Parameter adjustment device, parameter adjustment method, and program

Abstract

A parameter adjustment device (1) is provided with: a condition setting unit (121) that sets at least one operation condition and a control parameter to a control system (2); an evaluation unit (122) that calculates an evaluation value relating to the operation of the control system (2) when the control system (2) operates with at least one operation condition and a control parameter, for each of the at least one operation condition; a comprehensive evaluation unit (123) that calculates, as a comprehensive evaluation value, a weighted average value obtained using at least the maximum value and the minimum value from among the evaluation values; and an optimization unit (124) that calculates, using an optimization algorithm, a control parameter for operating the control system (2) next on the basis of the comprehensive evaluation value. The condition setting unit (121) sets at least one operation condition and the control parameter calculated by the optimization unit (124) to the control system (2).

Description

Parameter adjustment device, parameter adjustment method, and program

[0001] The present disclosure relates to a parameter adjustment device, a parameter adjustment method, and a program.

[0002] In DC-DC converters used in electric vehicles and the like, complex and sophisticated current and voltage control is performed by combining multiple feedback controls to control charging and discharging according to driving conditions. Also, in production equipment used in factories and the like (e.g., component mounters, assembly robots, etc.), complex and sophisticated operation control is performed by combining multiple servo-controlled motors. In these types of devices, the operation of the controlled object is controlled according to a large number of control parameters. Users empirically adjust (i.e., optimize) the control parameters to obtain desired performance regarding the operation of the controlled object.

[0003] In such a control parameter adjustment process, the operating conditions that define the operating pattern of the controlled object become more diverse and complex, and the greater the number of control parameters that are set to realize these operating conditions, the more difficult it becomes to optimize the control parameters. As a result, the number of trials required for the control parameter adjustment process increases, requiring a great deal of effort and time from the user. Therefore, there is a growing demand for automating the control parameter adjustment process.

[0004] For this reason, in optimizing the control parameters set to realize traffic light operation control, a parameter adjustment device has been disclosed for optimizing the control parameters by narrowing down the operating conditions to bottleneck time periods and bottleneck areas (see, for example, Patent Document 1).

[0005] Furthermore, in order to efficiently solve the minimax optimization problem in optimizing control parameters, there is a method called Adaptive Scenario Subset Selection (AS3) that prioritizes the execution and evaluation of the operating conditions with a proven track record that obtain the worst evaluation value in the optimization loop (see, for example, Non-Patent Document 1).

[0006] JP 2017-117140 A

[0007] Atsuhiro Miyagi, Kazuto Fukuchi, Jun Sakuma, Youhei Akimoto, Adaptive scenario subset selection for worst-case optimization and its application to well placement optimization, Applied Soft Computing, Volume 133, 2023, 109842.

[0008] However, when relatively difficult operating conditions are involved, such as those that cause control oscillation or output saturation for most control parameters, the parameter adjustment device described in Patent Document 1 may take a very long time to adjust the control parameters by using optimization that focuses on the bottleneck operating conditions.

[0009] Furthermore, if operating conditions that require a relatively long time to evaluate are included, the parameter adjustment device described in Non-Patent Document 1 may actively execute and evaluate such operating conditions, which may result in a very long time required to adjust the control parameters.

[0010] Therefore, an object of the present disclosure is to provide a parameter adjustment device, a parameter adjustment method, and a program that can reduce the time required for performance evaluation using simulation or an actual device in automatic adjustment of control parameters for multiple operating conditions.

[0011] In order to achieve the above goal, a parameter adjustment device according to one embodiment of the present disclosure is a parameter adjustment device that searches for optimal control parameters for operating a control system under a plurality of operating conditions by operating the control system while adjusting control parameters, and includes: a condition setting unit that sets at least one operating condition of the plurality of operating conditions and the control parameter to the control system; an evaluation unit that calculates, for each of the at least one operating condition, an evaluation value related to the operation of the control system when the control system operates under the at least one operating condition and the control parameter set by the condition setting unit; a comprehensive evaluation unit that calculates, as an overall evaluation value, a weighted average value obtained using at least the maximum and minimum values ​​of the evaluation values ​​calculated by the evaluation unit; and an optimization unit that calculates, using an optimization algorithm, control parameters for the next operation of the control system based on the overall evaluation value calculated by the comprehensive evaluation unit, and the condition setting unit sets the at least one operating condition and the control parameter calculated by the optimization unit to the control system.

[0012] In order to achieve the above-mentioned goal, a parameter adjustment method according to one embodiment of the present disclosure is a parameter adjustment method executed by a computer that searches for optimal control parameters for operating a control system under a plurality of operating conditions by operating the control system while adjusting control parameters, and includes: a condition setting step of setting at least one operating condition of the plurality of operating conditions and the control parameters in the control system; an evaluation step of calculating an evaluation value related to the operation of the control system when the control system operates under the at least one operating condition and the control parameters set in the condition setting step; a comprehensive evaluation step of calculating, as an overall evaluation value, a weighted average obtained using at least the maximum and minimum values ​​of the evaluation values ​​calculated in the evaluation step; and an optimization step of calculating, using an optimization algorithm, control parameters for next operating the control system based on the overall evaluation value calculated in the comprehensive evaluation step, wherein the at least one operating condition and the control parameters calculated in the optimization step are set in the control system in the condition setting step.

[0013] In order to achieve the above goal, a program according to one embodiment of the present disclosure causes the computer to execute the above parameter adjustment method.

[0014] The present disclosure provides a parameter adjustment device, a parameter adjustment method, and a program that can reduce the time required for performance evaluation using simulation or an actual device in automatic adjustment of control parameters for a plurality of operating conditions.

[0015] FIG. 1 is a block diagram illustrating the configuration of a system including a parameter adjustment device according to the present disclosure. FIG. 2 is a detailed configuration diagram of the control system illustrated in FIG. 1. FIG. 3 is a flowchart illustrating the operation of the parameter adjustment device according to the present embodiment. FIG. 4 is a flowchart illustrating the detailed operation of step S205 illustrated in FIG. 3. FIG. 5A is a diagram illustrating an example of an objective function corresponding to a plurality of operating conditions. FIG. 5B is a diagram illustrating fluctuations in the control input and the controlled variable with respect to time at points corresponding to the crosses a to i attached to the objective function in FIG. 5A. FIG. 6 is a diagram illustrating an example of transitions in the selection probability of an operating condition and the overall evaluation value when the selection probability update unit increases the selection probability of an operating condition corresponding to the maximum evaluation value, and the overall evaluation unit sets the maximum value among the evaluation values ​​obtained by the adjustment as the overall evaluation value. FIG. 7 is a diagram illustrating an example of transitions in the selection probability of an operating condition and the overall evaluation value when the selection probability update unit increases the selection probability of an operating condition corresponding to the maximum and minimum evaluation values, and the overall evaluation unit sets the average value of the maximum and minimum evaluation values ​​among the evaluation values ​​obtained by the adjustment as the overall evaluation value. FIG. 8 is a diagram illustrating an example of an evaluation cost set for each operating condition. FIG. 9 is a diagram showing an example of the transitions in the selection probability of an operating condition and the overall evaluation value when the selection probability update unit further performs the process of step S255 of FIG. 4 to reduce the total evaluation cost in the search. FIG. 10 is a diagram showing the relationship between the cumulative evaluation cost for each parameter adjustment method and the maximum value of the obtained evaluation value. FIG. 11 is a diagram showing an example of an image displayed on the input / output device by the input / output unit. FIG. 12 is a table showing an example of the estimated success or failure of multiple operating conditions in a certain number of search attempts. FIG. 13 is a diagram showing the relationship between the search progress and the correction coefficient. FIG. 14 is a diagram showing an example of the transitions in the selection probability of an operating condition and the overall evaluation value when the process of step S255 of FIG. 4 is performed. FIG. 15 is a diagram showing the transitions in the evaluation value when focusing on operating conditions cond2 and cond3. FIG. 16 is a diagram showing a modification of the image shown in FIG. 11.

[0016] Embodiments of the present disclosure will be described below with reference to the drawings. Each of the embodiments described below represents a specific example of the present disclosure. The numerical values, components, component placement and connection, steps, step order, and display examples shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, the present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope of the claims. Furthermore, the technical features described in each embodiment can be combined with each other. Furthermore, each drawing is not necessarily an exact representation. In each drawing, substantially identical components are designated by the same reference numerals, and redundant explanations are omitted or simplified.

[0017] 1 is a block diagram showing the configuration of a system including a parameter adjustment device 1 according to the present disclosure. The present disclosure is composed of the parameter adjustment device 1, a control system 2 that is the adjustment target of the parameter adjustment device 1, and an input / output device 3 that is a device through which a user makes various settings for the parameter adjustment device 1 and checks information on the results of the adjustment of control parameters.

[0018] The parameter adjustment device 1 is a device that searches for optimal control parameters for operating the control system 2 under multiple operating conditions by operating the control system 2 while adjusting control parameters. The parameter adjustment device 1 is, for example, a terminal device such as a personal computer. The parameter adjustment device 1 also includes, as its functional components, an input / output unit 11 and a control unit 12.

[0019] The input / output unit 11 is an interface that communicates with the input / output device 3. The input / output unit 11 receives information about setting conditions for adjusting the control parameters from the input / output device 3 and outputs the information to the control unit 12. The information about setting conditions for adjusting the control parameters includes operating conditions used to search for the control parameters, evaluation indices that are used by the evaluation unit 122 (described later) to calculate evaluation values, and a combination of operating conditions and evaluation costs for each operating condition. The evaluation indices are indices related to the performance of the control system 2 (e.g., the settling time and position deviation of the control system 2). The settling time is the time required for the control system 2 to move from its drive start position to an allowable position at which it can be evaluated as having reached the target position. The position deviation is the distance between the position of the control system 2 and the target position. The evaluation cost is the cost incurred when evaluating one operating condition, and corresponds to, for example, the time required for the evaluation, the energy and cost required for the evaluation, and the burden on the operator and the object to be adjusted (the control system 2).

[0020] Furthermore, the input / output unit 11 receives information on the results of the adjustment of the control parameters from the control unit 12, and outputs the information to the input / output device 3. The results of the adjustment of the control parameters will be described in detail later.

[0021] The control unit 12 is realized by a microcomputer, a processor, etc. That is, the functions of the control unit 12 are realized by the microcomputer, the processor, etc. executing a program stored in a memory.

[0022] The control unit 12 also includes, as functional components, a condition setting unit 121 , an evaluation unit 122 , a comprehensive evaluation unit 123 , an optimization unit 124 , a selection probability update unit 125 , and an operating condition selection unit 126 .

[0023] The condition setting unit 121 acquires the operating conditions received by the input / output unit 11. The condition setting unit 121 sets at least one operating condition and control parameters from the acquired operating conditions in the control system 2. Specifically, the condition setting unit 121 sets the control parameters output by the optimization unit 124 in the control system 2. The condition setting unit 121 also sequentially provides signals based on the operating conditions included in the operating condition set output by the operating condition selection unit 126 to the control system 2 (i.e., sequentially sets them). Furthermore, sequentially providing signals based on the operating conditions to the control system 2 means that after the operation of the control system 2 based on one operating condition is completed, another operating condition is set.

[0024] The condition setting unit 121 outputs the operating conditions and control parameters set in the control system 2 to the evaluation unit 122. In addition, the condition setting unit 121 outputs a set of operating conditions set in the control system 2 for each search to the input / output unit 11.

[0025] The evaluation unit 122 observes the operational control of the control system 2 based on the operating conditions given to the control system 2 by the condition setting unit 121. The evaluation unit 122 calculates an evaluation value for the observed operational control in accordance with the evaluation index accepted by the input / output unit 11. The evaluation unit 122 outputs a combination of the operating conditions and the evaluation value for each operating condition to the overall evaluation unit 123 and the selection probability update unit 125. The evaluation unit 122 also outputs the evaluation value of the control parameter obtained under each operating condition for each search trial to the input / output unit 11.

[0026] The overall evaluation unit 123 calculates an overall evaluation value from a combination of the operating conditions output by the evaluation unit 122 and the evaluation values ​​for each operating condition, and outputs the calculated overall evaluation value to the optimization unit 124. The overall evaluation unit 123 also outputs the overall evaluation value output to the optimization unit 124 to the input / output unit 11. A detailed description of the method for calculating the overall evaluation value will be given later.

[0027] The optimization unit 124 uses a black-box optimization algorithm based on the overall evaluation value output by the overall evaluation unit 123 to output to the condition setting unit 121 control parameters intended to search for control parameters that will minimize the overall evaluation value (i.e., the best value). The optimization unit 124 also outputs the control parameters for each search trial output to the condition setting unit 121 to the input / output unit 11. The black-box optimization algorithm is a known algorithm such as evolutionary computation or Bayesian optimization.

[0028] The selection probability update unit 125 updates the selection probability table and the cost-gradient selection probability table for operating conditions stored therein based on the combination of the operating condition and the evaluation value for each operating condition output by the evaluation unit 122 and the combination of the operating condition and the evaluation cost for each operating condition received by the input / output unit 11. The selection probability update unit 125 outputs at least one of the updated selection probability table and the cost-gradient selection probability table to the operating condition selection unit 126. The selection probability update unit 125 also outputs the selection probability table or the cost-gradient selection probability table output to the operating condition selection unit 126 to the input / output unit 11. The selection probability table is a table indicating the probability of selection by the operating condition selection unit 126 for each operating condition. The cost-gradient selection probability table is a table in which the probability for each operating condition in the selection probability table is weighted according to the evaluation cost for each operating condition. The selection probability table and the cost-gradient selection probability table will be described in detail below.

[0029] The operating condition selection unit 126 uses the selection probability table or the cost-gradient selection probability table output by the selection probability update unit 125 to probabilistically select at least one operating condition from all the operating conditions accepted by the input / output unit 11. The operating condition selection unit 126 outputs the selected operating conditions to the condition setting unit 121 as an operating condition set.

[0030] It should be noted that information acquired from the components of the parameter adjustment device 1 is stored inside the input / output unit 11 and is displayed on the input / output device 3 after undergoing simple information processing as appropriate.

[0031] The input / output device 3 is a device that inputs and outputs data to and from the parameter adjustment device 1, and is, for example, a terminal device such as a computer, tablet terminal, or smartphone. Communication between the input / output device 3 and the input / output unit 11 may be wired or wireless. The parameter adjustment device 1 and the input / output device 3 may be included in the same terminal device.

[0032] [Configuration of Control System 2] FIG. 2 is a detailed configuration diagram of the control system 2 shown in FIG.

[0033] 2, the control system 2 is composed of a control unit 21 and a controlled object 22, such as a power supply device. The control system 2 controls its operation based on control parameters set by a condition setting unit 121 and signals based on operating conditions given by the condition setting unit 121.

[0034] The signal based on the operating conditions may be, for example, a target signal given to the control unit 21 or an operating mode that switches the behavior of the control unit 21, or a disturbance signal given to the controlled object 22. The target signal is, for example, a signal that includes information such as a target voltage output by the control system 2. The operating mode is, for example, a signal that switches the control state of the control system 2, such as a signal that switches the control state of the power supply device to either constant voltage control, constant current control, or constant power control. The disturbance signal is, for example, a signal that includes information such as input voltage fluctuation or load current fluctuation that is given to the power supply device.

[0035] The control unit 21 is realized by a microcomputer, a processor, etc. That is, the functions of the control unit 21 are realized by the microcomputer, the processor, etc. executing a program stored in a memory.

[0036] The control unit 21 receives control parameters, a target signal, and an operation mode from the condition setting unit 121. The control unit 21 also outputs a control input signal, which is a signal for controlling the operation of the controlled object 22, to the controlled object 22. The control parameters are, for example, a proportional gain, an integral gain, and a differential gain used in PID (Proportional-Integral-Differential) control.

[0037] The controlled object 22 is an object controlled by the control unit 21, such as a power supply circuit, a battery, or a load. The controlled object 22 outputs an internal signal to the evaluation unit 122. The internal signal may be, for example, a signal including information on the control amount with which the control unit 21 controls the controlled object 22, or may be the value of a signal that the control unit 21 cannot directly control or observe. The internal signal may also be a control input signal that the control unit 21 provides to the controlled object 22. Specifically, the internal signal is a signal including information such as an output voltage or an inductor current.

[0038] The operation of the parameter adjustment node 1 will be explained below using a flowchart illustrating the operation.

[0039] [Overall Operation] FIG. 3 is a flowchart showing the operation of the parameter adjustment node 1 according to this embodiment.

[0040] After the operation of the parameter adjustment device 1 is started, in step S201, the user inputs operating conditions, evaluation indices, and combinations of operating conditions and evaluation costs for each operating condition to the input / output device 3. The input / output unit 11 accepts the information input by the user and outputs the operating conditions to the condition setting unit 121, the evaluation indices to the evaluation unit 122, and combinations of operating conditions and evaluation costs for each operating condition to the selection probability update unit 125. The operating conditions may thereafter be managed uniformly within the parameter adjustment device 1 as, for example, a combination of an operating condition number and a signal to be provided to the control system 2.

[0041] In step S202, the selection probability update unit 125 initializes the search progress, the selection probability table, and the cost gradient selection probability table.

[0042] The search progress degree (hereinafter also referred to as the search progress degree) is a scalar value held by the selection probability update unit 125, and represents the degree of progress in the control parameter search performed by the parameter adjustment device 1. The search progress degree takes a value between 0 and 1. Note that the selection probability update unit 125 sets the search progress degree to 0 during initialization in step S202.

[0043] The selection probability table is a table of table values ​​held in the selection probability update unit 125, and indicates the probability that the operating condition selection unit 126 will select each operating condition for all operating conditions output to the selection probability update unit 125 by the input / output unit 11. Note that the selection probability update unit 125 sets all of the table values ​​to 1 during initialization in step S202.

[0044] The cost-gradient selection probability table is a table of values ​​held by the selection probability update unit 125. The probability of each operating condition set in the selection probability table is corrected based on the cost of each operating condition, increasing the probability if the cost is greater than a certain standard and decreasing the probability if the cost is smaller. Note that the selection probability update unit 125 sets all of the table values ​​to 1 during initialization in step S202.

[0045] In step S203, the condition setting unit 121 sets the control parameters to the control system 2 and sequentially provides all the acquired operating conditions to the control system 2. Here, the control parameters set by the condition setting unit 121 may be predetermined initial values. Then, the control system 2 sequentially executes control based on the operating conditions provided by the condition setting unit 121 and the set control parameters.

[0046] In step S204, the evaluation unit 122 observes the operating waveforms under each operating condition from the control system 2, calculates an evaluation value for each operating condition based on the acquired evaluation index, and outputs a combination of the operating condition and the evaluation value for each operating condition to the selection probability update unit 125.

[0047] The parameter adjustment node 1 may also perform step S209 or step S210, which will be described later, after step S204.

[0048] In step S205, the selection probability update unit 125 updates the selection probability table and the cost gradient selection probability table in accordance with the combination of the operating condition and the evaluation value obtained in step S204 or step S208 (described later).

[0049] In step S206, the operating condition selection unit 126 selects an operating condition with a probability based on the value of the cost-gradient selection probability table updated in step S205, creates an operating condition set, and outputs the set to the condition setting unit 121. The operating condition selection unit 126 creates an operating condition set that includes at least one operating condition. If the operating condition selection unit 126 creates an operating condition set that does not include an operating condition by selecting an operating condition based on the cost-gradient selection probability table, the operating condition selection unit 126 creates a new operating condition set by, for example, selecting the operating condition with the highest probability in the cost-gradient selection probability table.

[0050] In step S207, the condition setting unit 121 sequentially provides the operating conditions included in the operating condition set output by the operating condition selection unit 126 in step S206 to the control system 2. Then, the control system 2 sequentially executes control based on the operating conditions provided by the condition setting unit 121 and the set control parameters.

[0051] In step S208, the evaluation unit 122 observes the operating waveforms under each operating condition from the control system 2, calculates an evaluation value for each operating condition based on the acquired evaluation index, and outputs a combination of the operating condition and the evaluation value for each operating condition to the overall evaluation unit 123.

[0052] In step S209, the overall evaluation unit 123 obtains the maximum and minimum values ​​of each evaluation value calculated by the evaluation unit 122 in step S208, and calculates the average of the maximum and minimum values ​​as the overall evaluation value. Note that the method of calculating the average of the maximum and minimum values ​​as the overall evaluation value is one example of a method of calculating the overall evaluation value, and the overall evaluation value may also be an appropriately weighted average of the maximum and minimum values. Furthermore, instead of the maximum value, the overall evaluation value may also be a weighted average of multiple top evaluation values ​​including the maximum value, and instead of the minimum value, the overall evaluation value may also be a weighted average of multiple bottom evaluation values ​​including the minimum value.

[0053] In step S210, the optimization unit 124 calculates the next control parameters to be tried based on the black-box optimization algorithm, using the overall evaluation value calculated by the overall evaluation unit 123 in step S209 as the objective function value to be minimized, and outputs the control parameters to the condition setting unit 121.

[0054] In step S211, the condition setting unit 121 determines whether the search operation for the control parameters defined in steps S205 to S210 has been performed a certain number of times. If the condition setting unit 121 determines that the search operation has not been performed a certain number of times (No in step S211), the flowchart proceeds to step S205. On the other hand, if the condition setting unit 121 determines that the search operation has been performed a certain number of times (Yes in step S211), the flowchart proceeds to step S212.

[0055] In step S212, the input / output unit 11 outputs the best parameters calculated by the evaluation unit 122 based on the evaluation index for each operating condition selected in each search operation to the input / output device 3. The best parameters are defined as the control parameters that give the smallest (best) value among the history of the largest (worst) evaluation values ​​for the operating conditions selected in each search operation.

[0056] In addition, the selection probability update unit 125 may update the selection probabilities set in the selection probability table and the selection probability table with cost gradient so that they have the same value, without applying a cost gradient (i.e., a weight according to the evaluation cost) in step S205.

[0057] Alternatively, the selection probability update unit 125 may update the selection probability table in step S205, and the operating condition selection unit 126 may create and output an operating condition set based on the selection probability table in step S206.

[0058] As explained above, the overall evaluation unit 123 calculates the weighted average of the maximum and minimum values ​​of the multiple evaluation values ​​as the overall evaluation value, and thus can obtain a variety of overall evaluation values ​​in each search operation. This allows the optimization unit 124 to calculate the next control parameter to be tried using the overall evaluation value obtained in each search operation as a hint.

[0059] Furthermore, the overall evaluation unit 123 calculates the weighted average of a plurality of evaluation values, including the maximum and minimum evaluation values, as the overall evaluation value, so that it is possible to obtain a wider variety of overall evaluation values ​​in each search operation.

[0060] [Operation of Selection Probability Update Unit 125] Hereinafter, detailed operations performed by the selection probability update unit 125 in step S205 of FIG. 3 will be described with reference to the flowchart of FIG.

[0061] FIG. 4 is a flowchart showing the detailed operation of step S205 shown in FIG.

[0062] In step S251, the selection probability update unit 125 identifies the operating condition with the largest evaluation value from the combination of the operating conditions and the evaluation values ​​for each operating condition obtained in step S204 or step S208, and updates the selection probability table so that the selection probability of that operating condition is higher than the selection probabilities of the other operating conditions. Specifically, if the selection probability of the operating condition with the largest evaluation value is ps, the selection probability update unit 125 increases the selection probability ps to be higher than the selection probabilities of the other operating conditions by using an appropriate positive value cp, as shown in Equation 1 below.

[0063] ps←ps+cp... (Formula 1)

[0064] In step S252, the selection probability update unit 125 identifies the operation condition with the smallest evaluation value from the combinations of the operation conditions and the evaluation values ​​for each operation condition obtained in step S204 or step S208, and updates the selection probability table so that the selection probability of that operation condition increases or decreases according to the search progress level. Specifically, if the selection probability of the operation condition with the smallest evaluation value is p s , the selection probability update unit 125 updates the selection probability table using appropriate positive values ​​c p and c n and the search progress level α, as shown in Equation 2 below. Note that Equation 2 below increases the selection probability of the operation condition when the search progress level α is smaller than a threshold, and decreases the selection probability of the operation condition when the search progress level α is greater than the threshold. Furthermore, Equation 2 below sets the selection probability p s of the operation condition when the search progress level α is greater than the threshold to be lower than the selection probability p s of the operation condition when the search progress level α is smaller than the threshold.

[0065] ps←ps+(1-α)×cp-α×cn...(Formula 2)

[0066] In step S253, the selection probability update unit 125 updates the selection probability table so as to decrease the selection probability of an operating condition whose evaluation value is neither the maximum nor the minimum value, based on the combination of the operating condition and the evaluation value for each operating condition obtained in step S204 or step S208. Specifically, when the selection probability of an operating condition whose evaluation value is neither the minimum nor the maximum value is ps, the selection probability update unit 125 updates the selection probability ps using an appropriate positive value cn as shown in the following equation 3 so as to decrease the selection probability ps.

[0067] ps←ps−cn...(Formula 3)

[0068] The selection probability update unit 125 may perform the processes of steps S251, S252, and S253 in reverse order.

[0069] The updated selection probabilities in steps S251, S252, and S253 are values ​​ranging from the minimum probability ε (described later) to 1. The threshold value in step S252 is a value ranging from the minimum probability ε (described later) to 1.

[0070] In step S254, the selection probability update unit 125 updates the search progress degree. Specifically, the selection probability update unit 125 updates the search progress degree α using an update step Δα according to the following equation 4.

[0071] α←α+Δα...(Formula 4)

[0072] In step S255, the selection probability update unit 125 multiplies the probability for each operating condition in the selection probability table by a coefficient corresponding to the evaluation cost for that operating condition, and updates the cost-gradient selection probability table. Specifically, the selection probability update unit 125 sets the cost-gradient selection probability psc as shown in Equation 5 below, where psc is the selection probability for an operating condition stored in the selection probability table, C is the evaluation cost for that operating condition, and Cmin is the lowest evaluation cost among all operating conditions.

[0073] psc←ps×Cmin / C...(Formula 5)

[0074] The selection probability update unit 125 may adjust the magnitude of the cost-based correction coefficient Cmin / C according to the search progress degree α. For example, when the search progress degree α exceeds a threshold, the selection probability update unit 125 may set the correction coefficient Cmin / C to 1.

[0075] In step S256, the selection probability update unit 125 restricts the selection probability of the operating condition held in the selection probability table and the cost-graded selection probability table to a value between the minimum probability ε and 1. The minimum probability ε is a value that prevents the selection probability from becoming completely 0 and allows for a slight possibility that the operating condition will be selected again.

[0076] As explained above, the parameter adjustment node 1 performs a search operation using the operating conditions selected by the operating condition selection unit 126, and therefore the number of evaluations required for each search operation can be reduced.

[0077] Furthermore, since the selection probability table sets the probability of selecting a plurality of operating conditions including the maximum and minimum evaluation values, the operating condition selection unit 126 can select a variety of operating conditions. This allows the parameter adjustment device 1 to perform each search operation using a variety of operating conditions, and the overall evaluation unit 123 to obtain a variety of overall evaluation values.

[0078] Furthermore, in the early stages of a search, the operating condition selection unit 126 is more likely to select an operating condition with a minimum value, and the parameter adjustment device 1 therefore frequently performs search operations that include that operating condition. This allows the optimization unit 124 to calculate the next control parameter to be tried using the comprehensive evaluation value of various values ​​as a hint in the early stages of a search. Furthermore, in the middle and later stages of a search, the operating condition selection unit 126 is less likely to select an operating condition with a minimum value, and the parameter adjustment device 1 therefore rarely performs search operations that include that operating condition. This allows the parameter adjustment device 1 to reduce the number of evaluations required for each search operation.

[0079] Furthermore, the operating condition selection unit 126 is more likely to select operating conditions with low evaluation costs and less likely to select operating conditions with high evaluation costs, so the parameter adjustment device 1 can reduce the evaluation costs associated with each search operation.

[0080] Furthermore, the operating conditions selected by the operating condition selection unit 126 are changed depending on the progress of the search, so the parameter adjustment device 1 can use a variety of operating conditions in the search.

[0081] [Relationship Between Objective Function, Control Input, and Controlled Variable] The relationship between the objective function, control input, and controlled variable obtained by the operation of the parameter adjustment device 1 will be described below using Fig. 5A and Fig. 5B as examples. Fig. 5A is a diagram showing an example of an objective function corresponding to a plurality of operating conditions. Fig. 5B is a diagram showing fluctuations in the control input and controlled variable over time at points corresponding to the crosses a to i on the objective function in Fig. 5A.

[0082] 5A is a diagram showing an example of objective functions corresponding to a plurality of operating conditions cond0 to cond7. The objective function is a function that represents the correspondence between control parameters on the horizontal axis and their evaluation values ​​on the vertical axis. The control parameters shown on the horizontal axis are control parameters set in the control system 2 by the condition setting unit 121, and the evaluation values ​​shown on the vertical axis are evaluation values ​​calculated by the evaluation unit 122. Note that FIG. 5A shows the objective function obtained when the parameter adjustment device 1 adjusts two types of control parameters while fixing one of the control parameters.

[0083] The objective functions corresponding to operating conditions cond0 and cond1 are objective functions with a minimum evaluation value of approximately 1. The objective functions corresponding to operating conditions cond4, cond5, cond6, and cond7 are objective functions with a minimum evaluation value of approximately 3.7. The objective functions corresponding to operating conditions cond2 and cond3 are objective functions with a minimum evaluation value of approximately 5. Furthermore, among the objective functions with a minimum evaluation value of approximately 1, the objective function for which the control parameter value when the minimum evaluation value is obtained is approximately 0 corresponds to operating condition cond0, and the objective function for which the control parameter value when the minimum evaluation value is obtained is approximately 0.3 corresponds to operating condition cond1. Of the objective functions with a minimum evaluation value of approximately 3.7, the objective function whose control parameter value when the minimum evaluation value is obtained is approximately −0.3 corresponds to operating condition cond7, the objective function whose control parameter value when the minimum evaluation value is obtained is approximately 0 corresponds to operating condition cond6, the objective function whose control parameter value when the minimum evaluation value is obtained is approximately 0.3 corresponds to operating condition cond5, and the objective function whose control parameter value when the minimum evaluation value is obtained is approximately 0.9 corresponds to operating condition cond4. Of the objective functions with a minimum evaluation value of approximately 5, the objective function whose control parameter value when the minimum evaluation value is obtained is a negative value corresponds to operating condition cond2, and the objective function whose control parameter value when the minimum evaluation value is obtained is a positive value corresponds to operating condition cond3. Furthermore, the objective function max indicated by the dashed line is the objective function drawn with the largest evaluation value among the evaluation values ​​obtained for each control parameter from the objective functions corresponding to multiple operating conditions cond0 to cond7.

[0084] The crosses a, b, and c are marks attached to the operating condition cond3. The crosses d, e, and f are marks attached to the operating condition cond7. The crosses g, h, and i are marks attached to the operating condition cond1. Furthermore, the crosses a, d, and g are points where the control parameter value is 1, the crosses b, e, and h are points where the control parameter value is 2, and the crosses c, f, and i are points where the control parameter value is 3.

[0085] (a) to (i) in Fig. 5B are diagrams showing the control inputs and controlled variables corresponding to the crosses a to i attached to the objective function in Fig. 5A. The control input is a signal output by the control unit 21 to the controlled object 22, and in the example in Fig. 5B, it is set to saturate at a minimum value of -1 and a maximum value of 1. The controlled variable is the output value of the controlled object 22 that can be observed by the evaluation unit 122, and in the example in Fig. 5B, the control unit 21 controls the output value with a target value of 1.

[0086] 5B, the evaluation unit 122 calculates the integral value of the deviation between the controlled variable and the target value as the evaluation value. Specifically, the evaluation unit 122 calculates the total area enclosed by the waveform representing the controlled variable and the straight line representing the target value over a certain period of time as the evaluation value.

[0087] In the responses shown in (a), (b), (c), (e), (f), and (i) of FIG. 5B, the waveform of the control input repeatedly saturates. In such responses, even if the condition setting unit 121 slightly increases or decreases the value of the control parameter and sets it in the control system 2, the waveform of the control input saturates to 1 or -1. Therefore, since the saturation amount in the input control is similar, the subsequent movement of the control amount will also be similar. Furthermore, since the movement of the control amount output to the evaluation unit 122 is similar, the waveform of the subsequent control input also similarly repeatedly saturates to 1 or -1. For this reason, the objective function shown in FIG. 5A has a flat section in which the evaluation value does not change with respect to the control parameter.

[0088] Figure 5B (a), (b), and (c) show the evaluation value and response for operating condition cond3. As shown in these figures, the waveform of the control input repeatedly saturates for all control parameters, and the movement of the controlled variable is oscillatory. Due to these characteristics, the objective function for operating condition cond3 shown in Figure 5A has the maximum evaluation value for most control parameters. In this objective function, the control parameters that improve the response and reduce the evaluation value are distributed in a very narrow range. Note that operating condition cond2 has the same characteristics and objective function shape as operating condition cond3.

[0089] (d), (e), and (f) of FIG. 5B are diagrams showing the evaluation value and response of operating condition cond7. In (e) and (f) of FIG. 5B, the waveform of the control input repeatedly saturates, resulting in oscillatory behavior of the controlled variable. However, in (d) of FIG. 5B, the waveform of the control input is not constantly saturated, resulting in improved response. Furthermore, the amplitude of the controlled variable in (d) of FIG. 5B is smaller than that in (e) and (f) of FIG. 5B. Furthermore, the amplitude of the controlled variable in (e) and (f) of FIG. 5B is also smaller than that in (b) and (c) of FIG. 5B when the same control parameters are set. Due to these characteristics, the objective function for cond7 shown in FIG. 5A has a narrower flat section and a wider valley section than those for operating conditions cond2 and cond3. Furthermore, the evaluation values ​​obtained for most control parameters are smaller than those obtained for operating conditions cond2 and cond3. The operating conditions cond4, cond5, and cond6 also have similar properties and objective function shapes to the operating condition cond7.

[0090] (g), (h), and (i) of FIG. 5B show the evaluation value and response of operating condition cond1. In (i) of FIG. 5B, the waveform of the control input repeatedly saturates, resulting in oscillatory behavior of the controlled variable. However, in (g) and (h) of FIG. 5B, the waveform of the control input is not constantly saturated, resulting in improved response. Furthermore, the amplitude of the controlled variable in (g) and (h) of FIG. 5B is smaller than that in (i) of FIG. 5B. Furthermore, the amplitude of the controlled variable in (i) of FIG. 5B is also smaller than that in (f) of FIG. 5B when the same control parameters are set. Due to these characteristics, the objective function for cond1 shown in FIG. 5A has a narrower flat zone and a wider valley zone than those for operating conditions cond4 to cond7. Furthermore, for most control parameters, the evaluation value is smaller than those for operating conditions cond4 to cond7. The operating condition cond0 also has similar properties and objective function shapes to the operating condition cond1.

[0091] It is assumed that the intervals where the objective functions for the respective operating conditions form valleys and their minimum values ​​differ locally but roughly coincide globally.

[0092] 5A is the objective function that takes the maximum value among the objective functions of all operating conditions, and is the objective function that the parameter adjustment device 1 ultimately wants to minimize. The objective function drawn by the dotted line can also be considered as the evaluation value of the operating condition that is the bottleneck for a certain control parameter.

[0093] In addition, for most control parameters, the operating conditions cond2 and cond3 are the maximum values ​​related to the operating conditions, so simply thinking about it, if the parameter adjustment device 1 performs control execution, evaluation, and optimization only for the operating conditions cond2 and cond3, this is equivalent to performing control execution, evaluation, and optimization for the maximum values.

[0094] Comparative Example A comparative example of the parameter adjustment device 1 for the objective functions of the multiple operating conditions shown in FIGS. 5A and 5B will be described below with reference to the explanatory diagram of FIG.

[0095] FIG. 6 is a diagram illustrating an example of the transition of the selection probability of an operating condition and the overall evaluation value when the selection probability update unit 125 increases the selection probability of the operating condition corresponding to the maximum evaluation value, and the overall evaluation unit 123 sets the maximum of the evaluation values ​​obtained by adjustment as the overall evaluation value. (a) of FIG. 6 is a diagram illustrating a table showing the selection probability of an operating condition versus the "trial" representing the number of search trials for the control parameters. The selection probability for each "trial" is the selection probability set in the cost-graded selection probability table for each search trial. (b) on the left side of FIG. 6 is a diagram illustrating the transition of the evaluation value for each operating condition versus the "trial" representing the number of search trials for the control parameters. (c) on the right side of FIG. 6 is a diagram illustrating the evaluation values ​​displayed using a color bar. The evaluation values ​​shown in (b) on the left side of FIG. 6 are displayed in the colors shown in the color bar on the right side of FIG. 6. FIG. 6C is a diagram showing a color bar indicating the transition of only the maximum value (i.e., the overall evaluation value) among the evaluation values ​​for each operating condition in each trial shown in FIG. 6B.

[0096] Fig. 6 is an explanatory diagram of the operation of the parameter adjustment device 1 when the processing for achieving the effects of the present disclosure is omitted or changed. That is, the operation performed by the parameter adjustment device 1 is the same as when the processing of steps S252 and S254 of Fig. 4 is omitted, the selection probability of the operating condition that takes the minimum value in step S253 of Fig. 4 is also lowered, the maximum value rather than the average value is used as the comprehensive evaluation value in step S209 of Fig. 3, and the cost gradient is not applied in step S255 of Fig. 4, and the selection probabilities set in the selection probability table and the cost-graded selection probability table are changed to the same value.

[0097] As shown in Fig. 6(a), it can be seen that the selection probabilities of all operating conditions at the beginning of the search are close to 1. This is because the selection probability update unit 125 initializes the selection probabilities set in the selection probability table and the cost-graded selection probability table to 1 in step S202 of Fig. 3. As a result, as shown in Fig. 6(b), at the beginning of the search, the parameter adjustment device 1 selects and evaluates almost all operating conditions.

[0098] However, as shown in (b) of FIG. 6 , due to the influence of the objective function shape shown in (a) of FIG. 5 , the evaluation values ​​of operating conditions cond2 and cond3 are always larger than those of the other operating conditions. In such a case, the selection probability update unit 125 performs the processes of steps S251 and S253 of FIG. 4 (however, the selection probability of the operating condition with the smallest value is also lowered). As a result, as shown in (a) of FIG. 6 , the selection probability update unit 125 updates the selection probability table and the cost-gradient selection probability table so that the selection probability of operating condition cond2 or cond3 increases and the selection probability of the other operating conditions decreases. However, because the selection probability is constrained to the range from the minimum probability ε to 1 by the process of step S256, it never takes a value greater than 1 or less than the minimum probability ε.

[0099] The operating condition selection unit 126 selects operating conditions based on the values ​​of the updated selection probability table and cost-gradient selection probability table (i.e., the process of step S206 in FIG. 3 is performed). As a result, as shown in FIG. 6B, the frequency with which operating conditions other than operating conditions cond2 and cond3 are evaluated decreases as the search progresses. As a result, the parameter adjustment device 1 performs control execution, evaluation, and optimization only for the operating conditions that are bottlenecks and that are likely to produce the overall evaluation value (i.e., the maximum value) shown in FIG. 6C.

[0100] However, this parameter adjustment method requires optimization for the nearly flat objective functions of the operating conditions cond2 and cond3 shown in FIG. 5A, making the search problem extremely difficult. In fact, as shown in FIG. 6C, it can be seen that the overall evaluation value remains almost constant up to the point of 200 trial. The black-box optimization algorithm used by the optimization unit 124 is an algorithm that estimates the control parameter region that minimizes the objective function using the obtained evaluation value (the overall evaluation value in this disclosure) as a hint. Therefore, if the black-box optimization algorithm can only obtain the same evaluation value, it cannot obtain a hint, and therefore the optimization unit 124 can only perform the same operation as if it were performing a random search without using the black-box optimization algorithm. As a result, the above-mentioned parameter adjustment method results in a stagnation in the search for optimal control parameters.

[0101] [Operation Example 1] An example of how to address the problem of search stalling will be described below as Operation Example 1 using FIG. 7 . FIG. 7 is a diagram illustrating an example of the transitions in the selection probability of an operating condition and the overall evaluation value when the selection probability update unit 125 increases the selection probability of an operating condition corresponding to the maximum and minimum evaluation values, and the overall evaluation unit 123 sets the average of the maximum and minimum evaluation values ​​obtained by adjustment as the overall evaluation value. FIG. 7 is an explanatory diagram illustrating an example of the operation of the parameter adjustment device 1 for the objective function of multiple operating conditions shown in FIGS. 5A and 5B . Note that the parameter adjustment device 1 described in FIG. 7 differs from the parameter adjustment device 1 described in FIG. 6 in that it adds the processing of steps S252 and S254 of FIG. 4 , does not lower the selection probability of an operating condition with the minimum value in step S253 of FIG. 4 , and instead sets the average value as the overall evaluation value in step S209 of FIG. 3 .

[0102] 7A and 7B are diagrams showing the same contents as FIG. 6A and FIG. 6B, respectively. FIG. 7C is a diagram showing a color bar indicating the transition of only the maximum value of the evaluation values ​​for each operating condition in each trial shown in FIG. 7B. FIG. 7D is a diagram showing the transition of the selection probability table and the transition of the search progress degree updated in step S254 of FIG. 4 for the trial representing the number of search trials for the control parameters. FIG. 7E is a diagram showing the transition of the selected operating condition and its evaluation value, as well as the transition of the overall evaluation value (the average value of the maximum and minimum values ​​in FIG. 7) calculated in step S209 for the trial representing the number of search trials for the control parameters.

[0103] As shown in (b) of Figure 7, due to the influence of the objective function shape shown in Figure 5A, the evaluation values ​​of operating conditions cond2 and cond3 are larger than those of the other operating conditions, and the evaluation values ​​of operating conditions cond0 and cond1 are smaller than those of the other operating conditions. As shown in (a) of Figure 7, in the trial when the search progress level α shown in (d) of Figure 7 is low (i.e., in the early stage of the search), the selection probability update unit 125 performs the process of step S252 of Figure 4. That is, unlike the case of (a) of Figure 6, the selection probability update unit 125 updates not only the selection probabilities of operating conditions cond2 and cond3 but also the selection probabilities of operating conditions cond0 and con1 so as to be higher. As a result, in the early stage of the search, the selection probabilities of operating conditions cond0 to cond3 remain near 1, and the selection probabilities of the other operating conditions are lower.

[0104] It can be seen that the selection probabilities close to 1 for operating conditions cond0 and cond1 are maintained up to about the 100th trial point when the search progress level α shown in Fig. 7(d) becomes 1. As a result, as shown in Fig. 7(b), the parameter adjustment device 1 evaluates operating conditions cond0 and cond1, as well as operating conditions cond2 and cond3, in most search operations up to about the 100th trial point.

[0105] In this parameter adjustment method, operating conditions cond0 and cond1 that are likely to yield the minimum value are selected and evaluated in step S252 of Fig. 4 and step 206 of Fig. 3, and an overall evaluation value is calculated using the average value in step S209 of Fig. 3. As a result, in this parameter adjustment method, the evaluation value shown in Fig. 7(c) and the overall evaluation value shown in Fig. 7(e) are improved with fewer search attempts than the evaluation value (overall evaluation value) shown in Fig. 6(c). In other words, in this parameter adjustment method, the evaluation values ​​of operating conditions cond2 and cond3, which were bottlenecks, and the evaluation value shown in Fig. 7(c) can be improved with fewer search attempts than in the parameter adjustment method described in Fig. 6.

[0106] This is because including the evaluation values ​​of operating conditions cond0 and cond1, which have few flat regions in the objective function value, in the overall evaluation value prevents a situation in which only a constant overall evaluation value is obtained, and narrows the search region for control parameters calculated by the optimization unit 124 to the vicinity of the true optimal solution. In fact, it can be seen that the overall evaluation value shown in (e) of Figure 7 takes on a variety of values ​​even while the evaluation values ​​of operating conditions cond2 and cond3 shown in (b) of Figure 7 are constant.

[0107] Thereafter, when the search progress level α shown in (d) of Fig. 7 becomes 1, the above formula 2 becomes equivalent to the above formula 3. Therefore, similar to the selection probability of the operating conditions cond0 and cond1, which are likely to have the smallest evaluation value shown in (a) of Fig. 6, the selection probability of the operating conditions cond0 and cond1 shown in (a) of Fig. 7 decreases as the search progresses. As a result, as shown in (b) of Fig. 7, the frequency with which the evaluation unit 122 evaluates the operating conditions cond0 and cond1 also decreases as the search progresses.

[0108] In this way, this parameter adjustment method prevents the search from stagnating due to only obtaining a certain evaluation value in the early stages of the search, and thereafter, it is possible to narrow down the operating conditions that become bottlenecks and perform control execution, evaluation, and optimization.

[0109] Furthermore, the overall evaluation unit 123 calculates the weighted average of the maximum and minimum values ​​of the multiple evaluation values ​​as the overall evaluation value, thereby obtaining a variety of overall evaluation values ​​in each search operation. This allows the optimization unit 124 to use the overall evaluation value obtained in each search operation as a hint to calculate the next control parameter to be tried. This allows the parameter adjustment device 1 to avoid adjustment stagnation at the beginning of the search.

[0110] Furthermore, in the early stages of a search, the operating condition selection unit 126 is likely to select the operating condition with the smallest value, so the parameter adjustment device 1 will often perform search operations that include that operating condition. This allows the optimization unit 124 to calculate the next control parameter to be tried using the comprehensive evaluation value of various values ​​as a hint in the early stages of a search, allowing the parameter adjustment device 1 to avoid adjustment stagnation. Furthermore, in the middle and later stages of a search, the operating condition selection unit 126 is unlikely to select the operating condition with the smallest value, so the parameter adjustment device 1 will rarely perform search operations that include that operating condition. This allows the parameter adjustment device 1 to reduce the number of evaluations required for each search operation.

[0111] [Operation Example 2] However, in the parameter adjustment method described in Operation Example 1, the total evaluation cost required for adjusting the control parameters may increase depending on the evaluation cost for each operating condition. Specifically, since the evaluation cost for each operating condition is not necessarily the same, actively selecting operating conditions with high evaluation costs may result in a problem of increasing the total evaluation cost. The parameter adjustment device 1 is required to minimize the overall evaluation value with as little evaluation cost as possible.

[0112] An example of how to address the above-mentioned problem will be described below as Operation Example 2 using FIGS. 8 and 9 . FIG. 8 is a diagram showing an example of evaluation costs set for each operating condition. FIG. 9 is a diagram showing an example of the transition of the selection probability of an operating condition and the overall evaluation value when the selection probability update unit 125 further performs the process of step S255 in FIG. 4 to reduce the total evaluation cost in the search. The evaluation cost of each operating condition shown in FIG. 9 is the evaluation cost shown in FIG. 8 . FIG. 9 is also an explanatory diagram showing an example of the operation of the parameter adjustment device 1 with respect to the objective functions of the multiple operating conditions shown in FIGS. 5A and 5B . Note that (a), (b), (c), (d), and (e) in FIG. 9 are diagrams showing the same content as (a), (b), (c), (d), and (e) in FIG. 7 , respectively.

[0113] It is assumed that the evaluation of the operating conditions requires evaluation costs as shown in Fig. 8. Specifically, the evaluation cost of even-numbered operating conditions is 1, and the evaluation cost of other operating conditions is 2.

[0114] Unlike the selection probability of odd-numbered operating conditions shown in (a) of Figure 7, which fluctuates within the range of 0 to 1 throughout the entire search, the selection probability of odd-numbered operating conditions shown in (a) of Figure 9 fluctuates within the range of 0 to 0.5 throughout the entire search. This is because the cost-gradient selection probability table updated based on the above-mentioned equation 5 in step S255 of Figure 4 is reflected. In other words, since the smallest evaluation cost Cmin among the operating conditions is 1 and the evaluation cost C of odd-numbered operating conditions is 2, the selection probability update unit 125 updates the cost-gradient selection probability table by multiplying the selection probability of odd-numbered operating conditions in the selection probability table by Cmin / C = 1 / 2 = 0.5.

[0115] As a result, as shown in (b) of FIG. 9, the selection probability of operating condition cond3 is lower than the selection probability of operating condition cond2, which has a similar objective function. Also, the selection probability of operating condition cond1 is lower than the selection probability of operating condition cond0, which has a similar objective function. Similarly, the selection probability of operating conditions cond5 and cond7 is lower than the selection probability of operating conditions cond4 and cond6, which have similar objective functions. However, the evaluation values ​​shown in (c) of FIG. 9 are improved from the beginning of the search, and, similar to the evaluation values ​​shown in (c) of FIG. 7, they prevent stagnation in the early stages of the search.

[0116] As a result, this parameter adjustment method can reduce the total evaluation cost in the search while avoiding the evaluation of operating conditions that have a large evaluation cost.

[0117] Furthermore, the operating condition selection unit 126 is more likely to select operating conditions with low evaluation costs and less likely to select operating conditions with high evaluation costs, so the parameter adjustment device 1 can reduce the evaluation costs associated with each search operation.

[0118] The selection probability update unit 125 may correct the evaluation cost according to the search progress degree α. For example, in the early stage of the search, the selection probability update unit 125 updates the cost-gradient selection probability table based on the above formula 5 in order to avoid evaluating operating conditions with high evaluation costs. On the other hand, in the later stage of the search, the selection probability update unit 125 corrects the evaluation cost in order to improve the accuracy of the control parameters, and updates the cost-gradient selection probability table so that the operating condition selection unit 126 can select many operating conditions that are important for adjusting the control parameters regardless of the evaluation cost.

[0119] This allows the parameter adjustment device 1 to use a variety of operating conditions in the search, thereby improving the accuracy of the control parameters.

[0120] [Relationship Between Accumulative Evaluation Cost and Evaluation Value] The relationship between the total amount of evaluation cost required for the comparative example, operation example 1, operation example 2, and other parameter adjustment methods and the maximum value of the obtained evaluation value will be described below with reference to FIG. 10. FIG. 10 is a diagram showing the relationship between the cumulative evaluation cost required for each parameter adjustment method and the maximum value of the obtained evaluation value. FIG. 10 is a diagram showing a table in which the horizontal axis represents the cumulative evaluation cost and the vertical axis represents the maximum value of the obtained evaluation value. The cumulative evaluation cost is a numerical value obtained by adding up all the evaluation costs required for each search operation.

[0121] The dashed-dotted line in FIG. 10 represents the relationship obtained when the parameter adjustment method of the comparative example is used. The solid line represents the relationship obtained when the parameter adjustment method of the operational example 1 is used. The two-dot-dash line represents the relationship obtained when the parameter adjustment method of the operational example 2 is used. The dotted line represents the relationship obtained when a parameter adjustment method that omits steps S251 and S256 in FIG. 4 is used among the operations performed by the parameter adjustment method of the comparative example. In other words, the parameter adjustment method that obtains the dotted line is an adjustment method that always controls and executes all operating conditions in the search operation, and evaluates the maximum value of the obtained evaluation values ​​as the overall evaluation value. The dashed line represents the relationship obtained when a parameter adjustment method that always controls and executes all operating conditions in the search operation, and evaluates the maximum value of the obtained evaluation values ​​as the overall evaluation value in step S209 in FIG. 3 is used among the operations performed by the parameter adjustment method that obtains the dotted line. In other words, the parameter adjustment method that obtains the dashed line is an adjustment method that always controls and executes all operating conditions in the search operation, and evaluates the average value of the maximum and minimum values ​​of the obtained evaluation values ​​as the overall evaluation value.

[0122] In the parameter adjustment methods that obtain the solid line, the two-dot chain line, and the dashed line, the average of the maximum and minimum of the obtained evaluation values ​​is evaluated as the overall evaluation value. In contrast, in the parameter adjustment methods that obtain the one-dot chain line and the dotted line, the maximum of the obtained evaluation values ​​is evaluated as the overall evaluation value. For example, when adjusting control parameters with the goal of reducing the maximum of the obtained evaluation values ​​by the parameter adjustment device 1 to 6 or less, a parameter adjustment device 1 using one of the former three parameter adjustment methods achieves this goal with a lower cumulative evaluation cost than a parameter adjustment device 1 using one of the latter two parameter adjustment methods. This is because including evaluation values ​​for operating conditions with few flat regions in the objective function value in the overall evaluation value prevents a situation in which only a certain evaluation value is obtained, and narrows the search range of the control parameters calculated by the optimization unit 124 to near the true optimal solution. As a result, a parameter adjustment device 1 using one of the former three parameter adjustment methods can avoid adjustment stagnation at the initial stage of the search.

[0123] Furthermore, in the parameter adjustment method when the dashed line is obtained, all operating conditions are controlled and executed in each search operation. On the other hand, in the parameter adjustment method when the solid line is obtained, operating conditions selected based on a cost-gradient selection probability table set to the same value as the selection probability of each operating condition set in the selection probability table are controlled and executed in each search operation. As a result, a parameter adjustment device 1 using the latter parameter adjustment method can reduce the number of evaluations required for each search operation compared to a parameter adjustment device 1 using the former parameter adjustment method. Therefore, a parameter adjustment device 1 using the latter parameter adjustment method can reduce the evaluation cost required for each search operation compared to a parameter adjustment device 1 using the former parameter adjustment method.

[0124] Furthermore, in the parameter adjustment method when the two-dot-dash line is obtained, the operating conditions selected based on a cost-graded selection probability table in which the selection probability of each operating condition set in the selection probability table is multiplied by a coefficient corresponding to the evaluation cost for each operating condition are controlled and executed in each search operation. Thus, unlike the parameter adjustment method when the solid line is obtained, the parameter adjustment method when the two-dot-dash line is obtained can reduce the total evaluation cost in the search while avoiding the evaluation of operating conditions with high evaluation costs. Therefore, the parameter adjustment device 1 using the former parameter adjustment method can reduce the evaluation cost for each search operation more than the parameter adjustment device 1 using the latter parameter adjustment method.

[0125] [Image Displayed on Input / Output Device 3] Next, an image (i.e., a UI diagram) displayed on the input / output device 3 will be described with reference to FIG. 11 . FIG. 11 is a diagram showing an example of an image displayed on the input / output device 3 by the input / output unit 11. As shown in FIG. 11 , the image includes a file selection button 401, an operating condition setting area 402, an evaluation index setting area 403, a cost setting area 404, an adjustment execution button 405, an adjustment status display area 406, an adjustment result display area 407, a comprehensive evaluation value display area 408, a maximum evaluation value display area 409, a selection probability table display area 410, a selection probability table with cost gradient display area 411, and a best parameter display area 412. Note that the image does not need to include all of the above-mentioned components, and may instead display only some of the components.

[0126] The user can provide pre-set operating conditions to the input / output unit 11 via the file selection button 401. In the example of Fig. 11, the input / output unit 11 reads a setting file and accepts eight operating conditions.

[0127] The user can set the operating conditions that the input / output unit 11 outputs to the condition setting unit 121 via the operating condition setting area 402. In the example of Fig. 11, for each of the eight operating conditions, the shape of the input signal that the condition setting unit 121 gives to the control system 2 is set by reading a setting file.

[0128] The user can set an evaluation index for the evaluation unit 122 to evaluate the operating waveform via the evaluation index setting area 403. In the example of Fig. 11, the amount of overshoot when an input signal is applied using pull-down is set as the evaluation index.

[0129] The user can set the evaluation cost of each operating condition set in the operating condition setting area 402 via the cost setting area 404. This setting may be set directly by the user using a bar graph as a slider UI, or may be set automatically by the parameter adjustment device 1 based on the length of the input signal, etc.

[0130] The user can execute the adjustment operation of the parameter adjustment device 1 via the adjustment execution button 405 based on the operating conditions set in the operating condition setting area 402, the evaluation index set in the evaluation index setting area 403, and the evaluation cost set in the cost setting area 404.

[0131] In the adjustment status display area 406, an image showing the control parameters set in the control unit 21 by the condition setting unit 121 and the operating conditions given to the control system 2 is displayed in real time.

[0132] In the adjustment result display area 407, the horizontal axis represents each trial from the start of adjustment of the control parameters to the present, and the vertical axis represents the operating conditions, and an image is displayed showing, in a heat map, the evaluation values ​​calculated by the evaluation unit 122 for the operating conditions set by the condition setting unit 121. However, the evaluation value of an operating condition that the condition setting unit 121 did not set in the control system 2 in a certain trial is displayed as blank.

[0133] In the overall evaluation value display area 408, the horizontal axis represents each trial from the start of the control parameter adjustment to the present, and an image showing the overall evaluation value calculated by the overall evaluation unit 123 in a heat map is displayed.

[0134] In the maximum evaluation value display area 409, the horizontal axis represents each trial from the start of control parameter adjustment to the present, and an image showing the maximum (worst) evaluation value for the operating conditions in each search trial in a heat map is displayed.

[0135] In the selection probability table display area 410 and the cost gradient selection probability table display area 411, each trial from the start of control parameter adjustment to the present is placed on the horizontal axis, and an image is displayed showing the progress of the selection probability of the selection probability table and the selection probability of the cost gradient selection probability table held by the selection probability update unit 125 in the form of a line graph.

[0136] The best parameter display area 412 displays the evaluation value when the smallest value was obtained among the history of the largest evaluation values ​​for the operating conditions selected in each search operation from the start of control parameter adjustment to the present, and the control parameter that produced that result.

[0137] To make it easier to understand the correspondence between operating conditions, the operating condition setting area 402, the cost setting area 404, and the adjustment result display area 407 may be displayed so that the vertical positions of the operating conditions displayed in each UI diagram match.

[0138] As explained above, the parameter adjustment device 1 can output the optimal control parameters obtained by the search to the input / output device 3.

[0139] [Modification] Below, a modification of the operation example 2 of the embodiment will be described. In the following modification, an example will be described in which the correction coefficient varies depending on the search progress degree α. Note that the multiple operating conditions cond0 to cond7 shown in the following description have the same objective function as the multiple operating conditions cond0 to cond7 shown in Figures 5A and 5B.

[0140] [Modification of the Operation of the Selection Probability Update Unit 125] In step S254 of Fig. 4, the selection probability update unit 125 may update the search progress rate α according to the following equations 6 and 7 instead of the above equation 4. Note that α shown in equation 6 ｋ is the estimated pass rate in the search trial number k (k is a natural number), Np is the number of operating conditions that are estimated to pass when the search trial number k is reached, and N is the total number of operating conditions. ｋ is a symbol indicating the maximum value for the number of search trials k.

[0141] α ｋ ←Np / (N-1)...(Formula 6)

[0142] α←max ｋ {α ｋ}...(Formula 7)

[0143] The selection probability update unit 125 uses the above formulas 6 and 7 to calculate the estimated pass rate α ｋ Here, the calculation of the estimated pass rate will be explained using the specific example shown in FIG. 12. FIG. 12 is a table showing an example of the estimated pass / fail results of a plurality of operating conditions cond0 to cond7 in a certain number of search trials k. In the example of FIG. 12, the operating conditions actually given to the control system 2 are cond2, cond3, cond4, and cond6.

[0144] As shown in Figure 12, the table shows the operating conditions, the selection probability of each operating condition in search attempt number k, the pass / fail of the operating conditions controlled and executed in search attempt number k, and the estimated pass / fail of each operating condition in search attempt number k.

[0145] The pass / fail column indicates whether the operating condition controlled and executed when search attempt count k is passed or failed. For example, if the evaluation value obtained by controlled execution of a certain operating condition is below the reference value, the operating condition passes, and if the evaluation value obtained by controlled execution of a certain operating condition is equal to or greater than the reference value, the operating condition fails.

[0146] The estimated pass / fail column shows pass / fail results including those for which control was not executed. Specifically, the pass / fail result for an operating condition for which control was not executed in the number k of search attempts is considered to be the pass / fail result of the operating condition with the next lowest selection probability after the operating condition for which control was not executed among the operating conditions for which control was executed in that number of search attempts. For example, the pass / fail result of cond7 is considered to be the pass / fail result (fail) of cond6, which has the next lowest selection probability after cond7, and the pass / fail results of cond0, cond1, and cond5 are considered to be the pass / fail result (pass) of cond4, which has the next lowest selection probability after cond0, cond1, and cond5.

[0147] Therefore, in the specific example of FIG. 12, the estimated pass rate α ｋ becomes:

[0148] α ｋ ←4 / (8-1)=4 / 7

[0149] 4, the selection probability update unit 125 may set the cost-graded selection probability psc as shown in the following equation 8 instead of the above equation 5. Note that β shown in equation 8 is a correction coefficient.

[0150] psc←ps×β...(Formula 8)

[0151] 13A and 13B are diagrams showing the relationship between the search progress degree α and the correction coefficient β, where (a) in FIG. 13A is a diagram showing the relationship between the search progress degree α and the cost-based correction coefficient Cmin / C in operation example 2, and (b) in FIG. 13B is a diagram showing the relationship between the search progress degree α and the correction coefficient β in a modified example of operation example 2.

[0152] 13A, the cost-based correction coefficient Cmin / C is a constant value regardless of the search progress level α. For example, when the evaluation cost is 1, the correction coefficient Cmin / C is 1, and when the evaluation cost is 2, the correction coefficient Cmin / C is 0.5.

[0153] As shown in (b) of Figure 13, for an operating condition with the minimum evaluation cost (the evaluation cost is 1 in (b) of Figure 13), the selection probability update unit 125 sets the correction coefficient β to 1. For an operating condition with a non-minimum evaluation cost (the evaluation cost is 2 in (b) of Figure 13), the selection probability update unit 125 varies the correction coefficient β between 0 and 1 according to the search progress degree α. In other words, for an operating condition with a non-minimum evaluation cost, the selection probability update unit 125 varies the correction coefficient β between 0 and 1 according to the search progress degree α.

[0154] As explained above, in the modified example of operation example 2, the parameter adjustment device 1 sets the selection probability of the operating conditions so that it is difficult to select operating conditions with high evaluation costs in the early stages of the search for control parameters, and it is easy to select operating conditions with high evaluation costs in the later stages of the search for control parameters. This allows the parameter adjustment device 1 to use a variety of operating conditions in the search.

[0155] The correction coefficient β is not limited to increasing linearly as the search progress level α increases, but may instead increase in a curved manner as the search progress level α increases.

[0156] [Comparison of Selection Probabilities of Operating Conditions] An example of the transitions in the selection probability of an operating condition and the overall evaluation value when the selection probability update unit 125 changes the correction coefficient β according to the search progress level α will be described below with reference to FIG. 14 . FIG. 14 is also an explanatory diagram showing an example of the operation of the parameter adjustment device 1 with respect to the objective functions of the multiple operating conditions shown in FIGS. 5A and 5B . Note that (a), (b), (c), (d), and (e) of FIG. 14 are diagrams showing the same content as (a), (b), (c), (d), and (e) of FIG. 9 , respectively. Furthermore, as in the description of Operation Example 2, it is assumed that the evaluation of the operating conditions shown in FIG. 14 incurs the evaluation costs shown in FIG. 8 .

[0157] Unlike the selection probability of odd-numbered operating conditions shown in (a) of FIG. 9, which fluctuates within the range of 0 to 1 throughout the entire search, the selection probability of odd-numbered operating conditions shown in (a) of FIG. 14 fluctuates around 0 in the early stages of the search (e.g., when the trial is less than 20), and rises to around 1 in the middle and later stages of the search (e.g., when the trial is 20 or more), fluctuating within the range of 0 to 1. This is because the cost-graded selection probability table updated based on the above-mentioned Equation 8 in step S255 of FIG. 4 is reflected. Specifically, this is because the value of the correction coefficient β increases with the increase in the search progress level α shown in (d) of FIG.

[0158] As a result, as shown in FIG. 14B , the selection probability of operating condition cond3 remains lower than the selection probability of operating condition cond2, which has a similar objective function, when trial is less than 20, but remains close to the selection probability of operating condition cond2 (near 1) when trial is 50 or greater. Furthermore, the selection probability of operating condition cond1 remains lower than the selection probability of operating condition cond0, which has a similar objective function, when trial is less than 20, but remains close to the selection probability of operating condition cond0 (within the range of 0 to 1) when trial is 50 or greater. Similarly, the selection probabilities of operating conditions cond5 and cond7 remain lower than the selection probabilities of operating conditions cond4 and cond6, which have similar objective functions, when trial is less than 20, but remains close to the selection probabilities of operating conditions cond4 and cond6 (within the range of 0 to 1) when trial is 50 or greater. 14(c) is improved from the beginning of the search, and has the same effect as the evaluation value shown in FIG. 9(c), preventing stagnation in the early stage of the search. In other words, the selection probability update unit 125 sets the selection probability table so that, according to the progress of the search, the selection probability of the operating conditions (cond2 and cond3) that are not below the reference value among the multiple operating conditions (cond0 to cond7) is increased.

[0159] As a result, this parameter adjustment method can reduce the total evaluation cost in the search while avoiding the evaluation of operating conditions that have a large evaluation cost.

[0160] [Comparison of Relationship Between Accumulative Evaluation Cost and Evaluation Value] The relationship between the cumulative evaluation cost and the evaluation value for Operation Example 2 and a modified example of Operation Example 2 will be described below with reference to FIG. 15 . FIG. 15 is a diagram showing the transition of the evaluation value when focusing on operation conditions cond2 and cond3. (a) of FIG. 15 is a diagram showing the transition of the evaluation value obtained in Operation Example 2. (b) of FIG. 15 is a diagram showing the transition of the evaluation value obtained in a modified example of Operation Example 2. (a) of FIG. 15 and (b) of FIG. 15 are diagrams in which the horizontal axis represents the cumulative evaluation cost and the vertical axis represents the evaluation value. Also, (c) of FIG. 15 is a diagram showing the transition of the search progress degree α with respect to the cumulative evaluation cost.

[0161] 15(a) and 15(b), the number of times the parameter adjustment node 1 searches using operating condition cond3 before the cumulative evaluation cost reaches 2500 is greater in the modified version of operation example 2 than in operation example 2. In other words, the number of times the parameter adjustment node 1 simultaneously selects operating conditions cond2 and cond3 before the cumulative evaluation cost reaches 2500 is greater in the modified version of operation example 2 than in operation example 2. Furthermore, in operation example 2, the maximum value (dashed line) of the obtained evaluation values ​​does not fall below the reference value (6.0 in the example of FIG. 15(a)) when the cumulative evaluation cost is 2500, whereas in the modified version of operation example 2, the maximum value (dashed line) of the obtained evaluation values ​​falls below the reference value (6.0 in the example of FIG. 15(b)) when the cumulative evaluation cost is approximately 2000.

[0162] As described above, the parameter adjustment device 1 can increase the number of times it searches for control parameters that simultaneously improve the operating conditions cond2 and cond3, which are bottlenecks where maximum values ​​are likely to appear, in the modified version of operation example 2 compared to operation example 2. That is, in the modified version of operation example 2, the parameter adjustment device 1 increases the number of times it simultaneously selects the operating conditions cond2 and cond3, which are bottlenecks where maximum values ​​are likely to appear, thereby increasing the number of times it searches for control parameters that simultaneously improve the operating conditions cond2 and cond3. As a result, in the modified version of operation example 2, the parameter adjustment device 1 can search for control parameters that will make the maximum value of the obtained evaluation values ​​less than (satisfy) the reference value, with less evaluation cost.

[0163] [Modification of image displayed on input / output device 3] Next, a modification of the image (i.e., UI diagram) displayed on the input / output device 3 will be described with reference to Fig. 16. Fig. 16 is a diagram showing a modification of the image shown in Fig. 11. In the description of Fig. 16, the same components included in the image shown in Fig. 11 are denoted by the same reference numerals and description thereof will be omitted, and components not included in the image shown in Fig. 11 are denoted by new reference numerals and description will be omitted.

[0164] As shown in FIG. 16, the image differs from the image shown in FIG. 11 in that it includes a pass criteria input area 413 and a pass rate display area 414 .

[0165] The user can provide the pass criteria for the set operating conditions to the input / output unit 11 via the pass criteria input area 413. This causes the parameter adjustment device 1 to search for control parameters that will cause the evaluation value of each set operating condition to fall below the pass criteria (6.0 in the example of FIG. 16 ).

[0166] In the pass rate display area 414, each trial from the start of control parameter adjustment to the present is placed on the horizontal axis, and an image is displayed showing the progress of the search α calculated by the selection probability update unit 125 using the above equations 6 and 7 in the form of a line graph.

[0167] As explained above, the parameter adjustment device 1 can accept the pass criteria specified by the user via the input / output device 3 .

[0168] [Effects] As described above, the parameter adjustment device 1 according to this embodiment is a parameter adjustment device 1 that searches for optimal control parameters for operating the control system 2 under a plurality of operating conditions by operating the control system 2 while adjusting the control parameters, and includes: a condition setting unit 121 that sets at least one operating condition and a control parameter among the plurality of operating conditions to the control system 2; an evaluation unit 122 that calculates, for each of the at least one operating condition, an evaluation value related to the operation of the control system 2 when the control system 2 operates under the at least one operating condition and the control parameter set by the condition setting unit 121; a comprehensive evaluation unit 123 that calculates, as a comprehensive evaluation value, a weighted average value obtained using at least the maximum and minimum values ​​of the evaluation values ​​calculated by the evaluation unit 122; and an optimization unit 124 that calculates, using an optimization algorithm, control parameters for the next operation of the control system 2 based on the comprehensive evaluation value calculated by the comprehensive evaluation unit 123. The condition setting unit 121 sets at least one operating condition and the control parameter calculated by the optimization unit 124 to the control system 2.

[0169] With this configuration, the overall evaluation unit 123 calculates the weighted average of the maximum and minimum values ​​of the multiple evaluation values ​​as the overall evaluation value, thereby enabling a variety of overall evaluation values ​​to be obtained in each search operation. This allows the optimization unit 124 to use the overall evaluation value obtained in each search operation as a hint to calculate the next control parameter to be tried. Therefore, the parameter adjustment device 1 can avoid adjustment stagnation at the beginning of the search, thereby reducing the time required for performance evaluation using simulation or an actual machine in the automatic adjustment of control parameters for multiple operating conditions.

[0170] Furthermore, in the parameter adjustment device 1 according to this embodiment, the overall evaluation unit 123 calculates the overall evaluation value as a weighted average value obtained using a plurality of upper evaluation values ​​including the maximum value among the evaluation values ​​and a plurality of lower evaluation values ​​including the minimum value among the evaluation values.

[0171] With this configuration, the overall evaluation unit 123 calculates the weighted average of multiple evaluation values, including the maximum and minimum evaluation values, as the overall evaluation value, thereby enabling a wider variety of overall evaluation values ​​to be obtained in each search operation. This allows the optimization unit 124 to use the overall evaluation value obtained in each search operation as a hint to calculate the next control parameter to be tried. Therefore, the parameter adjustment device 1 can avoid adjustment stagnation at the beginning of the search, thereby reducing the time required for performance evaluation using simulation or an actual machine in the automatic adjustment of control parameters for multiple operating conditions.

[0172] The parameter adjustment device 1 according to this embodiment further includes a selection probability update unit 125 that updates a selection probability table based on the evaluation value calculated by the evaluation unit 122. The selection probability table sets the probability of selecting the operating condition with the maximum evaluation value higher than the probability of selecting other operating conditions, and the probability of selecting the operating condition with the minimum evaluation value according to the progress of the search. The selection probability table is updated by the selection probability update unit 125, and the condition setting unit 121 sets the at least one operating condition selected by the operating condition selection unit 126 and the control parameters calculated by the optimization unit 124 in the control system 2.

[0173] With this configuration, the parameter adjustment server 1 performs a search operation using the operating conditions selected by the operating condition selector 126, thereby reducing the number of evaluations required for each search operation. This allows the parameter adjustment server 1 to reduce the time required for performance evaluation using simulations or actual machines when automatically adjusting control parameters for multiple operating conditions.

[0174] Furthermore, in the parameter adjustment device 1 according to this embodiment, the selection probability table is set so that the probability of selecting the upper multiple operating conditions including the maximum evaluation value is higher than the probability of selecting other operating conditions, and the probability of selecting the lower multiple operating conditions including the minimum evaluation value is set according to the progress of the search.

[0175] With this configuration, the selection probability table is set with the probability of selecting multiple operating conditions with maximum and minimum evaluation values, allowing the operating condition selection unit 126 to select a variety of operating conditions. This allows the parameter adjustment device 1 to perform each search operation using a variety of operating conditions, allowing the overall evaluation unit 123 to obtain a variety of overall evaluation values. Therefore, the parameter adjustment device 1 can avoid adjustment stagnation at the beginning of the search, thereby reducing the time required for performance evaluation using simulation or an actual machine in the automatic adjustment of control parameters for multiple operating conditions.

[0176] Furthermore, in the parameter adjustment device 1 according to this embodiment, the selection probability table is set so that when the degree of search progress is greater than the threshold, the probability of selecting the operating condition with the smallest evaluation value is lower than when the degree of search progress is less than the threshold.

[0177] With this configuration, in the early stages of a search, the operating condition selection unit 126 is more likely to select the operating condition with the smallest value, and the parameter adjustment device 1 is more likely to perform a search operation that includes that operating condition. This allows the optimization unit 124 to calculate the next control parameter to be tried using the comprehensive evaluation value of various values ​​as a hint in the early stages of a search, allowing the parameter adjustment device 1 to avoid adjustment stagnation. Furthermore, in the middle and subsequent stages of a search, the operating condition selection unit 126 is less likely to select the operating condition with the smallest value, and the parameter adjustment device 1 is less likely to perform a search operation that includes that operating condition. This allows the parameter adjustment device 1 to reduce the number of evaluations required for each search operation. Therefore, the parameter adjustment device 1 can reduce the time required for performance evaluation using a simulation or an actual machine when automatically adjusting control parameters for multiple operating conditions.

[0178] Furthermore, in the parameter adjustment device 1 according to this embodiment, the selection probability update unit 125 further updates a cost-graded selection probability table in which the probabilities set in the selection probability table are weighted by the evaluation cost, and the operating condition selection unit 126 further selects at least one operating condition based on the cost-graded selection probability table updated by the selection probability update unit 125.

[0179] This configuration makes it easier for the operating condition selector 126 to select operating conditions with low evaluation costs and less likely to select operating conditions with high evaluation costs, allowing the parameter adjustment device 1 to reduce the evaluation costs associated with each search operation. Furthermore, since the parameter adjustment device 1 performs a search operation using the operating conditions selected by the operating condition selector 126, the number of evaluations required for each search operation can be reduced. This allows the parameter adjustment device 1 to reduce the time required for performance evaluation using simulations or actual machines in the automatic adjustment of control parameters for multiple operating conditions.

[0180] Furthermore, in the parameter adjustment device 1 according to this embodiment, the selection probability update unit 125 corrects the evaluation cost according to the progress of the search.

[0181] This configuration allows the parameter adjustment device 1 to use a variety of operating conditions in the search, thereby improving the accuracy of the control parameters.

[0182] Furthermore, the parameter adjustment device 1 according to this embodiment further includes an input / output unit 11 that receives information about setting conditions for adjusting the control parameters from the input / output device 3 and outputs information about the results of the adjustment of the control parameters to the input / output device 3, and at least one of the condition setting unit 121, the evaluation unit 122, the overall evaluation unit 123, and the optimization unit 124 operates using the information about the setting conditions received by the input / output unit 11, and the results of the adjustment of the control parameters include at least one operating condition set in the control system 2 by the condition setting unit 121 and the control parameters calculated by the optimization unit 124.

[0183] With this configuration, the parameter adjustment device 1 can output the optimal control parameters obtained by the search to the input / output device 3.

[0184] Furthermore, in the parameter adjustment device 1 according to the embodiment, the selection probability update unit 125 determines whether the evaluation value calculated by the evaluation unit 122 is successful or not based on whether it is below a reference value, and updates the progress of the search based on whether the evaluation value is successful or not.

[0185] With this configuration, the parameter adjustment device 1 sets the selection probability table so that the probability of selecting an operating condition that does not fall below the reference value among multiple operating conditions increases depending on the progress of the search, thereby increasing the number of times multiple operating conditions that become bottlenecks and are likely to have a maximum value are simultaneously selected. In other words, the parameter adjustment device 1 can increase the number of times it searches for control parameters that simultaneously improve operating conditions that do not fall below the reference value. This allows the parameter adjustment device 1 to search for control parameters that will make the maximum of the obtained evaluation values ​​fall below the reference value, with less evaluation cost.

[0186] Furthermore, the parameter adjustment method according to this embodiment is a parameter adjustment method executed by a computer that searches for optimal control parameters for operating control system 2 under a plurality of operating conditions by operating control system 2 while adjusting the control parameters, and includes a condition setting step of setting at least one operating condition of a plurality of operating conditions and control parameters in control system 2; an evaluation step of calculating an evaluation value related to the operation of control system 2 when control system 2 operates under at least one operating condition and control parameter set in the condition setting step; a comprehensive evaluation step of calculating a weighted average value obtained using at least the maximum and minimum values ​​of the evaluation values ​​calculated in the evaluation step as an overall evaluation value; and an optimization step of calculating control parameters for next operating control system 2 using an optimization algorithm based on the overall evaluation value calculated in the comprehensive evaluation step, wherein at least one operating condition and the control parameters calculated in the optimization step are set in control system 2 in the condition setting step.

[0187] With this configuration, in the overall evaluation step, a weighted average of the maximum and minimum values ​​of the multiple evaluation values ​​is calculated as the overall evaluation value, so that a variety of overall evaluation values ​​can be obtained in each search operation. As a result, in the optimization step, the overall evaluation values ​​obtained in each search operation can be used as a hint to calculate the next control parameters to be tried. Therefore, the parameter adjustment method can avoid adjustment stagnation at the beginning of the search, and can reduce the time required for performance evaluation using simulation or an actual machine in automatic adjustment of control parameters for multiple operating conditions.

[0188] The program according to the present embodiment causes a computer to execute a parameter adjustment method.

[0189] According to such a program, the computer can reduce the time required for performance evaluation using simulation or an actual device in automatically adjusting control parameters for a plurality of operating conditions.

[0190] (Other Embodiments) While the parameter adjustment device and the like according to one or more aspects have been described above based on the above-described embodiments, the present disclosure is not limited to the above-described embodiments. As long as they do not deviate from the spirit of the present disclosure, various modifications conceivable by those skilled in the art to the above-described embodiments and configurations constructed by combining components of different embodiments may also be included within the scope of one or more aspects.

[0191] In the above-described embodiments, the parameter adjustment device or the like may be used for automatically adjusting control parameters in a simulation, or may be used for automatically adjusting control parameters in an actual device.

[0192] In the above embodiments, each component may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU or processor reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory.

[0193] Furthermore, some or all of the functions of the parameter adjustment device according to the above embodiments may be implemented by a processor such as a CPU executing a program.

[0194] Some or all of the components constituting each of the above devices may be configured as an IC card or a standalone module that can be attached to or detached from each device. The IC card or module is a computer system composed of a microprocessor, ROM, RAM, etc. The IC card or module may include a super multi-function LSI. The IC card or module achieves its functions when the microprocessor operates in accordance with a computer program. The IC card or module may be tamper-resistant.

[0195] A parameter adjustment device according to the present disclosure is useful, for example, as a device for searching for optimal control parameters.

[0196] REFERENCE SIGNS LIST 1 Parameter adjustment device 11 Input / output unit 12 Control unit 121 Condition setting unit 122 Evaluation unit 123 Overall evaluation unit 124 Optimization unit 125 Selection probability update unit 126 Operation condition selection unit 2 Control system 21 Control unit 22 Control target 3 Input / output device 401 File selection button 402 Operation condition setting area 403 Evaluation index setting area 404 Cost setting area 405 Adjustment execution button 406 Adjustment status display area 407 Adjustment result display area 408 Overall evaluation value display area 409 Maximum evaluation value display area 410 Selection probability table display area 411 Selection probability table with cost gradient display area 412 Best parameter display area 413 Pass criteria input area 414 Pass rate display area

Claims

1. A parameter adjusting device that searches for optimal control parameters for operating a control system under a plurality of operating conditions by operating the control system while adjusting control parameters, comprising: a condition setting unit that sets at least one operating condition of the plurality of operating conditions and the control parameters in the control system; an evaluation unit that calculates, for each of the at least one operating condition, an evaluation value related to the operation of the control system when the control system operates under the at least one operating condition and the control parameters set by the condition setting unit; a comprehensive evaluation unit that calculates, as an overall evaluation value, a weighted average obtained using at least the maximum and minimum values of the evaluation values calculated by the evaluation unit; and an optimization unit that calculates, using an optimization algorithm, control parameters for the next operation of the control system based on the overall evaluation value calculated by the comprehensive evaluation unit, wherein the condition setting unit sets the at least one operating condition and the control parameters calculated by the optimization unit in the control system.

2. The parameter adjustment device according to claim 1, wherein the comprehensive evaluation unit calculates the comprehensive evaluation value as a weighted average obtained using a plurality of top evaluation values including the maximum value among the evaluation values and a plurality of bottom evaluation values including the minimum value among the evaluation values.

3. The parameter adjustment device according to claim 1, further comprising: a selection probability update unit that updates a selection probability table based on the evaluation value calculated by the evaluation unit, in which the probability of selecting the operating condition with the maximum evaluation value is set higher than the probability of selecting other operating conditions, and the probability of selecting the operating condition with the minimum evaluation value is set according to the progress of the search; and an operating condition selection unit that selects at least one operating condition from the plurality of operating conditions based on the selection probability table updated by the selection probability update unit, wherein the condition setting unit sets the at least one operating condition selected by the operating condition selection unit and the control parameters calculated by the optimization unit in the control system.

4. The parameter adjustment device of claim 3, wherein the selection probability table is set so that the probability of selecting the top operating conditions including the maximum value among the evaluation values is higher than the probability of selecting other operating conditions, and the probability of selecting the bottom operating conditions including the minimum value among the evaluation values is set according to the progress of the search.

5. The parameter adjustment device according to claim 3, wherein the selection probability table is set so that when the degree of progress of the search is greater than a threshold, the probability of selecting the operating condition with the smallest evaluation value is lower than when the degree of progress of the search is less than the threshold.

6. The parameter adjustment device according to any one of claims 3 to 5, wherein the selection probability update unit further updates a cost-graded selection probability table in which the probabilities set in the selection probability table are weighted by evaluation costs, and the operating condition selection unit further selects at least one operating condition based on the cost-graded selection probability table updated by the selection probability update unit.

7. The parameter adjustment device according to claim 6, wherein the selection probability update unit corrects the evaluation cost according to the progress of the search.

8. The parameter adjustment device according to claim 7, further comprising an input / output unit that receives information about setting conditions for adjusting the control parameters from an input / output device and outputs information about results of the adjustment of the control parameters to the input / output device, wherein at least one of the condition setting unit, the evaluation unit, the comprehensive evaluation unit, and the optimization unit operates using the information about the setting conditions received by the input / output unit, and the results of the adjustment of the control parameters include the at least one operating condition set in the control system by the condition setting unit and the control parameters calculated by the optimization unit.

9. The parameter adjustment device according to claim 6, wherein the selection probability update unit determines success or failure based on whether the evaluation value calculated by the evaluation unit is below a reference value, and updates the progress of the search based on the success or failure of the evaluation value.

10. The parameter adjustment device according to claim 9, further comprising an input / output unit that receives information about setting conditions for adjusting the control parameters from an input / output device and outputs information about results of the adjustment of the control parameters to the input / output device, wherein at least one of the condition setting unit, the evaluation unit, the comprehensive evaluation unit, and the optimization unit operates using the information about the setting conditions received by the input / output unit, and the results of the adjustment of the control parameters include the at least one operating condition set in the control system by the condition setting unit and the control parameters calculated by the optimization unit.

11. The parameter adjustment device according to claim 7, wherein the selection probability update unit determines success or failure based on whether the evaluation value calculated by the evaluation unit is below a reference value, and updates the progress of the search based on the success or failure of the evaluation value.

12. The parameter adjustment device according to claim 11, further comprising an input / output unit that receives information about setting conditions for adjusting the control parameters from an input / output device and outputs information about results of the adjustment of the control parameters to the input / output device, wherein at least one of the condition setting unit, the evaluation unit, the comprehensive evaluation unit, and the optimization unit operates using the information about the setting conditions received by the input / output unit, and the results of the adjustment of the control parameters include the at least one operating condition set in the control system by the condition setting unit and the control parameters calculated by the optimization unit.

13. A parameter adjustment method executed by a computer for searching for optimal control parameters for operating a control system under a plurality of operating conditions by operating the control system while adjusting control parameters, the parameter adjustment method comprising: a condition setting step for setting at least one operating condition of the plurality of operating conditions and the control parameters in the control system; an evaluation step for calculating an evaluation value related to the operation of the control system when the control system operates under the at least one operating condition and the control parameters set in the condition setting step; a comprehensive evaluation step for calculating, as a comprehensive evaluation value, a weighted average obtained using at least the maximum and minimum values of the evaluation values calculated in the evaluation step; and an optimization step for calculating, using an optimization algorithm, control parameters for the next operation of the control system based on the comprehensive evaluation value calculated in the comprehensive evaluation step, wherein in the condition setting step, the at least one operating condition and the control parameters calculated in the optimization step are set in the control system.

14. A program for causing the computer to execute the parameter adjustment method according to claim 13.

Images