment and steps of a composition plan creating method performed using this apparatus 100 will be described in detail. Fig, 8 is a diagram illustrating a detailed structure of the composition plan creating apparatus 100 with respect to the basic structure of the composition plan creating apparatus 100 described using Fig. 2. Further, Fig. 9 is a flowchart illustrating the steps of the composition plan creating method performed using this apparatus 100.
[0100] To describe an overview of composition plan creation, as illustrated in Fig. 10 for example, use amounts (composition ratios (proportions)) of brands A to N in each of plural steel plants a to b are determined as a composition plan so as to achieve demand-supply balances of composition raw materials (brands) in the steel plants a to c (such as avoiding running out of stocks of the brands A to N), satisfy required properties, and minimize costs. [0101] (1) Obtaining input data and setting initial

values and conditions (input data obtaining unit 801 in Fig. 8, Step S901 in Fig. 9)
Information needed in this processing (arrival schedules of composition raw materials, stock conditions of composition raw materials, properties of composition raw materials, cost Information, and so on) is read online, and the operator makes a modification to them as necessary.
[0102] The input data obtaining unit 801 and Step S901 described above are examples of a data obtaining unit and processing thereof mentioned in the present invention.
[0103] (2) Setting a composition plan creation period (plan creation period setting unit 802 in Fig. 8, Step S902 in Fig. 9)
A period to create a composition plan is set. This creation period can be set to an arbitrary period as needed by the planner. Here, a plan for ten days is made as an example.
[0104] (3) Setting a composition plan creation time accuracy (time accuracy setting unit 803 in Fig. 8, Step S903 in Fig. 9)
A time accuracy and a simulation accuracy for creating the composition plan are set. These time accuracy and simulation accuracy can be set to arbitrary accuracies separately as needed by the planner. For example, the accuracies can be increased in the first half of the plan creation period in which high accuracies are required in

planning, and then the accuracies can be decreased in
the second half of the plan creation period in which
a rough plan will suffice, thereby enabling efficiënt
plan creation with sufficiënt accuracies within a
short time.
[0105] (4) Setting an optimizat ion period
(optimization period setting unit 804 in Fig. 8, Step
S904 in Fig. 9)
An optimization period for creating the composition plan is set. This optimization period can be set to an arbitrary target period separately as needed by the planner. Here, the optimization period is set to three days through the plan creation period as an example.
[0106] (5) Setting a plan determination period (plan determination period setting unit 805 in Fig. 8, step S905 in Fig. 9)
A plan determination period for determining a composition plan is set. This plan determination period can be set to an arbitrary period separately as needed by the planner. For example, the plan determination period can be shortened in the first half of the plan creation period in which high accuracies are required in planning, and then the plan determination period can be lengthened in the second half of the plan creation period in which a rough plan will suffice, thereby enabling efficiënt plan creation with sufficiënt accuracies within a short time. Here the plan determination period is

set to one day as an example. In this case, it is determined for the first day through the plan creation period with respect to the composition plan obtained as a result of simulation based on a solution for a formula model.
[0107] (6) Formulating demand-supply balance restrictions of the composition plan into a formula model (demand-supply balance model building unit 806 in Fig. 8 (corresponding to the demand-supply balance model building unit 203 in Fig. 2), step S906 in Fig. 9)
Based on all or part of data obtained by the input data obtaining unit 801, demand-supply balance restrictions are formulated into formula models for the set optimization period by the set time accuracy. An example of the demand-supply balance model is shown by following Formula (15) to Formula (17). Note that an added word "plant" in each formula represents a steel plant, and "brand" represents a brand. For example, a use amount (plant, brand) means the use amount of each brand in each steel plant.
E use amount (plant, brand) = total use amount (plant) ... Formula (15)
Supply Min (brand) < E incoming amount (plant, brand) < supply Max (brand) ... Formula (16)
E use amount (plant, brand) = E incoming amount (plant, brand) ... Formula (17) [0108] Note that the above-described Formula (15)

to Formula (17) are just an example and may be replaced with another demand-supply balance model, or another demand-supply balance model may be added. For example, the operator may set a target composition ratio from factors with respect to purchase of various kinds of composition raw materials and the like, and may demand that a composition plan with a composition ratio close to the given target composition ratio is created. That is, when the composition ratio becomes largely different from the operator's assumption, it is expected that the assumed purchase amount is not satisfied, the purchase amount is exceeded, and/or it causes an unreasonable operation in the operation equipment. Thus, a demand-supply balance model may be set so as to output a composition ratio close to the composition ratio given as a target. [0109] Further, when the composition ratio on a previous day and the composition ratio on a following day separate largely, it may cause a difficulty in operation, Specifically, it causes an increase in setup time for using a different composition raw material and a failure in equipment. Accordingly, a demand-supply balance model may be set which does not cause a large separation between the composition ratio on a previous day and the composition ratio on a following day.
[0110] (7) Formulating property restrictions of the composition plan into a formula model (property model

building unit 807 in Fig. 8 including a 1inearization unit 807a (corresponding to the property model building unit 204 in Fig. 2), step S907 in Fig. 9)
Based on all or part of data obtained by the input data obtaining unit 801, property restrictions are formulated into a formula model for the set optimization period by the set time accuracy. For example, when a composition plan of coal is created, there are properties such as CSR (Coke Strength after ' Reaction), Dl (Drum Index), VM (Volatile Matter), and expansion pressure, and these properties have to satisfy required property restrictions. An example of the property model after mixing is illustrated by following Formula (18). Incidentally, an example having a lower limit value S is shown by Formula (18), but there may be a situation having an upper limit value and a situation having both the upper limit value and the lower limit value.
f(XA, XB, XC, ..., XN) > S ... Formula (18)
XA to XN : composition ratios of composition raw materials (brands) A to N
S: lower limit value (constant) [0111] Here, for many properties, the formula f(Xa, XB, XC, ..., Xjg) included in the property model often becomes linear with respect to a composition ratio as illustrated in following Formula (19).
f (XA, XB, XQ, . . . , X^)
= WA X composition ratio of A + WB x composition ratio of B + ... + WN x composition ratio of N ...

Formula (19)
Wa to WN : property with respect to a component of interest included in a brand i of each brand [0112] However, the formula f(XA, XB, XC, ■•■, XN ) representing a property may be nonlinear depending on the property. In this case, as described below, a linear formula f'(XA, XB, XC, •■•, XN) is introduced instead of the nonlinear formula f(XA, XB, XC, •••, XN) by the linearizat ion unit 807a to formulate the formula model.
[0113] Processing in the linearizat ion unit 807a will be described. When the formula f (XA, XB, XC, ..., XN) representing a certain property is nonlinear, the linear formula f'(XA, XB, XC, •••, XN) is introduced instead. For this linear formula f'(XA, XB, XC, ..., XN), there is considered one that forms the lower limit of the nonlinear formula f(XA, XB, XC, ••., XN), that is, one by which a relation of following Formula (20) holds true. Incidentally, Formula (20) does not always need to hold true, and just needs to hold true in a necessary range.
f (XA, XB, XC , • • • , XN) ^ f' (XA, XB, XC, • • • , XN ) • . •
Formula (20)
[0114] For example, as the linear formula f'(XA, XB, XC, . . . , XN) , a weighted average illustrated by following Formula (21) is considered. The weighted average is a value obtained such that a property when using 100% of a single brand is obtained from the nonlinear formula f(XA, XB, XC, ..., XN), a composition

ratio is multiplied therewith, and amounts of used brands are added up.
Weighted average (plant) = E[composition ratio (= use amount (plant, brand)/sum of use amounts (plant))
X property for 100% single brand] ... Formula (21) [0115] For the sake of simplicity in description, an example in which the composition ratio of brand A is 90% and the composition ratio of brand C is 10% is considered. In this case, the weighted average to be the linear formula f'(90, O, 10, ..., 0) is represented by the following formula. f' (90, O, 10, . . . , 0) = 0.9 X f(100, O, ..., 0) + 0.1 X f(0, O, 100, ... 0)
[0116] From past records or the like, when this weighted average satisfy Formula (20), it can be used as the linear formula f'(XA, XB, XC, •.•, XN). That is, when weighted average > S is adopted as a restriction, a possibility to formulate it with which Formula (18) holds true is obtained.
[0117] In the linearization unit 807a, as illustrated by following Formula (18)', a temporary lower limit value S' = S - s (s: offset value) smaller than the lower limit value S for the nonlinear formula f(XA^ XB, XC, .-w XN) is set as the lower limit value for the linear formula f' (XA, XB, Xc, ..., XN) and formulated into a formula model.
f' (XA, XB, XC , ..., XN) > S' ... Formula (18)' [0118] (8) Optimizing a composition plan formula

model based on an objective function (composition plan solution obtaining unit 809 in Fig. 8
(corresponding to the planning unit 205 in Fig. 2), Step S909 in Fig, 9)
The demand-supply balance model and the property model formed of the linear and integer restriction formulas built above make up a composition plan formula model. A problem is solved as an optimization problem by mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like based on an objective function set in advance, so as to calculate an optimal use amount.
[0119] Here, an example using a linear formula for the objective function is shown. An object of this embodiment is to minimize costs, and an example of objective function J is shown by following Formula
(22) .
J = E(cost (plant, brand) (per unit amount) x use amount (plant, brand)) -> minimization ... Formula
(22)
[0120] Incidentally, above Formula (22) is an example and may be replaced with another objective function, or another objective function may be added. For example, when it is required to approximate a composition plan to a composition ratio close to the given target composition ratio and to create a composition plan by which the composition ratio on a previous day and the composition ratio on a following

day does not separate largely, such an objective function may be set.
[0121] An optimal solution for the composition plan formula model combining the demand-supply balance model and the property model is obtained by solving the above formulated formulas (formula models) by mixed integer programming,
[0122] The above-described composition plan solution obtaining unit 809 (the planning unit 205) and Step S909 are examples of an optimization calculating unit and processing thereof mentioned in the present invention.
[0123] (9) Judging the result of obtaining a solution by the optimization calculation (solution obtaining result judging unit 810 in Fig. 8, Steps S910, S911 in Fig. 9)
Whether the result of obtaining a solution by the optimization calculation using Formula (18)'
satisfies the formula model f(XA, XB, XC, ..., XN) > S
including the nonlinear formula or not is judged.
When the formula model f(XA/ XB, XC, •••, XN) ^ S
including the nonlinear formula is satisfied as a result, this solution obtaining result is taken as a calculation instruction to a property simulator 812 described later for performing a simulation. When
the formula model f(XA, XB, XC, ..., XN) ^ S including the nonlinear formula is not satisfied, the formula
model f' (XA, XB, XC, • . ., XN ) ^ S' including the linear formula is adjusted (Step S911 in Fig. 9).

Specifically, the temporary lower limit value S' is increased slightly.
[0124] Fig. 11 is a flowchart illustrating processing of Steps S907 to S911, that is, processing when the linear formula f'(XA, XB, XC/ ■■■, XN) is introduced instead of the nonlinear formula f(XA, XB, Xc, . . . , XN) . In Step SllOl, the optimizat ion calculation is performed based on the demand-supply balance model, the property model (formulated by introducing the linear formula f'(Xa, XB, XC, •••, XN) instead of the nonlinear formula f(XA, XB, XC, ..., XN)), and the objective function J.
[0125] In this case, as illustrated in Formula
(18)', the temporary lower limit value S' = S - s (s: offset value) smaller than the lower limit value S for the nonlinear formula f(XA, XB, XC, ..., XN) is set as the lower limit value for the linear formula f'(XA, XB / Xc , . . . , XN ) .
[0126] Next, in Step S1102, it is judged whether a solution obtaining result by the optimization calculation using the formula model f'(XA, XB, Xc, ..., XN) > S' including the linear formula satisfies the
formula model f(XA, XB, XC, •••, XN) ^ S including the nonlinear formula or not. That is, the solution obtaining result (the use amounts (composition ratios) of the brands A to N) by the optimization calculation in Step S1101 is substituted in Formula (18), and it is judged whether Formula (18) holds true or not.

[0127] When Formula (18) holds true as a result of Step S1102, this processing is finished (proceeding to Step S912 in Fig. 9). Otherwise, when Formula (18) does not hold true, it proceeds to Step S1103, the teinporary lower limit value S' is increased slightly by an increase/decrease rate set in advance, and the processing of Step SllOl is performed again. Specifically, until Formula (18) holds true, a convergence calculation is performed such that the temporary lower limit value S' is increased slightly, and solution obtaining by the optimization calculation is repeated.
[0128] In this embodiment, incidentally, the description has been given with an example in which the property restrictions after mixing have a lower limit value, but the same applies to the case where they have an upper limit value. In this case, for the linear formula f' (XA, XB, XQ, • • •, XN) , one that forms an upper limit of the nonlinear formula f (XA, XB, Xc, ..•, XN) is considered. Further, in Step SllOl, a temporary upper limit value larger than the upper limit value for the nonlinear formula f(X^, XB, Xc, ••., XN) is set as the upper limit value for the linear formula f' (XA, XB, XC, ■ ■ ■ , XN) .
[0129] (10) Simulating a stock transition based on the obtained solution (stock transition simulator 811 in Fig. 8 (corresponding to the stock transition simulator 201 in Fig. 2), Step S912 in Fig. 9)
Based on the solution for the composition plan

formula model and all or part of data obtained by the input data obtaining unit 801, a simulation for all or part of a targeted composition for the set plan determination period is performed by a set plan creation accuracy. In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in actual operation without any problem. Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation,
[0130] Further, as an example of a restriction which is difficult to handle by a formula model, a setup time needed for setting up equipment when the composition ratio is changed or the like is taken into the simulation so as to perform the simulation precisely, and thus it is possible to set out a composition plan considering detailed restrictions required in actual operation.
[0131] (11) Simulating a property based on the obtained solution (property simulator 812 in Fig. 8 (corresponding to the property simulator 202 in Fig. 2), Step S913 in Fig. 9)
Based on the solution for the composition plan

formula model, the stock transition simulated by the stock transition simulator 811, and all or part of data obtained by the input data obtaining unit 801, a property is simulated for all or part of a targeted composition for the set plan determination period by a set plan creation accuracy, so as to obtain a property result of a composition raw material after mixing.
[0132] In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in actual operation without any problem. Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation.
[0133] The above-described stock transition simulator 811 (stock transition simulator 201) and Step S912, and the property simulator 812 (property simulator 202) and Step S913 are examples of a simulator and processing thereof mentioned in the present invention.
[0134] (12) Determining the composition plan
(determination unit 813 in Fig. 8, Step S914 in Fig. 9)

The composition plan derived by the stock transition simulation and the property simulation is determined for the set plan determination period. As shown in Fig. 7, since the composition determination period is set to one day in this embodiment, first one day of the created composition plan is determined. The plan for the portion of the created composition plan that is not within the plan determination period is discarded without being determined. [0135] (13) Judging whether the plan for the plan creation period or the plan determination period is determined (judgment unit 814 in Fig. 8, Step S915 in Fig. 9)
It is judged whether the plan determination period determined at this point is determined for the plan creation period set in advance. In this embodiment, since the plan creation period is ten days, the plan for the plan determination period is determined at the point when the plan is determined in a tenth loop. Accordingly, the composition plan for ten days is created at the point when determination of the plan is finished in the tenth loop, and the process is finished. [0136] (14) Updating the planning start date (updating unit 815 in Fig. 8, Step S916 in Fig. 9)
When the determined plan determination period is not determined for the plan creation period set in advance, the date and time just after the determined composition plan period in the composition plan is

set as a new planning start date. In this embodiment, as illustrated in Fig. 7, the planning start date which is initially first day and zero o'clock in the first loop is updated to second day and zero o'clock, and the planning start date which is initially second day and zero o'clock in the second loop is updated to third day and zero o'clock.
[0137] (15) Outputting the composition plan (output unit 816 of Fig. 8, Step S917 in Fig. 9)
The composition plan created as described above is displayed on the screen of the display unit 103 and/or is transmitted as data to a not-illustrated external apparatus by the output unit 816.
[0138] The output unit 816 and Step 3917 described above are examples of an output unit and processing thereof mentioned in the present invention.
[0139] As described above, corresponding to the current stock transition state, formula models for the demand-supply balance restrictions and the ^property restrictions are built first by the plan creation time accuracy for the predetermined optimization period, a solution for the built composition plan formula model is obtained based on the objective function, and the stock transition and the property after mixing are simulated based on the obtained solution. The composition plan obtained from the simulation result is determined for the set plan determination period, and the date and time just after the plan determination period are set to the

new planning start date and time. Thus, a series of processing for determining a composition plan for a new plan target period can be performed sequentially and repeatedly by a predetermined number of times, so as to create a composition plan for a desired plan creation period. Accordingly, a composition plan requiring an arbitrary time accuracy can be optimized quickly in detail, and can further be applied as it is in actual operation.
[0140] Moreover, when creating a composition plan using mathematical programming, the composition plan can be created even when the formula model representing property restrictions after mixing includes a nonlinear formula. Accordingly, a composition plan can be created while avoiding running out of stocks, satisfying required properties, and minimizing costs.
[0141] (Modification example of the second embodiment)
As shown in Fig. 12, the composition plan is created at constant periods (beginning,middle,and end of a month for example). Further, the property model may become nonlinear for plural properties a, p. Incidentally, in Fig. 12, "O" means that the property restriction is satisfied (Formula (18) holds true), and "X" means that the property restriction is not satisfied. That is, in the example of Fig. 12, nonconformity to the property a has occurred in plural periods (the beginning and the end of April),

and similarly nonconformity to the property β has occurred in plural periods (the beginning and the end of April).
[0142] In this case, a long time is taken for calculation processing when the convergence calculation described with Fig. 11 is performed separately for each period and for each property, specifically, the convergence calculation is performed for the property a and subsequently the convergence calculation is performed for the property
β in the beginning of April, and further the convergence calculation is performed for the property α and subsequently the convergence calculation is performed for the property β in the end of April. [0143] Accordingly, the convergence calculation described with Fig, 11 is performed at once for the target period and property. For example, the convergence calculation is performed at once for the properties α, β in the beginning and the end of April (temporary lower limit values of the properties α, β are slightly increased (or temporary upper limit values are slightly decreased) simultaneously in Step S1103 of Fig. 11), so as to increase the processing speed.
[0144] (Modification example of the second embodiment)
In the second embodiment, it is described that the temporary lower limit value S' is slightly ncreased (or the temporary upper limit value is

slightly decreased) in Step S1103 of Fig. 11, and thereafter the processing of Step s1101 is performed again. In this case, the formula models having no change in the convergence calculation, specifically, the demand-supply balance model of Formula (15) to Formula (17) and the property model which is originally linear are retained. Then, the processing speed can be increased by an arrangement such that when the temporary lower limit value is slightly increased (or the temporary upper limit value is slightly decreased) and the processing of Step SllOl is performed again, only the formula model that has a change in the convergence calculation, specifically, the formula model in which the temporary lower limit value is slightly increased (or the temporary upper limit value is slightly decreased) is changed. [0145] (Example)
Fig. 13 illustrates past records before applying the present invention (upper side) and a composition plan (lower side) by the composition plan creation method applying the present invention. As illustrated in the lower side of Fig. 13, it can be seen that the composition plan creation method applying the present invention enables to set out a composition plan satisfying the supply amount Min and the supply amount Max for each of brands A to N. Further, there is obtained a result that the used brands are reduced in each of steel plants a to e, contributing to cutting costs or the like.

[0146] -Third Embodiment-
When creating a composition plan, costs are determined as an important index, and it is required to minimize purchase costs and manufacturing costs, as well as shipping costs for shipping raw materials, and so on. Here, many raw materials in steel industry are purchased from mines in foreign countries, and hence the raw materials are transported by ship. Accordingly, freights which are expenses required for transporting raw materials by ship are main shipping costs. Furthermore, regarding the shipped raw materials, it is required to plan a composition over plural days while changing composition ratios for avoiding running out of stocks of the composition raw materials in a steel plant
(which is also simply referred to as a "plant" alternatively) which is an unloading port to discharge the raw materials.
[0147] To achieve a composition plan that satisfies the above requirements, it is necessary to grasp vast amounts of Information such as various properties of purchased various kinds of composition raw materials, stock Information of various kinds of composition raw materials, purchase costs, shipping costs, and so on, Accordingly, to create the composition plan manually, all of these vast amounts of Information should be grasped to determine a composition, which requires a considerable amount of time. Further, accompanying low response to changes of Information, there also

occurs a problem such that properties after mixing do not fall within a required range in reality. [0148] Here, a purchase plan of raw materials and a ship allocation plan for shipping raw materials are generally created based on a composition plan. However, when the composition plan is created without considering shipping costs, there is a risk of creating a composition plan to use raw materials with high shipping costs. In this case, it is difficult to lower the shipping costs no matter how the shipping is thought out. For example, when there are raw materials X, Y with almost equal qualities and it is possible to use either of the raw materials X, Y in unloading ports (steel plants) A, B, if the cost to ship the raw material X to the unloading port A is $20/ton, the cost to ship the raw material Y thereto is $40/ton, the cost to ship the raw material X to the unloading port B is $40/ton, and the cost to ship the raw material Y thereto is $20/ton, there is a risk to set out a plan to use the raw material Y in the unloading port A and the raw material X in the unloading port B when the shipping cost is not considered, even though originally it is better in view of shipping cost to set out a plan to use the raw material X in the unloading port A and the raw material Y in the unloading port B.
[0149] Accordingly, when considering costs, not only purchase costs of composition raw materials but also shipping costs thereof should be considered. An

object of a third embodiment is to enable creation of a composition plan over plural days while satisfying requirements of demand-supply balances of composition raw materials and properties after mixing and suppressing costs including shipping costs, when creating a composition plan to receive and mix plural kinds of composition raw materials. [0150] A system structure example including a composition plan creating apparatus according to this embodiment and a basic structure of a composition plan creating apparatus 100 are the same as those illustrated in Figs. 1, 2, and will be described here also with reference to Figs. 1, 2.
[0151] As illustrated in Fig. 1, in the composition plan creating apparatus 100, when creating a composition plan, data of restriction conditions and prerequisites including a plan creation period, an arrival schedule of composition raw materials including incoming amounts by a ship allocation plan, stock conditions of composition raw materials, properties of composition raw materials (characteristics (including qualities), conditions, and so on), cost Information (purchase cost Information representing unit costs of composition raw materials, and shipping cost Information when using a ship), which are needed for setting out a composition plan, are set by an operator or obtained from a process computer 105 or a business computer 106.

[0152] The composition plan creating apparatus 100 creates a mixing plan for receiving and mixing various kinds (plural brands) of composition raw materials by performing a simulation, and obtains use amounts (composition ratios) and incoming amounts of respective brands as a composition plan so as to satisfy demand-supply balance restrictions of composition raw materials and property restrictions after mixing. The composition plan creating apparatus 100 attempts to optimize the composition plan by building a formula model representing the demand-supply balance restrictions of composition raw materials (also referred to as a "demand-supply balance model") and a formula model representing the property restrictions after mixing (also referred to as a "property model") using mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like, details of which will be described later. [0153] A display unit 103 displays use amounts of respective brands (composition ratios) and incoming amounts obtained by the composition plan creating apparatus 100, a stock transition graph, and various forms.
[0154] An operator evaluation unit 104 enables the operator to evaluate an obtained composition plan from various aspects (for example, stock transition, property, and so on), and modify a composition ratio and so on as necessary if the result is

unsatisfactory, When this happens, the operator changes the weight of an objective function or an index of evaluation, and/or changes a target period and a plan determination period for building the formula models as necessary. Further, the unit is able to reflect the operator's intention such as fixing all use amounts or a use amount of specified processing, and the like. Then the composition plan is created again in the composition plan creating apparatus 100 .
[0155] As illustrated in Fig. 2, the composition plan creating apparatus 100 is structured including a simulator (a stock transition simulator 201 and a property simulator 202), a model building unit (a demand-supply balance model building unit 203 and a property model building unit 204), and a planning unit 205 functioning as an optimization calculating unit, and further has input and output units. [0156] The stock transition simulator 201 is a simulator which calculates the demand-supply condition (stock transition) of each composition raw material. The property simulator 202 is a simulator which calculates the property after mixing a composition raw material. By the stock transition simulator 201 and the property simulator 202 operating in conjunction with each other, the stock transition of a composition raw material and the property after mixing are calculated. [0157] In this embodiment, based on input data 206

of a plan creation period, an arrival schedule of composition raw materials including incoming amounts by a ship allocation plan, stock conditions of composition raw materials, properties of composition raw materials, purchase cost Information representing unit costs of composition raw materials, shipping costs when using a ship, and so on, which are needed for setting out a composition plan, the formula model representing the demand-supply balance restrictions (stock restrictions) is built by the demand-supply balance model building unit 203 and the formula model representing the property restrictions is built by the property model building unit 204, with an optimization period set in advance from a planning start date and time of the composition plan being a target, based on a time accuracy set in advance, and conforming to mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like. [0158] Using the formula models built by the demand-supply balance model building unit 203 and the property model building unit 204, the planning unit 205 performs an optimization calculation so as to create a composition plan while avoiding running out of stocks, satisfying required properties, and minimizing costs (purchase costs and shipping costs of composition raw materials), and calculates a calculation instruction to the stock transition simulator 201 and the property simulator 202. Upon

receipt ot this caiculation instruction, the stock transition simulator 201 simulates the stock transition, and the property simulator 202 simulates the property.
[0159] With the composition plan creating apparatus according to this embodiment, the caiculation instruction is not performed based on predetermined rules as is conventional. The caiculation instruction based on a result of the optimization caiculation performed by the planning unit 205 is outputted to the stock transition simulator 201 and the property simulator 202. Thus, it becomes possible to securely perform an optimal caiculation instruction corresponding to an event on a certain occasion.
[0160] Further, for example, when a simulation for a plan determination period set in advance as illustrated in Fig. 7 by the stock transition simulator 201 and the property simulator 202 is completed, the formula model representing the stock restrictions is built by the demand-supply balance model building unit 203 based on Information of stock transition and property in a final state of the plan determination period, and the formula model representing the property restrictions is built by the property model building unit 204, which are given to the planning unit 205. When given Information of the stock transition and the property, the planning unit 205 performs the optimization caiculation.

[0161] As described above, an optimal composition plan can be created by carrying out a detailed simulation with the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the property model building unit 204), and the planning unit 205 operating in conjunction. That is, the simulation performed in this embodiment is not based on predetermined rules as is conventional but is performed based on results of the optimization calculation, and thus a logical optimal solution can be obtained securely by just carrying out one time of simulation. Accordingly, it is not necessary to perform simulations again and again repeatedly after evaluating a simulation result as is conventional, and a simulation result 207 can be created quickly with high accuracy. Therefore, it is sufficiently possible to create a plan within a practical time period even when the target of creating a composition plan is large in scale. [0162] Further, when the plan creation period becomes long, the period to be considered becomes long, and there occurs a problem that it becomes not possible to obtain a solution because the scale of problem becomes large by a conventional method. However, with this approach, it is possible to reduce the scale of problem by dividing into optimization periods, and thus it is possible to solve the problem even when the plan creation period becomes long. The

simulation result 207 obtained as described above is outputted as a composition plan.
[0163] Further, when the scale of a model built by the demand-supply balance model building unit 203 or the property model building unit 204 is quite large or when there are great many complicated restriction conditions, the demand-supply balance model building unit 203 and the property model building unit 204 are configured to obtain only an important part having a large influence on creation of composition plan out of the demand-supply balance restrictions and the property restrictions described in the stock transition simulator 201 and the property simulator 202. Thus, the scales of formula models of the demand-supply balance model building unit 203 and the property model building unit 204 come within an appropriate range, and the optimization calculation can be performed within a practical time period. The stock transition simulator 201 and the property simulator 202 are capable of describing all demand-supply balance restrictions and property restrictions which should be considered, and thus it is assured that the composition plan created by performing one time of simulation is practicable in reality.
[0164] As described above, in this embodiment, the composition plan is created by operating the simulator (the stock transition simulator 201 and the property simulator 202), the model building unit (the demand-supply balance model building unit 203 and the

property model building unit 204), and the planning unit 205 in conjunction. Thus, (1) the composition plan can be created without performing the simulation repeatedly. (2) The calculation time can be reduced by allowing the planning unit 205 to obtain only an important part having a large influence on creation of composition plan, and (3) it becomes possible to solve a large-scale problem.
[0165] Hereinafter, with reference to Figs. 14 to 20, a structure of the composition plan creating apparatus 100 according to this embodiment and steps of a composition plan creating method performed using this apparatus 100 will be described in detail, Fig, 14 is a diagram illustrating a detailed structure of the composition plan creating apparatus 100 with respect to the basic structure of the composition plan creating apparatus 100 described using Fig. 2. Further, Fig. 15 is a flowchart illustrating the steps of the composition plan creating method performed using this apparatus 100.
[0166] To describe an overview of composition plan creation, as illustrated in Fig. 10 for example, use amounts (composition ratios (proportions)) and incoming amounts of brands A to N in each of plants a to c are determined as a composition plan so as to achieve demand-supply balances of composition raw materials (brands) in the plural plants (unloading ports) a to c (such as avoiding running out of stocks of the brands A to N), satisfy required properties,

and minimize costs (purchase costs and shipping costs of composition raw materials). Here, the scheduled use amounts as sums of use amounts are given as input data for each plant, and thus a composition ratio (%) = use amount/scheduled used amount x 100 holds true. Accordingly, when one of the use amount and the composition ratio is determined, the other is also determined.
[0167] (1) Obtaining input data (input data obtaining unit 1401 in Fig. 14, Step S1501 in Fig. 15)
Information needed in this processing (arrival schedules of composition raw materials including incoming amounts by a ship allocation plan, stock conditions of composition raw materials, properties of composition raw materials, cost Information, and so on) is read online, and the operator makes a modification to them as necessary.
[0168] Here, the arrival schedules of composition raw materials obtained by the input data obtaining unit 1401 include scheduled incoming amounts by a receipt plan based on a target receipt amount, and Information representing incoming amounts of a ship allocation plan (planned regarding items including loading port, date and time of arrival at a loading port, loaded brand, loaded amount, unloading port, date and time of arrival at an unloading port, unloaded brand, and unloaded amount of each ship) if it is already created. Here, the target receipt

amount is Information representing target receipt amounts (scheduled receipt amounts) by mine (loading place) and by brand. Each mine is contracted about the amount of receipt by brand annually for example, and the target receipt amount per month can be obtained by dividing this amount by the number of months. It is required to receive an amount close to this target receipt amount, but a variation to an extent of a few tens of thousand tons more or less than this amount per year is within a tolerable range in negotiation with each mine. Further, a contract not to receive a predetermined brand in a predetermined period is also conceivable depending on the contract, and such Information may be included. For example, when the target receipt amount of raw material A in a certain month is 50,000 tons, a variation more or less than the target receipt amount per year is 60,000 tons (5,000 tons per month), and the ship to transport this raw material is not decided in this month, the scheduled incoming amount in this month has an upper limit (scheduled incoming amount upper limit) 50,000 tons + 5,000 tons and a lower limit (scheduled incoming amount lower limit) 50,000 tons - 5,000 tons. When a period as a target of setting out a composition plan is coming close, the ship to transport this raw material is often decided already. For example, when it is decided to receive 30,000 tons of the raw material by ship X and 22,000 tons by ship Y in this month, the scheduled

incoming amount is 30,000 tons + 22,000 tons = 52,000 tons .
[0169] In a ship allocation plan as illustrated in Fig. 16 for example, operation schedules of ships listed on a ship list as illustrated in Fig. 17 are arranged. Ships used for transporting composition raw materials include continuous voyage ships (continuous ships), irregular ships, and spot ships (Spot). The continuous ships are ships contracted for making continuous voyages during a contract period. The irregular ships are ships contracted for making a number of voyages in a contract period or making a voyage only in a contract period of voyage. The spot ships are ships normally not contracted but can be requested for making a voyage on the spot basis. For the continuous ships, there are described chartered ship code, contract segment, contract period (start date and finish date), maximum load capacity, and ship name. For the irregular ships, there are described chartered ship code, contract segment, contract period (start date and finish date), annual contracted number (the number of voyages contracted to be allocated annually), or scheduled allocation year and month (contract year and month of voyage to allocate a ship), maximum load capacity, and ship name. These continuous ships and irregular ships are listed separately, but the spot ships are listed by name of region where a ship is able to voyage and the size of a ship, and there are

described chartered ship code (describing a name of region and size), contract segment, and maximum load capacity. Incidentally, Pmax representing the ship type of the spot ships means a ship able to pass the Panama Canal (this type of ship is generally called Panamax), Cape means a ship able to pass the Cape of Good Hope and the Cape Horn (this type of ship is generally called Capesize), and VL (Very Large) means a large ship. Here, Panamax normally refers to a ship having a length not more than 900 feet and a width not more than 106 feet with a maximum load capacity of 60,000 to 80,000 ton class. Further, Capesize generally refers to a ship with a maximum load capacity of 150,000 to 170,000 ton class. In the ship allocation plan, a plan is set out for items including loading port, date and time of arrival at a loading port, loaded brand, loaded amount, unloading port, date and time of arrival at an unloading port, an unloaded brand, and unloaded amount for the ships listed on the ship list.
[0170] For example, in the ship allocation plan illustrated in Fig. 16, on the voyage No. 1 of the continuous ship A, the ship arrivés at the offing of a loading port (X1 port) at 21 o'clock on November 19, 2007, gets to the berth denoted by code "1" in the loading port (X1 port) at 21 o'clock on December 13, 2007, and leaves the loading port (XI port) at 21 o'clock on December 14, 2007. At this time, 40,000 t of raw material brand A is loaded, and 35,000 t of

brand B is loaded. Thereafter, the ship voyaged for 16,980 minutes, arrivés at the offing of the unloading port (A port) at 16 o'clock on December 26, 2007, gets to the berth denoted by code "4" in the unloading port (A port) at one o'clock on December 27, 2007, and leaves the unloading port (A port) at 16 o'clock on December 28, 2007. At this time, 25,000 t of the raw material brand A and 15,000 t of the brand B are unloaded. Thereafter, the ship voyaged for 3,060 minutes, arrivés at the offing of the unloading port (B port) at 19 o'clock on December 30, 2007, gets to the berth denoted by code "13" in the unloading port (B port) at 19 o'clock on December 30, 2007, and leaves the unloading port (B port) at 23 o'clock on January 1, 2008. At this time, 15,000 t of the raw material brand A and 20,000 t of the brand B are unloaded.
[0171] The stock conditions of composition raw materials are Information representing stock amounts
(m tons) by plant and by brand on the first day of the plan creation period. The properties of composition raw materials are Information representing properties such as ingredients of each composition raw material. For example, as properties of iron ore which is composition raw material, property Information such as Fe2O3, Fe304, Si02, AI2O3, and the like is included.
[0172] The purchase cost Information of composition raw materials is Information representing unit values

(dollar/ton) of composition raw materials by mine (loading place) and by brand.
[0173] The shipping cost Information when using a ship includes Information representing freights when using ships listed on the ship list. Fig. 18 illustrates an example of a table of freights by ship (chartered ship), by loading port (loading place), and by unloading port (unloading place). As illustrated in this diagram, chartered ship code, loading port, 1 unloading port, 2 unloading port, 3 unloading port, and freight (dollar/ton) are described for each ship listed on the ship list. For example, the freight of the continuous ship A to voyage from the loading port XI to the unloading port A is 16.00, and the freight to voyage from the loading port X1 to the unloading ports A, B is 16.24. Incidentally, as can be seen from the freight list, using a continuous ship is generally cheaper in freight than using an irregular ship or spot ship. [0174] Further, the shipping cost information when using a ship also includes information representing deemed freights by brand and by unloading port (unloading place). Originally, shipping costs are determined uniquely by the above-described freights by ship, by loading port, and by unloading port. However, regarding reception, a ship for loading a raw material is normally determined from about a few weeks to one month before reception when it is a raw material shipped from Australia, or two to three

months from Brazil. When making a long-term plan such as a composition plan for one year, a ship for loading a raw material is normally undetermined for a ship in the three months ahead at the point of making the composition plan. The freights by brand and by unloading port are needed for estimating shipping costs for such raw materials for which a loading ship is not determined. Here, the freights by brand and by unloading port cannot be determined uniquely because originally the freights are different depending on the choice of ship for loading raw materials, and the like. Accordingly, instead of the freights by brand and by unloading port, the Information of deemed freights by brand and by unloading port which are estimated values is obtained. As the deemed freights by brand and by unloading port, for example, there are listed in advance freights by brand and by unloading port set based on experience and the like, or deemed freights by brand and by unloading port obtained by a statistic method from past records, for example, by collecting freights of past records by brand and by unloading port and deeming average values thereof as freights by brand and by unloading port.
[0175] The input data obtaining unit 1401 and Step S1501 described above are examples of a data obtaining unit and processing thereof mentioned in the present invention. [0176] (2) Setting a composition plan creation

period (plan creation period setting unit 1402 in Fig. 14, Step S1502 in Fig. 15)
A period to create a composition plan is set. This creation period can be set to an arbitrary period as needed by the planner. Here, a plan for ten days is made as an example.
[0177] (3) Setting a composition plan creation time accuracy (time accuracy setting unit 1403 in Fig. 14, Step S1503 in Fig. 15)
A time accuracy and a simulation accuracy for creating the composition plan are set. These time accuracy and simulation accuracy can be set to arbitrary accuracies separately as needed by the planner. For example, the accuracies can be increased in the first half of the plan creation period in which high accuracies are required in planning, and then the accuracies can be decreased in the second half of the plan creation period in which a rough plan will suffice, thereby enabling efficiënt plan creation with sufficiënt accuracies within a short time.
[0178] (4) Setting an optimization period (optimization period setting unit 1404 in Fig. 14, Step S1504 in Fig. 15)
An optimization period for creating the composition plan is set. This optimization period can be set to an arbitrary target period separately as needed by the planner. Here, the optimization period is set to three days through the plan creation

period as an example.
[0179] (5) Setting a plan determination period (plan determination period setting unit 1405 in Fig. 14, step S1505 in Fig. 15)
A plan determination period for determining a composition plan is set. This plan determination period can be set to an arbitrary period separately as needed by the planner. For example, the plan determination period can be shortened in the first half of the plan creation period in which high accuracies are required in planning, and then the plan determination period can be lengthened in the second half of the plan creation period in which a rough plan will suffice, thereby enabling efficiënt plan creation with sufficiënt accuracies within a short time. Here the plan determination period is set to one day as an example. In this case, it is determined for the first day through the plan creation period with respect to the composition plan obtamed as a result of simulation based on a solution for a formula model
[0180] (6) Formulating demand-supply balance restrictions of the composition plan into a formula model (demand-supply balance model building unit 1406 in Fig. 14 (corresponding to the demand-supply balance model building unit 203 in Fig. 2), step S1506 in Fig. 15)
Based on all or part of data obtained by the input data obtaining unit 1401, demand-supply balance

restrictions are formulated into formula models for the set optimization period by the set time accuracy,
[0181] A variable representing a use amount of each brand is defined as shown by following Formula (23). Further, a variable representing a stock amount of each brand is defined as shown by following Formula
(24). Furthermore, a variable representing an incoming amount of each brand is defined as shown by following Formula (25).
[0182]
Xp,i,d : use amount of brand i on day d in plant p ... Formula (23)
[0183]
Sp,i,d : stock amount of brand i on day d in plant p ... Formula (24)
[0184]
yp,i,d : incoming amount of brand i on day d in plant p ... Formula (25)
[0185] A formula model built based on demand-supply Information, that is, a demand-supply balance restriction model is shown below. It is required that the stock amount of each brand is not less than a constant value called a safe stock amount. The restriction in this case is represented by following Formula (26).
[0186]
Sp,i,d ^ Saf eStockp, i,d : Saf eSt ockp, i, d is a safe stock amount (constant) of brand i on day d in plant p. . . Formula (26)

[0187] Further, the stock amount of each brand is determined from the stock amount on a previous day, the incoming amount on the previous day, and the use amount on the previous day. A restriction formula representing the relation in this case is represented by following Formula (27). That is, the stock amount of a current day is a value obtained by subtracting the use amount on the current day from a value obtained by adding the stock amount of the previous day and the amount coming in (unloaded) on the current day. [0188]
Sp,i,d = Sp,i,d-1 + yp,i,d-1 - Xp,i,d-1 ... Formula (27) [0189] Further, it is required that the sum of use amounts of each brand on a certain day matches a scheduled use amount with respect to the sum of all brands on this day. A restriction formula representing a relation in this case is represented by following Formula (28) [0190]
' : Up,d is use amount (constant) on day d
in plant p ... Formula (28)
[0191] Further, the operator sets a target composition ratio from factors with respect to purchase of various kinds of raw materials and the like, and demands that a composition plan with a composition ratio close to the given target composition ratio is created. That is, when the

composition ratio becomes largely different from the operator's assumption, it is expected that the assumed purchase amount is not satisfied, the purchase amount is exceeded, and/or it causes an unreasonable operation in the operation equipment. Thus, it is necessary to output a composition ratio close to the composition ratio given as a target. Restrictions to achieve the aforementioned function are shown below. That is, a value obtained by subtracting a target use amount (targeted composition ratio) (constant) from a use amount of a brand is defined as a variable of an overuse amount from the target use amount. Here, the closer the use amount and the target use amount, the better the plan is. Therefore, this overuse amount is the smaller the better. Due to this reason, this overuse amount is added as an item of the objective function as will be described later and is minimized. Similarly, a value obtained by subtracting a use amount from a target ,use amount of a brand is defined as a variable of deficiënt amount from the target use amount. Here, the closer the use amount and the target use amount, the better the plan is. Therefore, this deficiënt amount is the smaller the better. Due to this reason, this deficiency is added as an item of the objective function as will be described later and is minimized. In this case, a restriction formula representing the relation of use amount, target use amount, overuse amount, and deficiënt amount of each brand is

represented by following Formula (29). That is, when subtracting the overuse amount from or adding the deficiënt amount to the use amount, the result matches the target use amount. [0192]
Xp,i,d - OverUp,i,d + CoverUp,i,d = xTargetp,i,d ••• Formula (29)
xTargetp,i,d is the target use amount (constant) of brand i on day d in plant p.
OverUp,i,d is an overuse amount from the target use amount of brand i on day d in plant p.
CoverUp,i,d is a deficiënt amount from the target use amount of brand i on day d in plant p. [0193] Further, when the composition ratio on a previous day and the composition ratio on a following day separate largely, it causes a difficulty in operation. Specifically, it causes an increase in setup time for using a different raw material and a failure in equipment. Accordingly, it is required to create a composition plan which does not cause a large separation between the composition ratio on a previous day and the composition ratio on a following day. To achieve this function, a variable representing an upper limit amount of the difference between the use amount of a brand on a day of interest and the use amount on a previous day is defined as shown by following Formula (30). [0194]
Diffp,i,d : difference between a use amount of

brand i on day d and a use amount of brand i on day d-1 in plant p ... Formula (30)
[0195] A restriction to achieve the above function using this variable is shown bellow. Specifically, the value obtained by subtracting the use amount of a brand on the previous day of a day of interest from the use amount on the day of interest is not more than the difference between the use amount on the day of interest and the use amount on the previous day of the day of interest. Here, the closer the use amount on the day of interest and the use amount on the previous day of the day of interest, the better the plan is. Therefore, this difference in use amount is the smaller the better. Due to this reason, this difference in use amount is added as an item of the objective function as will be described later and is minimized. Similarly, also the value obtained by subtracting the use amount of a brand on the day of interest from the use amount on the previous day of the day of interest is also formulated as a restriction formula. [0196]
Xp,i,d ~ Xp,i,d-1 ≤ Diffp,i,d
Xp,i,d-i - Xp,i,d ≥ Diffp,i,d ... Formula (31) Diffp,i,d is a difference between a use amount of brand i on day d and a use amount of brand i on day d-1 in plant p.
[0197] Further, it is required that the incoming amount of each brand comes within an amount range

given as a scheduled incoming amount. A restriction formula representing the relation in this case is represented by following Formula (32) and Formula
(33). That is, it is required that the sum of incoming amounts in a month of interest is not more than the upper limit of a scheduled incoming amount of the month of interest and not less than the lower limit of the scheduled incoming amount.
[0198]

yLoweri,m is the lower limit of a scheduled incoming amount (constant) of brand i in month m, {didate included in month m} ... Formula (32)

yUpperi,in is the upper limit of a scheduled incoming amount (constant) of brand i in month m, {didate included in month m} ... Formula (33) [0199] Note that the above-described demand-supply balance restrictions are just an example and may be replaced with other restrictions, or other restrictions may be added.
[0200] (7) Formulating property restrictions of the composition plan into a formula model (property model building unit 1407 in Fig. 14 including a linearizat ion unit 1407a (corresponding to the property model building unit 204 in Fig. 2), step S1507, S1507a in Fig. 15)

Based on all or part of data obtained by the input data obtaining unit 1401, property restrictions are formulated into a formula model for the set optimization period by the set time accuracy. For example, when a composition plan of iron ore is created, there are properties such as iron, Si02, A12O3, and the like, and when a composition plan of coal is created, there are properties such as CSR (Coke Strength after Reaction), Dl (Drum Index), VM (Volatile Matter), and expansion pressure, and these properties have to satisfy required property restrictions. An example of the property model after mixing is illustrated by aforementioned Formula (18). [0201] Here, for many properties, the formula f(XA, XB, XC, ..., XN) included in the property model often becomes linear with respect to a composition ratio as illustrated in aforementioned Formula (19). For example, when Si02 is mixed such that the composition ratio of brand A is 40% and the Si02 component is 1%, or the composition ratio of brand A is 60% and the Si02 component is 2%, the property with respect to the Si02 component after mixing is 1 x 0.4 + 2 x 0.6 = 1.6%.
[0202] However, the formula f(XA, XB, XC, ..., XN ) representing a property may be nonlinear depending on the property. In this case, as described below, a linear formula f'(XA, XB, XQ, •••, XN) is introduced instead of the nonlinear formula f(XA, XB, XC, ..., XN) by the linearization unit 1407a to formulate the

formula model.
[0203] Processing in the linearization unit 1407a will be described. When the formula f(XA, XB, XC, ..., XN) representing a certain property is nonlinear, the linear formula f'(X^, XB, XC/ ..., XN) is introduced instead. For this linear formula f'(Xa, XB, XC, ..•, XN), there is considered one that forms the lower limit of the nonlinear formula f{XA, XB, XQ, ••., XN), that is, one by which a relation of Formula (20) holds true.
[0204] For example, as the linear formula f'(Xa, XB, Xc, • . • , XN) , a weighted average illustrated by Formula (21) is considered. The weighted average is a value obtained such that a property when using 100% of a single brand is obtained from the nonlinear formula f(XA, XB, XC, ••., XN), a composition ratio is multiplied therewith, and amounts of used brands are added up.
[0205] For the sake of simplicity in description, an example in which the composition ratio of brand A is 90% and the composition ratio of brand C is 10% is considered. In this case, the weighted average to be the linear formula f'(90, O, 10, ..., 0) is represented by the following formula. f' (90, O, 10, . . . , 0) = 0.9 X f(100, O, ..., 0) + 0.1 X f(0, O, 100, ... 0)
[0206] From past records or the like, when this weighted average satisfy Formula (20), it can be used

as the linear formula f' (XA, XB, XC, • • •/ XN) . That is, when weighted average > S is adopted as a restriction, a possibility to formulate it with which Formula (18) holds true is obtained.
[0207] In the linearization unit 1407a, as illustrated by Formula (18)', a temporary lower limit value S' = S - s (s: offset value) smaller than the lower limit value S for the nonlinear formula f (XA, XB, Xc, ••■, XN) is set as the lower limit value for the linear formula f' (Xa, XB, XQ, . • • , ^N) and formulated into a formula model.
[0208] In the foregoing, the description is given with an example in which the property restrictions after mixing have a lower limit value. Note that the above-described property restrictions are just an example and may be replaced with other restrictions, or other restrictions may be added (including the case where the property restrictions after mixing have an upper limit value).
[0209] The demand-supply balance model building unit 1406 (demand-supply balance model building unit 203) and Step S1506, and the property model building unit 1407 (property model building unit 204) and Steps S1507, S1507a described above are examples of a model building unit and processing thereof mentioned in the present invention.
[0210] (8) Fixation extraction processing (fixation extraction processing unit 1408 in Fig. 14, Step S1508 in Fig. 15)

As illustrated in Fig. 19, fixed one that is unchangeable one is extracted from loading port, loaded brand, loaded amount, unloading port, unloaded brand, and unloaded amount as items of the ship allocation plan. When "loading port, loaded brand, loaded amount, unloading port", "loading port, loaded brand, loaded amount, unloading port, unloaded brand", and "loading port, loaded brand, loaded amount, unloading port, unloaded brand, unloaded amount" are fixed, the freights by ship, by loading port, and by unloading port (see Fig. 18) are used. That is, regarding the shipping cost by a ship, the freight is decided at the time when items up to the unloading port are decided (fixed). Thus, regarding the above three patterns, the freights by ship, by loading port, and by unloading port are used to enable a precise shipping cost calculation when a ship to load raw materials is decided.
[0211] Further, when none of them is fixed, or only the "loading port", "loading port, loaded brand", "loading port, loaded brand, loaded amount" are fixed, the deemed freights by brand and by unloading port are used. That is, when the unloading port is not decided, it is possible to change to an unloading port for which the shipping cost is lower by changing the unloading port with respect to the ship of interest. In this case, with respect to this ship, using the deemed freights by brand and by unloading port, changing from the unloading port of this ship

to an unloading port for which a shipping cost is cheaper than the unloading port is planned by optimization described later. This enables to create a plan with a lower shipping cost. Incidentally, regarding a same chartered ship, the state of a record which is fixed most unlikely is considered as a fixed status of this chartered ship. This fixation extraction processing is not necessarily at the timing illustrated in Fig. 15, and may be performed for example when starting the composition plan creation.
[0212] The above function enables to set out a highly precise composition plan considering the operation status of a ship, specifically, items that can be changed and items that cannot be changed with respect to this ship among loading port, loaded brand, loaded amount, unloading port, unloaded brand, and unloaded amount (normally, when the date and time of arrival of a raw material is far away, the ship to load the raw material can be changed, and it becomes unchangeable as the date and time get closer). [0213] (9) Optimizing a composition plan formula model based on an objective function (composition plan solution obtaining unit 1409 in Fig. 14 (corresponding to the planning unit 205 in Fig. 2), Step S1509 in Fig. 15)
The demand-supply balance model and the property model formed of the linear and integer restriction formulas built above make up a composition plan

formula model. A problem is solved as an optimization problem by mathematical programming such as LP (linear programming), MIP (mixed integer programming), QP (quadratic programming), and the like based on an objective function set in advance, so as to calculate an optimal use amount and incoming amount.
[0214] Here, an example using a linear formula for the objective function is shown. An object of this embodiment is to minimize costs (purchase costs and shipping costs of composition raw materials), and an example of objective function J is shown by following Formula (34). When obtaining a solution using the objective function, purchase cost Information and shipping cost Information set in Step S1508 are used.
[0215]

... Formula (34)
Vi is purchase unit cost (constant) of brand i.
Fs,i,p is freight (constant) of ship s, brand i, and plant (unloading port) p.
ys,p,i,d is incoming amount (variable) of brand i on day d in plant p by ship s in ys,p,i,d
Fi,p is deemed freight (constant) of brand i, and plant (unloading port) p.
[0216] Incidentally, Formula (34) is an example of the objective function and may be replaced with

another objective function, or another objective function may be added.
[0217] For example, when it is required to approximate a composition plan to a composition ratio close to the given target composition ratio and further to create a composition plan by which the composition ratio on a previous day and the composition ratio on a following day does not separate largely, an item to minimize the overuse amount from the target use amount, the deficiënt amount, and the difference between the use amount on a day of interest and the use amount on the previous day of the day of interest is added to the objective function, as shown by following Formula (35). [0218]

... Formula (35)
Woveruse: degree (constant) of penalty with respect to the overuse amount
Wdeficient: degree (constant) of penalty with respect to the deficiënt amount
Wdifference: degree (constant) of penalty with respect to the difference in use amount [0219] An optimal solution for the composition plan formula model combining the demand-supply balance

model and the property model is obtained by solving the above formulated formulas (formula models) by mixed integer programming. That is, as described with the above-described items (6) to (9), a formula to be minimized is formulated as an objective function, and each formula to be satisfied is formulated as a restriction formula. This restriction formula is represented by a linear equation or inequality, and the formula model as a model in which an objective function is represented by a linear formula and the objective function are built. A problem formulated in this manner is generally well known as a linear programming problem, and this problem can be optimized. [0220] The above-described composition plan solution obtaining unit 1409 (the planning unit 205) and Step S1509 are examples of an optimization calculating unit and processing thereof mentioned in the present invention.
[0221] (10) Judging the result of obtaining a solution by the optimization calculation (solution obtaining result judging unit 1410 in Fig. 14, Steps S1510, S1511 in Fig. 15)
Whether the result of obtaining a solution by the optimization calculation using Formula (18)'
satisfies the formula model f(XA, XB, XC, '••, XN) ^ S including the nonlinear formula or not is judged.
When the formula model f (XA, XB, XQ, ••., XN) ≥ S including the nonlinear formula is satisfied as a

result, this solution obtaining result is taken as a calculation instruction to a property simulator 1412 described later for performing a simulation. When
the formula model f(XA, XB, XQ, •, XN) ≥ S including the nonlinear formula is not satisfied, the formula
model f' (XA, XB, XC, • . ., XN) > S' including the linear formula is adjusted (Step S1511 in Fig. 15). Specifically, the temporary lower limit value S' is increased slightly.
[0222] Fig. 20 is a flowchart illustrating processing of Steps S1507 to S1510, that is, processing when the linear formula f' (Xa, XB, XC, • • • , XN) is introduced instead of the nonlinear formula f(XA, XB, XC, ..., XN). In Step S2001, the optimization calculation is performed based on the demand-supply balance model, the property model
(formulated by introducing the linear formula f' (XA, XB, XC, ..., XN ) instead of the nonlinear formula f(XA, XB, XC, . • ., XN) ) , and the objective function J.
[0223] In this case, as illustrated in Formula
(18)', the temporary lower limit value S' = S - s (s: offset value) smaller than the lower limit value S for the nonlinear formula f(XA, XB, XC, ..•, XN) is set as the lower limit value for the linear formula f'(XA, XB , Xc, • • • , XN) .
[0224] Next, in Step S2002, it is judged whether a solution obtaining result by the optimization calculation using the formula model f' (XA, XB, Xc, • • • , XN)≥ S' including the linear formula satisfies the

formula model f(XA, XB, XC, ..., XN) ^ S including the nonlinear formula or not. That is, the solution obtaining result (the use amounts (composition ratios) of the brands A to N) by the optimization calculation in Step S2001 is substituted in Formula (18), and it is judged whether Formula (18) holds true or not.
[0225] When Formula (18) holds true as a result of Step S2002, this processing is finished (proceeding to Step S1512 in Fig. 15). Otherwise, when Formula (18) does not hold true, it proceeds to Step S2003, the temporary lower limit value S' is increased slightly by an increase/decrease rate set in advance, and the processing of Step S2001 is performed again. Specifically, until Formula (18) holds true, a convergence calculation is performed such that the temporary lower limit value S' is increased slightly, and solution obtaining by the optimization calculation is repeated.
[0226] In this embodiment, incidentally, the description has been given with an example in which the property restrictions after mixing have a lower limit value, but the same applies to the case where they have an upper limit value. In this case, for the linear formula f' (XA, XB, XQ, . . . XN) , one that forms an upper limit of the nonlinear formula f (XA, XB, Xc, .., XN) is considered. Further, in Step S2001, a temporary upper limit value larger than the upper limit value for the nonlinear formula f(XA, XB,

Xe, •.• XN) is set as the upper limit value for the linear formula f' (XA, XB, XC, • • •, XN) .
[0227] (11) Simulating a stock transition based on the obtained solution (stock transition simulator 1411 in Fig. 14 (corresponding to the stock transition simulator 201 in Fig. 2), Step S1512 in Fig. 15)
Based on the solution for the composition plan formula model and all or part of data obtained by the input data obtaining unit 1401, a simulation for all or part of a targeted composition for the set plan determination period is performed by a set plan creation accuracy. In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in actual operation without any problem. Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation.
[0228] Further, as an example of a restriction which is difficult to handle by a formula model, a setup time needed for setting up equipment when the composition ratio is changed or the like is taken into the simulation so as to perform the simulation

precisely, and thus it is possible to set out a composition plan considering detailed restrictions required in actual operation.
[0229] (12) Simulating a property based on the obtained solution (property simulator 1412 in Fig. 14 (corresponding to the property simulator 202 in Fig. 2), Step S1513 in Fig. 15)
Based on the solution for the composition plan formula model, the stock transition simulated by the stock transition simulator 1411, and all or part of data obtained by the input data obtaining unit 1401, a property is simulated for all or part of a targeted composition for the set plan determination period by a set plan creation accuracy, so as to obtain a property result of a composition raw material after mixing. In this simulation, restriction conditions, operation rules, and so on which could not be incorporated in the composition plan formula model are incorporated to perform the simulation, and thereby the solution provided as a result of obtaining a solution for the composition plan formula model is changed to a composition plan usable in actual operation without any problem. Accordingly, it becomes possible to set out a composition plan considering a time accuracy required in actual operation and detailed restrictions required in actual operation.
[0230] The above-described stock transition simulator 1411 (stock transition simulator 201) and

step S1512, and the property simulator 1412 (property simulator 202) and Step S1513 are examples of a simulator and processing thereof mentioned in the present invention.
[0231] (13) Determining the composition plan (determination unit 1413 in Fig. 14, Step S1514 in Fig. 15)
The composition plan derived by the stock transition simulation and the property simulation is determined for the set plan determination period. As shown in Fig. 7, since the plan determination period is set to one day in this embodiment, first one day of the created composition plan is determined. The plan for the portion of the created composition plan that is not within the plan determination period is discarded without being determined.
[0232] (14) Judging whether the plan for the plan creation period or the plan determination period is determined (judgment unit 1414 in Fig. 14, Step S1515 in Fig. 15)
It is judged whether the plan determination period determined at this point is determined for the plan creation period set in advance. In this embodiment, since the plan creation period is ten days, the plan for the plan determination period is determined at the point when the plan is determined in a tenth loop. Accordingly, the composition plan for ten days is created at the point when determination of the plan is finished in the tenth

loop, and the process is finished.
[0233] (15) Updating the planning start date
(updating unit 1415 in Fig. 14, Step S1516 in Fig.
15)
When the determined plan determination period is not determined for the plan creation period set in advance, the date and time just after the determined composition plan period in the composition plan is set as a new planning start date. In this embodiment, as illustrated in Fig. 7, the planning start date which is initially first day and zero o'clock in the first loop is updated to second day and zero o'clock, and the planning start date which is initially second day and zero o'clock in the second loop is updated to third day and zero o'clock.
[0234] (16) Outputting the composition plan (output unit 1416 of Fig. 14, Step S1517 in Fig. 15)
The composition plan created as described above is displayed on the screen of the display unit 103 and/or is transmitted as data to a not-illustrated external apparatus by the output unit 1416. [0235] The output unit 1416 and Step S1517 described above are examples of an output unit and processing thereof mentioned in the present invention. [0236] As described above, corresponding to the current stock transition state, formula models for the demand-supply balance restrictions and the property restrictions are built first by the plan creation time accuracy for the predetermined

optimization period, a solution for the built composition plan formula model is obtained based on the objective function, and the stock transition and the property after mixing are simulated based on the obtained solution. The composition plan obtained from the simulation result is determined for the set plan determination period, and the date and time just after the plan determination period are set to the new planning start date and time. Thus, a series of processing for determining a composition plan for a new plan target period can be performed sequentially and repeatedly by a predetermined number of times, so as to create a composition plan for a desired plan creation period. Accordingly, a composition plan requiring an arbitrary time accuracy can be optimized quickly in detail, and can further be applied as it is in actual operation.
[0237] (Modification example of the third embodiment)
As a composition plan (for example a use amount (composition ratio)), there are many cases to set out a long-term plan such as an annual plan, a periodic plan, or a monthly plan. Thus, it is also important to create a long-term composition plan in advance, use this composition plan as a reference composition plan, and prevent large separation of a shorter-term composition plan created by a composition plan creation technique adopting the present invention from the reference composition plan.

[0238] Accordingly, in addition to the objective function J built with respect to costs (purchase costs and shipping costs of composition raw materials) illustrated in Formula (34), an objective function J' built with respect to prevention of separation from the composition plan as a reference created in advance may be used. An example of the objective function J' is shown by Formula (36).
J' = ∑(|reference composition ratio (brand) -composition ratio (brand, day) | ) → minimization . , . Formula (36)Reference composition ratio: composition ratio in a composition plan as a reference

[0239] In the above example, an example of creating a daily composition plan with a periodic plan being a reference composition plan in a monthly plan is shown. In this case, the daily sum of difference between the composition ratio (brand, day) and the reference composition ratio of each brand is minimized. As another example, when making the periodic plan, the plan may be created with an annual plan being a composition plan as a reference. In this case, when the composition ratio (brand, month) is decided in the monthly plan, the monthly sum of difference between the composition ratio (brand, month) and the reference composition ratio of each brand is minimized.

[0240] Incidentally, the reference composition plan is created based on past records for example, and any method of creation can be used for this. Of course, a long-term plan may be created in advance by a composition plan creation technique adopting the present invention, and this may be used as the reference composition plan.

[0241] Fig. 21 illustrates a hardware structure example of a computer apparatus 1200 capable of functioning as a composition plan creating apparatus of the present invention. The apparatus is made up of a CPU 1201 as a central processing unit controlling the entire apparatus, a display unit 1202 displaying various input conditions, results, and so on, a storage unit 1203 such as a hard disk storing results and the like, a ROM (read only memory) 1204 storing a control program, various application programs, data, and the like, a RAM (random access memory) 1205 as a work area used by the CPU 1201 when performing processing, an input unit 1206 such as a keyboard and a mouth, and so on.

[0242] Furthermore, when program codes of software for achieving functions of the above-described embodiments are supplied to a computer within an apparatus or system connected to various types of devices so as to operate the various devices for achieving the functions of the above-described embodiments, and the various devices are operated according to a program stored in the computer (CPU or MPU) of the system or apparatus to implement the functions, such functions are also included in the concept of the present invention. In this case, the program codes themselves of the above software achieve the functions of the above-described embodiments, and the program codes themselves and a unit supplying the program codes to a computer, for example, a recording medium storing such program codes constitutes the present invention. As the recording medium storing the program codes, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a magnetic tape, a non-volatile memory card, a ROM, or the like can be used.
Industrial Applicability

[0243] The structure of the present invention makes it possible to build a formula model representing demand-supply balance restrictions of composition raw materials and a formula model representing property restrictions after mixing using mathematical programming or the like, and operate a simulator and an optimization calculating unit in conjunction so as to create a composition plan over plural days while avoiding running out of stocks, satisfying properties, and minimizing costs, when creating a composition plan for mixing a plurality of kinds of composition raw materials.

CLAIMS

What is claimed is:

1. A composition plan creating apparatus creating a composition plan for mixing a plurality of kinds of composition raw materials, the apparatus comprising:

a simulator calculating supply-demand conditions of the composition raw materials and properties after mixing;

a data obtaining unit obtaining data containing arrival schedules of the composition raw materials, stock conditions of the composition raw materials, properties of the composition raw materials, and cost Information;

a model building unit building a formula model . representing a demand-supply balance restriction of the composition raw materials and a formula model representing a property restriction after mixing, with an optimization period set in advance from a planning start date and time being a target, based on the data obtained by the data obtaining unit; and

an optimization calculating unit performing an optimization calculation based on an objective function built with respect to costs using the formula models built by the model building unit, so as to calculate an instruction to the simulator.

2. The composition plan creating apparatus according to claim 1, further comprising
an output unit outputting a composition plan as a simulation result from the simulator.

3, The composition plan creating apparatus according to claim 1,

wherein with data obtained by the data obtaining unit being given, first,

(a) the optimization calculating unit creates a calculation instruction for an optimization period from a planning start date and time,

(b) with a calculation instruction created by the optimization calculating unit being given, the simulator performs a simulation only for a simulation period set in advance,

(c) a result of the simulation is determined as a composition plan only for a plan determination period set in advance, and

(d) a date and time just after the determination is set as a new planning start date and time,

wherein with an already determined composition plan being given, a series of the processing (a) to (d) to determine a composition plan for a new plan determination period is performed repeatedly until a composition plan for a plan creation period is determined, to thereby create a composition plan of a plan creation period.

4.The composition plan creating apparatus according to claim 1,
wherein when creating the composition plan, there is planned a composition approximated to a composition ratio given as a target with respect to a part or all of the composition raw materials.

5. The composition plan creating apparatus according to claim 1,wherein when creating the composition plan, there is planned a composition which does not cause a large separation between the composition ratio of a previous day and the composition ratio of a following day.

6. The composition plan creating apparatus according to claim 1,
wherein when creating the composition plan, if a stock of a composition raw material used on a previous day still exists on a following day, a composition using the composition raw material is planned.

7. The composition plan creating apparatus according to claim 1,
wherein when creating the composition plan, it is possible to specify a part of the composition plan in advance.

8. The composition plan creating apparatus according to claim 1, further comprising:

a linearization unit formulating, when a formula model representing the property restriction after mixing includes a nonlinear formula, a formula model by introducing a linear formula instead of the nonlinear formula; and a judgment unit judging whether or not a solution obtaining result by the optimization calculation unit using the formula model formulated by the linearization unit satisfies a formula model including the nonlinear formula.

9. The composition plan creating apparatus according to claim 8,

wherein when the property restriction after mixing has a lower limit value, the linear formula is a formula forming a lower limit of the nonlinear formula, and

wherein when the property restriction after mixing has an upper limit value, the linear formula is a formula forming an upper limit of the nonlinear formula.

10. The composition plan creating apparatus according to claim 8,

wherein in formulating a formula model by introducing the linear formula instead of the nonlinear formula, when the property restriction after mixing has a lower limit value, the linearization unit sets a temporary lower limit value smaller than the lower limit value, and when the property restriction after mixing has an upper limit value, the linearization unit sets a temporary upper limit value larger than the upper limit value.

11. The composition plan creating apparatus according to claim 10,

wherein when the solution obtaining result by the optimization calculating unit using a formula model formulated by the linearization unit does not satisfy
the formula model including the nonlinear formula, solution obtaining by the optimization calculating unit is repeated while slightly increasing the
temporary lower limit value or slightly decreasing the temporary upper limit value.

12. The composition plan creating apparatus according to claim 1,

wherein the data obtaining unit obtains an arrival Schedule of composition raw materials including incoming amounts by a ship allocation plan as the arrival schedule of composition raw materials purchase cost information of the composition raw materials and shipping cost Information when using a ship as the cost Information/ and

wherein the optimization calculating unit performs an optimization calculation based on an objective function built with respect to purchase costs and shipping costs of the composition raw materials using the formula model built by the model building unit, so as to calculate an instruction to the simulator.

13. The composition plan creating apparatus cording to claim 12, further comprising

an extraction unit extracting a fixed item from predetermined items of the ship allocation plan.

14. The composition plan creating apparatus according to claim 13,

wherein the predetermined items of the ship allocation plan are loading port, loaded brand, loaded amount, unloading port, unloaded brand, and unloaded amount.

15. The composition plan creating apparatus according to claim 13,

wherein the shipping cost Information obtained by the data obtaining unit includes Information of freights by ship, by loading port, and by unloading port and Information of freights by loaded brand and by unloading port, and
wherein which of the freights by ship, by loading port, and by unloading port and the freights by loaded brand and by unloading port are used in the optimization calculating unit is determined according to the fixed item extracted by the extraction unit.

16. The composition plan creating apparatus according to claim 12,

wherein in the optimization calculating unit, the optimization calculation is performed based on an objective function built with respect to prevention of separation from a reference composition plan created in advance, in addition to the objective function built with respect to purchase costs and shipping costs of the composition raw materials.

17. A composition plan creating method creating a composition plan for mixing a plurality of kinds of composition raw materials, the method comprising the
steps of:

obtaining data containing arrival schedules of the composition raw materials, stock conditions of the composition raw materials, properties of the composition raw materials, and cost Information;

building a formula model representing a demand-supply balance restriction of the composition raw materials and a formula model representing a property restriction after mixing, with an optimization period set in advance from a planning start date and time being a target, based on the obtained data; and

performing an optimization calculation based on an objective function built with respect to costs using the built formula models, so as to calculate an instruction to a simulator which calculates supply-demand conditions of the composition raw materials and properties after mixing.