Description:Complete Specification:

The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed.
Field of the invention
[0001] This invention is related to a method of optimizing an inventory and a control unit thereof.

Background of the invention

[0002] The major challenges in inventory management are high inventory costs and low service levels. Each material has a different consumption pattern hence the optimal technique used to optimize inventory (calculate safety stock and reorder point) may be different for each material. Using a non-optimal technique does not optimize the inventory in the most efficient way which results in high inventory cost and/or increase in stock outs. There are many conventional independent inventory optimization techniques currently used in the inventory management systems, but there is no ensemble learning based system that leverages multiple artificial intelligence (AI) techniques and identifies the optimal technique for each material.

[0003] A US patent application 20200143313 discloses Systems and methods for inventory management and optimization. The present disclosure provides systems and methods that may advantageously apply machine learning to accurately manage and predict inventory variables with future uncertainty. In an aspect, the present disclosure provides a system that can receive an inventory dataset comprising a plurality of inventory variables that indicate at least historical (i) inventory levels, (ii) inventory holding costs, (iii) supplier orders, and/or (iv) lead times over time. The plurality of inventory variables can be characterized by having one or more future uncertainty levels. The system can process the inventory dataset using a trained machine learning model to generate a prediction of the plurality inventory variables.

Brief description of the accompanying drawings
[0004] Figure 1 illustrates a control for optimizing an inventory, in accordance with an embodiment of the invention; and
[0005] Figure 2 illustrates a flow chart of a method of optimizing an inventory, in accordance with the present invention.

Detailed description of the embodiments
[0006] Figure 1 illustrates a control for optimizing an inventory, in accordance with an embodiment of the invention. The control unit 10 classifies a material 14 stored in the inventory 12 based on a consumption pattern using a demand classification technique and selects at least one operational model 16 based on the consumption pattern/demand type for optimizing an execution time. The control unit 10 calculates a re-order point (ROP) and safety stock value related to above selected at least one operational model and identifies an optimal ROP and safety stock having a minimum positive value using an ensemble learning technique. The control unit 10 selects the operational model 16 having the minimum positive value to ensure no stock out situation. The control unit 10 performs/re-does the above-mentioned steps for a predefined number of days for optimizing the inventory 12.

[0007] The inventory 12 comprises multiple materials 14 stored that are used in a manufacturing plant. If the user don’t calculate the re-order point (ROP) and the safety stock value for each of the material, there will be a no-stock situation or overload of the stock in the plant. In order to avoid , this the present invention provides a solution in optimizing the stock in the inventory by properly calculating the ROP and the safety stock for each of the material 14 and avoiding the execution of all the operational models 16 on all the materials 14. The ROP is a time/date at which the material needs to be ordered. For example, the ROP for a “X” material is every Tuesday at 10AM. The control unit 10 comprises a memory unit 18 that is loaded with the historical demand data, i.e.., multiple ROP’s for corresponding multiple materials 14 and their corresponding next lead time values. The next lead time values are the ones that are calculated in advance regarding the next order that needs to be placed.

[0008] The control unit 10 uses the demand classification technique for classifying the materials 14 and the classification comprises smooth, erratic, intermittent and lumpy. The at least one operational model 16 uses any one of the approaches comprising a statistical approach or an intelligence network learning approach. The control unit 10 uses any one of the intelligence network learning approaches comprising a machine-learning approach, an artificial intelligence approach, a deep-learning approach and the like. The control unit 10 is chosen from a group of control units comprising a micro-processor, a micro-controller, a digital circuit, an integrated chip and the like.

[0009] Figure 2 illustrates a flowchart of a method of optimizing an inventory according to the present invention. In step S1, a material 14 stored in the inventory 12 is classified based on a consumption pattern using a demand classification technique. In step S2, at least one operational model 16 is selected based on the consumption pattern/demand type for optimizing an execution time. In step S3, a re-order point (ROP) and safety stock related to above selected at least one operational model 16 is calculated. In step S4, an optimal ROP and the safety stock value having a minimum positive value using an ensemble learning technique and selecting the operational models 16 having the positive value to ensure no stock out situation. In step S5,above steps are performed for a predefined number of days for optimizing the inventory 12.

[0010] The method is explained in detail. Multiple materials are stored in the inventory 12. The materials 14 are classified based on the demand classification technique. The materials 14 are classified into multiple groups as mentioned above. After this, the control unit 10 selects at least one operational model 16 based on the consumption pattern/demand type at material level and at plant level. For instance, there are two plants plant A and plant B , both consume materials X,Y and Z . In the plant A the service level is 98% and plant B has a service level of 90%, i.e.., the consumption of materials in plant A will be more and will need faster replenishment to continue with the service level of 98%,whereas in plant B it would be comparatively less as the service level is 90% only.

[0011] The selection of a set of operational models 16 is based on various recommendations for different type of demand. The control unit 10 optimizes the execution time by avoiding execution of all the model(s) 16 for all the materials 14.

[0012] Using one operational model 16 on all the materials or multiple operational models on all the materials in the category will not result in an optimal value of safety stock and ROP. Hence, the control unit 10 selects at least one operational model 16 and calculates the ROP and the safety stock value of that particular operational model 16 for the materials on a given time of the day or on a particular day. The control unit 10 reduces the execution time by avoiding execution of all models 16 for all the materials present in the inventory by selecting the operational models 16 having the positive values. The control unit 10 calculates ROP and the safety stock value for each of the selection model 16.

[0013] The control unit 10 calculates multiple ROP values using the selected models 16 for a given time of a day, wherein the given time of the day is taken from a historical demand data. For instance, the control unit 10 calculates the ROP values using all the methods for a particular day (preferably latest day) in historical demand data. The control unit 10 further calculates the difference between the calculated ROP and safety stock (from all the methods) and next lead time demand value. The control unit 10 selects the operational models 16 with only positive difference between calculated ROP and next lead time demand value.

[0014] The negative difference (where the calculated ROP < next lead time demand value) will lead to stock out situation and the control unit 10 ignores those operational models 16. The control unit 10 further identifies the operational models 16 that has minimum positive difference, and the positive value ensures no stock out situation and the minimum values ensures that the stock is minimum (No over stocking of the material). This methodology not only avoids the no stock situation but also ensures minimum stock situation.

[0015] The control unit 10 repeats or performs the above-mentioned steps for a window of “N days” or the predefined number of days. For instance, the predefined number of days is 30. The control unit 10 identifies the at least one operational model 16 with the maximum best result among the N days as the optimal operational model for a particular material and the ROP and the safety stock calculated using this operational model 16 is the optimal ROPs and safety stock values.

[0016] The above disclosed method is explained in detail with an example. The ROP calculation for material a material named “F002”. The control unit 10 calculates the ROP (for material “F002” using all the methods for 5/30/2022 and the demand during next lead time is 172.
Date Material Method Name ROP Difference between ROP and Next LT Demand
5/30/2022 F002 Corston Forecasting Error Method 216 44
5/30/2022 F002 Statistical Method 207 35
5/30/2022 F002 Conventional Method 4770 4598
5/30/2022 F002 Corston Avg. Forecasted Demand Method 201 29
5/30/2022 F002 Random Forest Forecasting Error Method 129 -43
5/30/2022 F002 Random Forest Avg. Forecasted Demand Method 165 -7
5/30/2022 F002 Prophet Forecasting Error Method 174 2
5/30/2022 F002 Prophet Avg. Forecasted Demand Method 208 36

[0017] The difference between the calculated ROP and the safety stock value and the demand during next lead time (172) is calculated for all the different methods for a particular day (5/30/2022) and for the material F002 as shown in the above disclosed table. The control unit identifies the ‘Prophet Forecasting Error Method’ is selected as the best method for ‘5/30/2022’ as it has the lowest positive difference (+2) as the best operational model as the difference between the calculated ROP (174) and next lead time demand (172) is positive and is also minimum.

[0018] The above disclosed steps are repeated for all the days in the window of N days (for instance, if N=10) and the best methods are identified for each day. The below table provides information on the different methods and their ROP’s calculated and selection of the one of the best methods/operational models based on the minimum positive difference for a continuous ten days.

Date Material Best method name
5/21/2022 F002Z98265 Corston Forecasting Error Method
5/22/2022 F002Z98265 Prophet Forecasting Error Method
5/23/2022 F002Z98265 Conventional Method
5/24/2022 F002Z98265 Prophet Forecasting Error Method
5/25/2022 F002Z98265 Random Forest Forecasting Error Method
5/26/2022 F002Z98265 Random Forest Avg. Forecasted Demand Method
5/27/2022 F002Z98265 Prophet Forecasting Error Method
5/28/2022 F002Z98265 Prophet Forecasting Error Method
5/29/2022 F002Z98265 Prophet Forecasting Error Method
5/30/2022 F002Z98265 Prophet Forecasting Error Method

Based on the above analysis, the control unit 10 identifies the operational model/method ‘Prophet Forecasting Error Method’ as the best method for material ‘F002’ as this is the best method for maximum number of days out of 10 days.

[0019] With the above disclosed method, the inventory 12 is always maintained with a minimum required stock. No stock situation can be avoided and also over stocking of the material can be reduced as with the selection of the optimal re-order point and the safety stock value. This improves customer satisfaction and increases overall supply chain efficiency. By optimizing inventory levels, businesses can reduce excess inventory costs while ensuring they have enough inventory to meet orders and avoid stockouts. Reorder point (ROP) and safety stock values work together to minimize stockouts and meet customer demand while keeping inventory levels low.

[0020] It should be understood that embodiments explained in the description above are only illustrative and do not limit the scope of this invention. Many such embodiments and other modifications and changes in the embodiment explained in the description are envisaged. The scope of the invention is only limited by the scope of the claims.
, Claims:We Claim:

1. A control unit (10) for optimizing an inventory (12) , said control unit (10) adapted to :
- classify a material (14) stored in said inventory (12) based on a consumption pattern using a demand classification technique;
- select at least one operational model (16) based on said consumption pattern/demand type for optimizing an execution time ;
- calculate a re-order point (ROP) and a safety stock value related to above selected at least one operational model (16);
- identify an optimal ROP and safety stock value having a minimum positive value using an ensemble learning technique and select said operational model (16) having said minimum positive value to ensure no stock out situation ;
- re-doing above steps for a predefined number of days for optimizing said inventory (12).

2. The control unit (10) as claimed in claim 1, wherein said control unit (10) uses said demand classification technique on any one of the following said consumption pattern/demand types of said material (14) comprising smooth, erratic, intermittent and lumpy.

3. The control unit (10) as claimed in claim 1, wherein said at least one operational model (16) using any one of approaches comprising a statistical approach or an intelligence network learning approach.

4. The control unit (10) as claimed in claim 3, wherein said control unit (10) adapted to select said at least one operational model (16) based on said demand type at a material level and a plant level and said control unit (10) reduces execution time by avoiding execution of all models for all the materials (14) present in said inventory (12).

5. The control unit (10) as claimed in claim 1, wherein said control unit (10) calculates ROP value and the safety stock value for each of said selection model (16) .

6. A method of optimizing an inventory (12) by a control unit (10) , said method comprising :
- classifying a material (14) stored in said inventory (12) based on a consumption pattern using a demand classification technique;
- selecting at least one operational model (16) based on said consumption pattern/demand type for optimizing an execution time ;
- calculating a re-order point (ROP) and a safety stock value related to above selected at least one operational model (16);
- identifying an optimal ROP and safety stock value having a minimum positive value using an ensemble learning technique and selecting said operational models (16) having said minimum positive value to ensure no stock out situation ;
- re-doing above steps for a predefined number of days for optimizing said inventory (12).

7. The method as claimed in claim 6, wherein calculating multiple ROP values and safety stock values using said selected models for a given time of a day, wherein said given time of said day is taken from a historical demand data.

8. The method as claimed in claim 6, wherein calculating a difference between said calculated multiple ROP’s and next lead time demand values and selecting said operational models (16) with said positive values.

9. The method as claimed in claim 8, wherein identifying an operational model (16) for optimizing said inventory (12) for a particular said material (14) for said predefined number of days, based on said calculated difference of said ROP of a corresponding said operational model (16) and said next lead time demand value.

10. The method as claimed in claim 1, wherein selection of said at least one operational model (16) for optimizing the inventory (12) is based on the safety stock value.