Manufacture or control a technical system using an optimized parameter set

Abstract

The invention relates to a computer-implemented method for manufacturing or controlling a technical system, comprising the method steps: (a) inputting (S1) a sample of parameter sets suitable for manufacturing or controlling the technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible or technically infeasible or erroneous in terms of manufacturing or controlling the technical system, (b) generating (S2) a computerized surrogate model of the technical system by means of a regression method based on the corresponding parameter sets of the sample, said parameter sets being marked as technically feasible, (c) determining (S3) an optimized parameter set based on the surrogate model by means of a computerized optimization method, and (d) outputting (S4) an optimized parameter set for manufacturing or controlling the technical system.

Description

Technical Field

[0001] The invention relates to a computer-implemented method, an apparatus and a computer program product for generating an optimized parameter set for manufacturing or controlling a technical system. Background Art

[0002] Before manufacturing or controlling a product or technical system, it is necessary to determine suitable production, control and / or construction parameters. During, for example, the design phase, computerized optimization methods are often used to optimize parameter settings for manufacturing products or controlling technical systems. Such an optimization process is often run in an iterative manner, and may therefore be time and / or resource consuming, because the parameter values ​​obtained from the optimization need to be evaluated. Evaluation is often performed using a computerized simulation of the product or technical system, such as a finite element analysis (FEA) or computational fluid dynamics (CFD) simulation. However, typical optimizers often do not have enough information about such computerized models, so that the optimizer can generate parameters that cannot be evaluated and / or do not satisfy system or product constraints, resulting in useless parameter sets.

[0003] EP 3 246 769 A1 describes a controller for simulation and optimization of the operation of a power plant, wherein a dynamic model of the power plant based on geometric and operational data is created, and surrogate models for specific performance indicators based on the dynamic model are generated, which are incorporated into the optimization process to optimize the operation of the power plant for the specific performance indicators. Summary of the invention

[0004] It is therefore an object of the present invention to improve the generation of optimized parameter sets for manufacturing or controlling a product or a technical system.

[0005] The objects of the invention are achieved by the features of the independent claim. The dependent claims contain further developments of the invention.

[0006] According to a first aspect, the invention provides a computer-implemented method for producing or controlling a technical system, comprising the method steps of:

[0007] (a) input a sample of parameter sets suitable for manufacturing or controlling a technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible, with respect to manufacturing or controlling the technical system, if it meets a given system criterion for the technical system, or technically infeasible, if it does not meet the given system criterion, or erroneous, if an evaluation of the parameter set using a computerized simulation results in an error,

[0008] (b) generating a computerized alternative model of the technical system by means of a regression method based on corresponding parameter sets of the sample, said parameter sets being marked as technically feasible, but not erroneous and not infeasible parameter sets,

[0009] (c) determining an optimized parameter set based on the surrogate model by means of a computerized optimization method, and

[0010] (d) Output an optimized set of parameters for manufacturing or controlling a technical system.

[0011] If not indicated differently, the terms "operate", "execute", "computer-implemented", "calculate", "determine", "generate", "configure", "reconstruct", etc. preferably relate to actions and / or processes and / or steps of changing and / or generating data, wherein the data may be particularly presented as physical data and may be executed by a computer or processor. The term "computer" may be interpreted broadly and may be a personal computer, a server, a mobile computing device or a processor, such as a central processing unit (CPU) or a microprocessor.

[0012] An important advantage of the invention is the use of a surrogate model which represents and / or approximates at least part of the technical system. The surrogate model is preferably a computerized model. Preferably, the surrogate model is less complex than a detailed computerized simulation model of the technical system, such as a CFD simulation model. The surrogate model can be stored, for example, in a data structure. It is generated with the aid of a regression method and is generated from an input sample of a parameter set for manufacturing or controlling the technical system. The surrogate model is generated based on at least one technically feasible parameter set and a corresponding evaluation result of the parameter set.

[0013] A parameter set comprises parameter values ​​which at least define properties or functions of a technical system. A parameter set may be understood as a data set. A parameter set may, for example, comprise construction, production, control, operating and / or design parameters. A construction parameter set comprises parameters of at least a component of a technical system which may, for example, specify dimensions, weight, material, etc. Furthermore, a construction or control parameter set may comprise system- or product-specific setting parameters for a manufacturing unit or machine. A control parameter set may, for example, comprise system-specific control parameters for setting a control unit of a technical system for controlling the technical system.

[0014] The parameter set may be used, for example, as an input to a manufacturing unit, wherein the manufacturing unit is configured to manufacture a technical system according to the corresponding construction parameter set. Thus, the sample of parameter sets is a plurality of parameter sets, wherein each parameter set is suitable for manufacturing a technical system. Each parameter set is marked with a feasibility identifier, thereby allowing identification of a more suitable parameter set.

[0015] The optimized parameter set is determined based on the surrogate model, thereby accelerating the optimization process. This is because the surrogate model is preferably generated based on a feasible and not erroneous and preferably not infeasible parameter set. The optimization is then run on the pre-constrained model. In other words, the optimization process for determining the optimized parameter set takes into account information about, for example, the system constraints of the technical system. As a result, a more suitable optimized parameter set for manufacturing or controlling the technical system can be generated. As a result, the parameter optimization is accelerated compared to standard methods.

[0016] According to an embodiment of the computer-implemented method, a feasibility identifier of the optimization parameter set may be determined by means of a computerized evaluation method.

[0017] The computerized evaluation method can be trained, for example, based on a machine learning method, so that it reproduces the feasibility identifier distribution of a training data set, wherein the training data set includes parameter sets labeled with corresponding feasibility identifiers. The computerized evaluation method is, for example, an artificial neural network, etc. Using such a computerized evaluation method, information embedded in the data, i.e., in the parameter set, can be used to identify whether the parameter set is feasible, infeasible or erroneous.

[0018] For example, an output optimized parameter set can be evaluated by means of such a trained computerized evaluation method to determine whether it is technically feasible or technically infeasible or erroneous in terms of manufacturing or controlling a technical system.

[0019] According to further embodiments of the computer-implemented method, the set of optimization parameters may be determined by means of a computerized optimization method and according to its feasibility identifier.

[0020] The values ​​of the corresponding feasibility indicators of the outputted optimized parameter sets, i.e. technically feasible or technically infeasible or erroneous, can be further evaluated by the optimization method as additional optimization constraints. With such additional information, the optimization method allows the generation of parameter sets that avoid particularly erroneous parameter combinations and / or unnecessary calculation iterations.

[0021] According to further embodiments of the computer-implemented method, the optimized parameter set may be added to the sample of parameter sets and method steps (b) to (d) of claim 1 may be repeated.

[0022] By adding the outputted optimized parameter set to the sample of parameter sets, an optimized surrogate model can be generated, thereby further improving the generation of optimized parameter sets.

[0023] According to further embodiments of the computer-implemented method, an input sample of parameter sets may be generated and output by means of a second computerized optimization method, which is configured to determine at least one parameter set suitable for manufacturing or controlling a technical system.

[0024] The input samples of parameter sets are preferably generated by means of a second computerized optimization method and evaluated using a computerized simulation model of the technical system in order to determine the technical feasibility of each parameter set. The second computerized optimization method may for example be identical to the aforementioned optimization method.

[0025] According to further embodiments of the computer-implemented method, the regression method may be an incremental local Gaussian regression method.

[0026] To generate surrogate models, the Incremental Local Gaussian Regression algorithm (ILGR) can be applied to the trained model. This algorithm combines the advantages of Local Weighted Regression (LWR) and Gaussian Process Regression (GPR). It uses a localized function basis and approximate inference techniques to build a Gaussian Process Regression algorithm with increasing locality and similar computational complexity to LWR. It allows, for example, streamlining of datasets of unknown size, making it particularly suitable for generating surrogate models during iterative optimization processes. Model generation can be done incrementally when new parameter sets are available.

[0027] According to further embodiments of the computer-implemented method, the output optimized parameter set may be used to produce or control a technical system.

[0028] The outputted optimized parameter set can for example be transmitted to a manufacturing unit or a machine for manufacturing the technical system according to the construction parameter set. Alternatively, the outputted optimized parameter set can be transmitted to a control unit for controlling the technical system.

[0029] According to a second aspect, the invention provides an apparatus for producing or controlling a technical system, comprising:

[0030] (a) an input unit configured to input a sample of parameter sets suitable for manufacturing or controlling a technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible in terms of manufacturing or controlling the technical system if it meets a given system criterion for the technical system, or technically infeasible if it does not meet the given system criterion, or erroneous if an evaluation of the parameter set using a computerized simulation results in an error,

[0031] (b) a generator configured to generate, by means of a regression method, a computerized alternative model of the technical system based on a parameter set, said parameter set being marked as a technically feasible, but not erroneous and not infeasible parameter set,

[0032] (c) an optimizer configured to determine an optimization parameter set based on the generated surrogate model by means of a computerized optimization method, and

[0033] (d) An output unit configured to output an optimized parameter set for manufacturing or controlling a technical system.

[0034] The device and / or at least one of its units may also include at least one processor or computer to perform method steps according to the present invention. The corresponding unit may be implemented in hardware and / or software. If the unit is implemented in hardware, it may be embodied as a device, such as a part of a computer or processor or system. If the unit is implemented in software, it may be embodied as a computer program product, a function, a routine, a program code or an executable object. The output unit preferably provides a data structure including an optimization parameter set. This data structure may, for example, be transmitted to a manufacturing unit or a control unit.

[0035] The apparatus may further comprise a manufacturing unit configured to manufacture the technical system using the outputted optimized parameter set. Alternatively, the apparatus may be connected to the manufacturing unit, for example via a network.

[0036] The device may further comprise a control unit configured to control the technical system using the outputted optimized parameter set.Alternatively, the device may be connected to the control unit, for example via a network.

[0037] The invention also comprises a computer program product directly loadable into the internal memory of a digital computer, comprising software code portions for performing the steps of the method when said product is run on a computer.

[0038] The computer program product such as computer program means may be embodied as a memory card, a USB stick, a CD-ROM, a DVD or as a file which can be downloaded from a server in a network. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The present invention will be explained in more detail with reference to the accompanying drawings.

[0040] Figure 1 A flow chart is shown which includes method steps involved in an embodiment of a method for producing or controlling a technical system;

[0041] Figure 2 A schematic representation of an embodiment of a method for producing or controlling a technical system is shown; and

[0042] Figure 3 A schematic representation of an exemplary embodiment of an apparatus for producing or controlling a technical system is shown.

[0043] Identical parts in different figures are marked with the same reference numerals. DETAILED DESCRIPTION

[0044] Figure 1A flow chart illustrating method steps of a computer-implemented method according to the invention for producing or controlling a technical system using an optimized parameter set is shown. The technical system may be, for example, a machine, such as a turbine or a motor, a hardware component of a machine, a mechanical part, a product, etc. Furthermore, the technical system may be a complex system comprising a plurality of individual parts, such as a transport network for water, oil, packaging, etc.

[0045] In a first step S1, a sample of a parameter set suitable for manufacturing or controlling a technical system is input. Depending on the respective application, ie, for example, manufacturing or controlling a technical system, the parameter set comprises construction or production parameters or control parameters.

[0046] Each parameter set is marked with a feasibility identifier, which marks it as technically feasible or technically infeasible or erroneous with respect to the production or control of the technical system. The respective feasibility identifier can be determined, for example, by means of a classification or evaluation method and / or by means of an analytical method that evaluates the respective parameter set, for example using a computerized simulation of the technical system.

[0047] For example, if a parameter set meets a given system criterion for a technical system, the parameter set is marked as technically feasible. For example, if a parameter set does not meet a given system criterion, the parameter set is marked as technically infeasible. If, for example, the evaluation of the parameter set using a computerized simulation leads to an error, the parameter set is marked as erroneous, for example. Therefore, whether a parameter set is considered technically feasible is determined by the formed optimization problem. If a parameter set violates one of the constraints, it is infeasible. If a parameter set meets all constraints, it is marked as feasible. The distinction between feasible or infeasible is an error case. While the former has been determined by optimization, the latter is the result of an erroneous evaluation, for example an error from an evaluation software tool.

[0048] In a next step S2, a computerized surrogate model of the technical system is generated by means of a regression method. Preferably, an incremental local Gaussian regression method is used, which is suitable for training surrogate models. The corresponding parameter sets preferably define functions or properties of the technical system, such as size, mass, width, height, etc. The computerized surrogate model is configured to represent and / or approximate the technical system. Thus, the surrogate model can be generated based on technically feasible and / or technically infeasible input parameter sets. Preferably, only technically feasible parameter sets are used for the surrogate model generation. In other words, for the surrogate model generation, only a part of the sample of parameters is used, preferably only the technically feasible ones.

[0049] In the next step S3, a computerized optimization method is used to determine an optimization parameter set based on the generated computerized surrogate model. A black box optimizer may be used. Since only the evaluation of the objectives and constraints for the variable values, i.e. the corresponding parameter set, is available, a black box optimizer may be used to propose variable values ​​and evaluate them.

[0050] The optimized parameter set is output, step S4, for manufacturing or controlling the technical system. The optimized configuration parameter set can be used, for example, as input to a manufacturing machine that manufactures the technical system, or as input to a production unit for producing a product.

[0051] Alternatively, the feasibility identifier of the optimization parameter set is determined by means of a computerized evaluation method, step S6. The computerized evaluation method has been trained to determine whether an input parameter set is technically feasible, technically infeasible or erroneous in terms of manufacturing or controlling a technical system. Therefore, the computerized evaluation method is preferably trained or configured to represent the distribution of feasibility identifiers of an initial input sample of parameter sets. The value of the feasibility identifier of the optimization parameter set can also be used during the optimization, i.e. step S3, as a further optimization constraint. Therefore, when determining the optimization parameter set, the optimization method can be adapted to additionally evaluate the corresponding feasibility identifier. Preferably, the optimization parameter set evaluated as erroneous is selected.

[0052] Furthermore, the optimized parameter set which is preferably evaluated as a technically feasible output can be added to the initial input sample of parameter sets and method steps S1 to S4 can be repeated, see arrow REP. Such iteration provides an improved optimized parameter set which can be output for manufacturing or controlling a technical system.

[0053] Figure 2 A schematic representation of an embodiment of the invention is shown.A computer-implemented method for manufacturing or controlling a technical system using an optimized parameter set may comprise two method parts 10,20.

[0054] An initial input data set, i.e. a sample of parameter sets PS1, can be provided, for example, by an optimization method OPT2. Such an optimization method OPT2 can, for example, generate a sample of parameter sets PS1 for a manufacturing technical system. A feasibility identifier is assigned to each of the samples of PS1 parameter sets, which marks each parameter set as technically feasible, technically infeasible or erroneous. The corresponding feasibility identifier can be determined, for example, using a simulation-based evaluation. In other words, the parameter set PS1 can be understood as a labeled data set.

[0055] This sample of parameter sets PS1 can be used, for example, as a training data set in order to train a computerized evaluation method NN, for example an artificial neural network. The computerized evaluation method NN is trained such that it reproduces the feasibility identifier distribution of the training data set. The method part 10 of the method shown represents such a data classification.

[0056] The labeled data set can for example be divided into a training data set and a test data set, so that the computerized evaluation method NN can be trained with the training data set and validated with the test data set. Preferably, all parameter sets are used, i.e. technically feasible, technically infeasible or erroneous parameter sets. The trained evaluation method NN can then be used to determine a feasibility identifier FI for the output parameter set.

[0057] Furthermore, the results of the computerized evaluation method NN can be used as additional constraints for the optimization method, step S3, to determine the optimized parameter set. Such constraints can provide feedback to the optimization method and guide the optimization problem to generate an optimized parameter set according to the feasibility identifier FI, i.e., a parameter set that is preferably free of errors.

[0058] The method part 20 comprises the generation of surrogate models SM for objectives and constraints and parameter optimization based on the surrogate models. Preferably, using only a part of the initial sample of parameter sets PS2, which only include parameter groups classified as technically feasible, a surrogate model SM of the technical system is generated by means of a regression method, step S2.

[0059] The generation of the surrogate model SM preferably includes the following steps. The first main step is the training process. Here the incremental local Gaussian regression algorithm (ILGR) is suitable for training the surrogate model SM, which combines the advantages of local weighted regression (LWR) and Gaussian process regression (GPR). The second step is to cross-validate the generated surrogate model. The generated surrogate model SM is verified by a validation set (e.g., a part of a partial sample of the parameter set PS2). The root mean square error (RMSE) and R square value are evaluated with the validation set, and the number of sub-models (also referred to as local models) of ILGR is the input of the third step: hyperparameter optimization. These three items constitute the objective function, and the optimization problem is solved by a black box optimizer, such as BOBYQA from the NLopt library. Then, the surrogate model generation process is restarted with the updated hyperparameters until the stopping criteria are met. Therefore, the hyperparameters of ILGR are determined by minimizing the RMSE and the number of sub-models and maximizing R^2.

[0060] Part 10 and part 20 can be two independent methods to speed up the optimization. They can be applied together or separately.

[0061] Based on the surrogate model SM, an optimization parameter set PS_opt can be determined with the aid of an optimization method, step S3. With the aid of the optimization method, the parameters of the surrogate model SM are modified in order to find optimal values. In the case of manufacturing, the optimization method can, for example, determine an optimized parameter set for a product, resulting in an optimized size, weight or other physical property of the product. In other words, the optimization method can determine the optimized parameters of the surrogate model with respect to a given optimization goal.

[0062] If method part 10 is applied, the parameter set can be evaluated based on the computerized evaluation method NN, resulting in a feasibility identifier FI. If only part 20 is applied, the optimization method in step S3 will evaluate alternative models. If both parts 10, 20 are applied, the optimization method in step S3 includes an additional constraint depending on the feasibility identifier FI and evaluates the resulting alternative model SM. Taking into account this additional constraint, the optimization parameter set PS_opt is determined.

[0063] Figure 3 A schematic representation of an embodiment of an apparatus for producing or controlling a technical system is shown. The apparatus 100 for producing or controlling a technical system comprises an input unit 101, a generator 102, an optimizer 103 and an output unit 104. The apparatus and / or at least one of its units may also comprise at least one processor or computer to perform the method steps according to the invention.

[0064] The input unit 101 is configured to input a sample of parameter sets suitable for manufacturing or controlling a technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible or technically infeasible or erroneous in terms of manufacturing or controlling the technical system.

[0065] The generator 102 is configured to generate a computerized surrogate model of a technical system by means of a regression method based on a parameter set, preferably said parameter set being marked as technically feasible.

[0066] The optimizer 103 is configured to determine an optimization parameter set based on the generated surrogate model by means of a computerized optimization method.

[0067] The output unit 104 is configured to output an optimized parameter set for manufacturing or controlling a technical system.

[0068] The apparatus preferably comprises an evaluation unit 105 configured to determine a feasibility identifier for the input parameter set by means of a computerized evaluation method, such as an artificial neural network or a random forest algorithm.

[0069] The device further preferably further comprises a manufacturing unit (not shown) or is coupled to the manufacturing unit, wherein the manufacturing unit is configured to manufacture the technical system using the output optimized parameter set. Additionally or alternatively, the device may comprise a control unit (not shown) or is coupled to the control unit, wherein the control unit is configured to control the technical system using the output optimized parameter set.

[0070] While the present invention has been described in detail with reference to advantageous embodiments, it should be understood that the invention is not limited to the disclosed examples and that numerous additional modifications and variations may be made thereto by those skilled in the art without departing from the scope of the present invention.

Claims

1. A computer-implemented method for producing or controlling a technical system, comprising the method steps: (a) input (S1) a sample of parameter sets suitable for manufacturing or controlling a technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible with respect to manufacturing or controlling the technical system if it meets a given system criterion for the technical system, or technically infeasible if it does not meet the given system criterion, or erroneous if an evaluation of the parameter set using a computerized simulation resulted in an error, (b) generating (S2) a computerized alternative model of the technical system by means of a regression method based on corresponding parameter sets of the sample, said parameter sets being marked as technically feasible, but not erroneous and not infeasible parameter sets, (c) determining (S3) an optimization parameter set based on the surrogate model by means of a computerized optimization method, and (d) Output (S4) Optimized parameter set (PS_opt) for manufacturing or controlling the technical system. 2 . The computer-implemented method according to claim 1 , wherein the feasibility identifier (FI) of the optimization parameter set (PS_opt) is determined ( S6 ) by means of a computerized evaluation method (NN). 3 . The computer-implemented method according to claim 1 , wherein the optimization parameter set (PS_opt) is determined by means of a computerized optimization method (NN) and according to its feasibility identifier (FI).

4. The computer-implemented method of any of the preceding claims, wherein the optimized parameter set is added to the sample of parameter sets and method steps (b) to (d) of claim 1 are repeated (REP).

5. Computer-implemented method according to one of the preceding claims, wherein the input samples of parameter sets are generated and output by means of a second computerized optimization method (OPT2), which is configured to determine at least one parameter set suitable for manufacturing or controlling a technical system. The computer-implemented method of claim 1 , wherein the regression method is an incremental local Gaussian regression method. 7 . The computer-implemented method according to claim 1 , wherein the technical system is manufactured or controlled using the outputted optimized parameter set ( S5 ).

8. Apparatus (100) for producing or controlling a technical system, comprising: (a) an input unit (101) configured to input a sample of parameter sets suitable for manufacturing or controlling a technical system, wherein a feasibility identifier is assigned to each parameter set, wherein the feasibility identifier marks each parameter set as technically feasible in terms of manufacturing or controlling the technical system if it meets a given system criterion for the technical system, or technically infeasible if it does not meet the given system criterion, or erroneous if an evaluation of the parameter set using a computerized simulation results in an error, (b) a generator (102) configured to generate, by means of a regression method, a computerized alternative model of the technical system based on a parameter set, said parameter set being marked as a technically feasible, but not erroneous and not infeasible parameter set, (c) an optimizer (103) configured to determine an optimization parameter set based on the generated surrogate model by means of a computerized optimization method, and (d) An output unit (104) configured to output an optimized parameter set (PS_opt) for manufacturing or controlling a technical system.

9. The apparatus according to claim 8, comprising a manufacturing unit configured to use the outputted optimized parameter set to manufacture the technical system.

10. The device according to claim 8, comprising a control unit configured to control a technical system using the outputted optimized parameter set.

11. Computer program product directly loadable into the internal memory of a digital computer, comprising software code portions for performing the steps according to one of claims 1 to 7 when said computer program product is run on a computer.

Images