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] The present disclosure relates to control systems and more specifically to calibration of an engine control unit (ECU)

Background of the invention

[0002] Calibration for an engine control unit (ECU) is the process of continuously enhancing the engine performance by measuring changes in the ECU data. A tool (such as one based on MATLAB) can be used to simulate the behavior of the ECU mimicking the functions of ECU and generating responses with different data sets, thus helping to find approximate calibration points. The time and cost can be saved if the whole behavior of ECU to control a process is made available as a simulation model. The conventional methods of engine calibration require large number of metering tests to get the final calibration data. The calibration data increases with increased number of inputs due to generation of one or many outputs. In spite the effort, the calibration value obtained by manual means may not be the best.

[0003] Proportional and Integral controllers (PI controllers) are often used to calibrate engines in order to regulate engine parameters such as engine speed, air flow, engine torque etc. The controller continually monitors a process variable (such as engine RPM) and compares it to a desired target value or a setpoint. Based on this comparison, the controller adjusts a control signal to bring the process variable closer to a set point.

[0004] In order to tune a PI controller, the proportional and integral gain parameters are adjusted to get a desired response characteristic. The gain parameters are adjusted till the controller provides the desired result. In order to adjust these gain parameters a model based tuning or adaptive tuning algorithms can be used to automatically adjust the gain parameters based on engine’s behavior. The iteration cycle is repeated by the controller with new parameter values as input values until the optimum value to achieve the target output is obtained.

[0005] In the prior art US6751510 BA , a predictive and self-tuning PI control apparatus for expanded process control applications is disclosed. The predictive and self-tuning PI controller includes an apparatus for variably assigning gains thereto in accordance with specified time functions following a change in set-point. The control gains are computed in accordance with a plurality of input parameters and subsequently continuously adjusted to set-point errors based on a GPC (generalized predictive control) approach thereby to optimize performance indices derived from simple and classical user specifications. Variation and tuning of control gains in accordance with set-point errors permits the PI controller to be useful in the general area of process control and applicable to a wider range of processes compared to traditional PI control, including time-delay processes, unstable processes and processes with time-varying dynamics.

[0006] The present invention proposes a system and a method to calibrate an engine control unit (ECU) to obtain a target output, said target output dependent on a target value of at least one parameter of the ECU. The disclosure proposes a manner to perform an optimization based tuning of a calibration parameter for any given automotive engine control unit (ECU) implemented as a simulation model.

[0007] In the present disclosure, in order to find the gain parameters for the PI controller, a Recursive Modified Pattern Search (RMPS) based optimization technique is proposed. The RMPS is suitable for design problems that have multiple constraints and objectives. At each iteration, new possible set of solutions are found by adding a set of derived step-size vectors to the initial starting point. While deriving these step-size vectors, precautions and adjustments are considered so that the set of new possible solution points remain within the constrained space.

[0008] Therefore, the present disclosure proposes an optimization-based tuning of the calibration parameter for any given ECU implemented as a model-based design means through MATLAB/SIMULINK. Since the calibration is done is automatic fashion , it reduces all sorts of manual efforts.

Brief description of the accompanying drawings
[0009] An embodiment of the invention is described with reference to the following accompanying drawings:
[0010] Figure 1 depicts a system to calibrate an ECU to obtain a target output
[0011] Figure 2 depicts a flow chart for a method to calibrate an ECU to obtain a target output.

Detailed description of the drawings

[0012] Figure 1 depicts a system to calibrate an ECU to obtain a target output. Said target output depends on a target value of at least one parameter of the ECU.

[0013] The system (10) depicted may be a control system. A person skilled in the art is expected to know the concept of Proportional -Integral controller, control-loop feedback systems and their use in control systems. For the sake of brevity, an elaborate description of what is already known in the art is omitted.

[0014] A control system is a system that manages, regulates and directs the systems or processes. The present control system, herein referred to as the system(10) is implemented as a closed loop feedback system to continuously adjust input parameters of the engine in order to achieve the target output.

[0015] The system may be implemented as model on MATLAB that is designed to simulate the behaviour of engine in different operating conditions. This system may be used to identify the optimal set of values for input parameters to produce a target output. The system includes a Controller (1). The controller may be a Proportional-integral (PI) controller used in feedback control systems.

[0016] In an example, the controller may be a part of the ECU or the ECU itself may be implemented as the controller. The controller processes the input data and makes appropriate adjustment to achieve the target output. The behavior of the ECU to control a process based on the operating conditions set by input values of different engine parameters is simulated by an actuator or a simulation model (2). The controller responds to various inputs fed thereto. A value of a target output is set up into the controller. Based on an input value (6) of a parameter fed into the system by a user, the simulation model (2 )generates the output (7). An error e is computed by the controller that represents a difference between the target output and the output of the simulation model.
[0017] Preferably, the target output, objective functions, computation algorithms and other data variables are held in a Random Access Memory (not explicitly shown herein) of the microprocessor/digital computer with an input and output interface. The software used for this purpose by the present invention is the same as in other digitally implemented controllers and, accordingly, a detailed description thereof is omitted.

[0018] The controller(1) then computes ‘gains’ based on this error, mainly the proportional and integral gains. A proportional gain determines the relation between the error e and the output of the simulation model. Higher the proportional gain, higher will be the correction for the error e to reach the target output. The proportional gain determines the immediate response of the control system to the error whereas the integral gain determines the long term response and the ability of the system to eliminate steady state.

[0019] The integral gain eliminates the steady state errors that occur due to disturbances in the processs. It is a constant that describes the rate at which a control system accumulates correction over time. Generally, integral gain is used to correct any residual error that remains after proportional correction has been applied.

[0020] The performance of the PI controller is evaluated through performance indices. The indices such as Integral Squared error (ISE), Integral of time weighted square error (ITSE), Integral of Absolute error (IAE) may be used. These representative equations of some of the indices are given only for exemplary purposes

[0021] where, e(t) is the error response of a system at a time t:

Integral Square Error (ISE)
ISE=∫e2(t)dt

Integral of the absolute magnitude of error (IAE)
IAE=∫│ e (t) │dt

Integral Time-absolute error (ITAE)
ITAE=∫ t.│ e (t) │dt

Integral Time-Square error (ITSE)
ITSE=∫ e2 (t) dt

[0022] Based on these performance indices an objective function is obtained by the controller. An objective function defines the desired behaviour of the process implemented by the simulation model to achieve a target output. A typical objective function may be expressed as weighted sum of the ISE , ITSE and IAE expressed as : Objective function = (w1*ISE + w2*ITSE +w3*IAE) , where w1, w2, w3 are weighing factors that reflect the relative importance of each function. By minimising (or maximizing) this objective function, or in other words, by ‘optimizing’ this function, the target output can be achieved. The choice of a performance index and the weighting factors depend upon the requirement of ECU being calibrated.

[0023] An optimizing element (3) as a part of the controller optimizes this objective function to get the desired output. In the present disclosure, the objective function is optimized using the RMPS technique. RMPS or the recursive modified pattern search is a derivative free optimization algorithm, known in the art, that aims to find the minimum of the objective function without the need for calculating its gradient. RMPS iteratively refines its search space using a combination of pattern moves and recursive sub-sampling. At each iteration, RMPS starts with an initial set points that are evenly spaced in a search space. It then performs a pattern move by constructing a new set of points by translating the existing points in various directions. This pattern move allows the RMPS to explore a larger region of the search space by maintaining an even distribution of points. After the pattern move, RMPS evaluates the objective function at all new points and retains that best points. It then performs a recursive sub sampling of the search space by dividing it into smaller regions of the search space that contains the minimum of the function. RMPS continues iterating until a stopping criterion is such as the best points (or values) found during the optimization process.

[0024] In the present disclosure, while optimizing the objective function using RMPS technique, for each of the at least one parameter having n dimensions in space, the value of the objective function is evaluated at 2n neighboring points. These 2n neighboring points are obtained by making 2n coordinate-wise movements of plurality of step-sizes. (A step size is the amount by which each iteration is adjusted in the optimization process. The step size determines how quickly the optimization process converges into a solution. )
[0025] In an example, assume that an ECU is to be calibrated for the vehicles alternator speed . The parameters upon which the vehicle’s alternator speed depends include (non exhaustively) the engine speed, the electrical load, the battery state of charge, the temperature and the type of alternator. In an example the PI controller regulates the alternator speed based on the drivers demand for a set temperature in the car. A transfer function that relates the electrical load and temperature to vehicles alternator speed is implemented by the simulation model. The simulation model uses a design specification , specifying how the system should behave.

[0026] In an example, the proportional and integral gains are determined for a target alternator speed as the target output. The objective function is determined based on performance indices as explained above. This objective function is charachterised by tuning (optimizing) using the RMPS technique by the optimizing element (3). At each iteration with an input value, starting from the current solution, new possible set of solutions are found by adding a set of derived step-size vectors to the initial starting point. While deriving these step-size vectors, precautions and adjustments are considered so that the set of new possible solution points remain within the constrained space. The iteration cycle is repeated till a target value of at least one parameter is achieved that gives the target output.

[0027] Referring again to Figure 1, disclosed is the system (10) to calibrate an engine control unit (ECU) (not shown) to obtain a target output, said target output dependent on a target value of at least one parameter of the ECU. The system comprises a simulation model (2) of a process to be controlled by the ECU and a controller (3). The controller is adapted to receive at least one value for said at least one parameter as an input value (6); receive an output (7) computed by said simulation model based on the input value (6) for said at least one parameter; compute at least one performance index based on the target output and the output computed by said simulation model; compute an objective function based on the said at least one performance index; and, obtain the target value of said at least one parameter by optimizing the objective function using a Recursive Modified Pattern Search (RMPS) technique. The optimizing element (3) performs this optimization.

[0028] The objective function represents a function of an error e between the target output and the output computed by the simulation model. Further, while optimizing the objective function using RMPS technique, for each of the at least one parameter having n dimensions in space, the value of the objective function is evaluated at 2n neighboring points; and the 2n neighboring points are obtained by making 2n coordinate-wise movements of plurality of step-sizes. The optimizing element (3) performs this optimization.

[0029] The obtained target value is received as input value (6) by the controller until the target output is achieved or in other words, when (7) becomes the target output. Therefore, the iteration cycle is repeated with new calibration parameter values (target values) as input to the controller until optimum calibration parameters (the target value(s) that gives the target output) are obtained.

[0030] Referring to Figure 2, the same depicts a flowchart of the method (100) to calibrate an engine control unit (ECU) to obtain a target output, said target output dependent on a target value of at least one parameter of the ECU. The method implemented by the system as described in Fig 1, comprises the several steps wherein, the first step (101) includes receiving, by the controller, at least one value for said at least one parameter as an input value. This is followed by the next step (102) of receiving by the controller, an output computed by a simulation model of a process to be controlled by the ECU, based on the input value for said at least one parameter. This is followed by step (103) which includes- computing, by the controller, at least one performance index based on the target output and the output computed by said simulation model. The next step (104) includes computing, by the controller, an objective function based on the said at least one performance index. In this method step, the characterizing step (105) is obtaining the target value of said at least one parameter, by the controller, by optimizing the objective function using a Recursive Modified Pattern Search (RMPS) technique.
[0031] While optimizing the objective function, for each of the at least one parameter having n dimensions in space, evaluating the value of the objective function at 2n neighboring points; and obtaining the 2n neighboring points by making 2n coordinate-wise movements of plurality of step-sizes. The target value for the parameter achieved is again used as an input value by the controller until a target output is achieved.

[0032] The present advantageously provides an optimization based tuning of the calibration parameter for any given ECU implemented as a model-based design means through MATLAB/SIMULINK. Since the calibration is done is automatic fashion , it reduces all sorts of manual efforts. The calibration through RMPS provides a faster, more precise and derivative free tuning

, Claims:
We Claim:
1.A system (10) to calibrate an engine control unit (ECU) to obtain a target output, said target output dependent on a target value of at least one parameter of the ECU,
the system comprising:

-a simulation model (2) of a process to be controlled by the ECU;
-a controller(1) adapted to :
receive at least one value for said at least one parameter as an input value (6);

receive an output (7) computed by said simulation model based on the input value for said at least one parameter;

compute at least one performance index based on the target output and the output (7) computed by said simulation model;

compute an objective function based on the said at least one performance index; and
obtain the target value of said at least one parameter by optimizing the objective function using a Recursive Modified Pattern Search (RMPS) technique.

2. The system (10) as claimed in Claim 1, wherein, said objective function represents a function of an error between the target output and the output (7) computed by the simulation model.

3. The system (100) as claimed in Claim 1, wherein, while optimizing the objective function using RMPS technique, for each of the at least one parameter having n dimensions in space, the value of the objective function is evaluated at 2n neighboring points; and
the 2n neighboring points are obtained by making 2n coordinate-wise movements of plurality of step-sizes.

4. The system (10) as claimed in claim 1, wherein, the obtained target value is received as input value (6) by the controller until the target output is achieved.

5. The system (10) as claimed in claim 1, wherein, the controller is the ECU.

6. A method (100) to calibrate an engine control unit (ECU) to obtain a target output, said target output dependent on a target value of at least one parameter of the ECU, the method comprising the steps of:

-receiving, by the controller, at least one value for said at least one parameter as an input value (101);

-receiving by the controller, an output computed by a simulation model of a process to be controlled by the ECU, based on the input value for said at least one parameter (102);

-computing, by the controller, at least one performance index based on the target output and the output computed by said simulation model (103);

-computing, by the controller, an objective function based on the said at least one performance index (104),

the method characterized by:
obtaining the target value of said at least one parameter, by the controller, by optimizing the objective function using a Recursive Modified Pattern Search (RMPS) technique (105).

6. The method (100) as claimed in Claim 6, wherein, optimizing the objective function using RMPS technique, in that,
for each of the at least one parameter having n dimensions in space,
evaluating the value of the objective function at 2n neighboring points; and

obtaining the 2n neighboring points by making 2n coordinate-wise movements of plurality of step-sizes.

7. The method (100) as claimed in claim 6, wherein, receiving the target value as input value by the controller until the target output is achieved.