Intelligent optimization method and system for equipment model

Abstract

The invention discloses an intelligent optimization method and system for an equipment model. The intelligent optimization method comprises the steps of obtaining multiple pieces of model test data of equipment; carrying out sparse sampling on the multiple pieces of test data by adopting a central composite design algorithm or a Benkine box design algorithm to obtain a data sample; inputting the data sample into a pre-constructed intelligent optimization model to obtain a prediction equipment model; wherein the intelligent optimization model is obtained by training a polynomial response surface model by adopting historical sample data. According to the system and the method, the artificial intelligence technology is utilized, historical test data and simulation data are combined, forward and reverse fusion is carried out, an intelligent optimization model is constructed, a designer does not need to carry out simulation calculation, optimal model parameters are rapidly obtained through the intelligent optimization model, and therefore manpower and time are saved.

Description

Technical Field

[0001] The present invention relates to the technical field of model construction, and specifically to an intelligent optimization method and system for an equipment model. Background Art

[0002] Traditional equipment research and development design relies on CAD modeling, simulation calculation, and optimization algorithms, which belongs to a forward design method. However, the traditional design method requires designing and constructing a CAD model, then performing mesh division, boundary condition setting, material property setting, simulation solution calculation, comparing whether the simulation results meet the expected goals, and cooperating with the optimization algorithm to iteratively perform CAD modeling and simulation calculation. The whole process is time-consuming and labor-intensive. Summary of the Invention

[0003] To solve the problems raised in the background art. In a first aspect, the present invention proposes an intelligent optimization method for an equipment model, including:

[0004] Obtaining multiple model test data of the equipment;

[0005] Performing sparse sampling on the multiple test data by using a central composite design algorithm or a Box-Behnken design algorithm to obtain data samples;

[0006] Inputting the data samples into a pre-constructed intelligent optimization model to obtain a predicted equipment model;

[0007] Wherein, the intelligent optimization model is obtained by training a polynomial response surface model with historical sample data.

[0008] Preferably, the construction process of the intelligent optimization model includes:

[0009] Obtaining historical sample data of the equipment; using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical equipment model; based on the historical sample data, obtaining a historical measured equipment model; taking the minimum error between the predicted historical equipment model and the historical measured equipment model as the goal, training the polynomial response surface model, and iterating repeatedly until a preset iteration stop condition is reached, and outputting the model with the corresponding parameters reaching the iteration stop condition as the intelligent optimization model.

[0010] Preferably, using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical equipment model satisfies the following formula: a first-order formula or a polynomial;

[0011] The first-order formula satisfies the following formula:

[0012]

[0013] In the above formula: x iis the i-th dimension of the historical sample data, n is the total number of dimensions, and β i represents the univariate polynomial coefficient, and β i is the univariate polynomial coefficient, is the predicted historical device model;

[0014] The polynomial satisfies the following formula:

[0015]

[0016] In the above formula: β ii is the bivariate square term coefficient, and β ij is the bivariate cross-term coefficient, x j is the cross-term of the historical sample data, and j is the j-th dimension of the cross-term of the historical sample data.

[0017] Preferably, the data sample satisfies the following formula:

[0018]

[0019] In the above formula: d is the dimension of the sample data, and N sampling is the number of sample data.

[0020] Preferably, obtaining multiple model test data of the device includes:

[0021] Based on the multiple model test data, obtaining the data volume of the model test data;

[0022] When the data volume is less than one hundred, performing multiple simulation calculations on the device to obtain multiple groups of simulated model test data.

[0023] In a second aspect, the present application also proposes an intelligent optimization system for a device model, including:

[0024] An acquisition module for acquiring multiple model test data of the device;

[0025] A data sample acquisition module for performing sparse sampling on the multiple test data by using a central composite design algorithm or a Box-Behnken design algorithm to obtain a data sample;

[0026] A predicted device model acquisition module for inputting the data sample into a pre-constructed intelligent optimization model to obtain a predicted device model; wherein, the intelligent optimization model is trained by using historical sample data for a polynomial response surface model.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0028] This system and method utilize artificial intelligence technology, combined with historical test data and simulation data, to conduct forward and reverse integration, constructing an intelligent optimization model. Designers do not need to perform CAD modeling and simulation calculations, and can quickly obtain the optimal model parameters through the intelligent optimization model, thus saving manpower and time. Description of the Drawings

[0029] Figure 1 is the traditional equipment R & D and design process of the background technology;

[0030] Figure 2 is the flowchart of an intelligent optimization method for an equipment model of the present invention;

[0031] Figure 3 The usage and construction process of the intelligent optimization model in Embodiment 1 of the present invention. Detailed Embodiments

[0032] Next, the technical solutions of the present invention in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the present invention in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0033] Embodiment 1:

[0034] As Figure 2-3 shown, the present invention proposes an intelligent optimization method for an equipment model, including the following steps:

[0035] Step 1: Obtain multiple model test data of the equipment;

[0036] Step 2: Perform sparse sampling on the multiple test data by using the central composite design algorithm or the Box - Behnken design algorithm to obtain data samples;

[0037] Step 3: Input the data samples into a pre - constructed intelligent optimization model to obtain a predicted equipment model; wherein, the intelligent optimization model is trained by using historical sample data for a polynomial response surface model.

[0038] In the above - mentioned Step 1, the obtaining of multiple model test data of the equipment includes:

[0039] Based on the multiple model test data, obtain the data volume of the model test data;

[0040] When the data volume is less than one hundred, perform multiple simulation calculations on the equipment to obtain multiple groups of simulated model test data to make up for the situation of less test data.

[0041] In step 3 described above, the construction process of the intelligent optimization model includes: obtaining the historical sample data of the device; using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical device model; based on the historical sample data, obtaining a historical measured device model; aiming at minimizing the error between the predicted historical device model and the historical measured device model, training the polynomial response surface model, and iterating repeatedly until a preset iteration stop condition is reached, and outputting the model with the corresponding parameters reaching the iteration stop condition as the intelligent optimization model.

[0042] Build an intelligent optimization model and use the "Minimum Prediction Strategy (MP)", as Figure 2 shown. Assuming that the surrogate model is sufficiently accurate, call a robust optimization solver to find the minimum value of this surrogate model. After a large number of iterations, the global optimal solution will always be obtained. At this time, high-precision computational simulation is performed at the predicted optimal solution, and there are often deviations between the high-precision response obtained and the predicted response from the surrogate model. Subsequently, this set of high-precision results is supplemented to the original database to reconstruct the surrogate model. Repeating this process makes the deviation between the predicted value and the true value smaller and smaller, and finally the true minimum value is obtained. MP is a relatively simple and intuitive sample update strategy, which only needs to use the predicted optimal solution or a solution near it as the updated sample. However, the disadvantage is that it will fall into a local optimal region during the optimization process and cannot jump out. After establishing the surrogate models of the objective function and the constraint function, solve the following optimization problem, where n is the number of constraint functions. The intelligent optimization model can be constructed by using the Polynomial Response Surface Method (PRS). PRS has now been widely and effectively applied to a large number of engineering designs, can accurately express convex function problems, and its approximate expression is obtained by the least squares method.

[0043] In the above, using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical device model satisfies the following formula: first-order formula or polynomial;

[0044] The first-order formula satisfies the following formula:

[0045]

[0046] In the above formula: x i is the i-th dimension of the historical sample data, n is the total number of dimensions, β i represents the univariate polynomial coefficient, β i is the univariate polynomial coefficient, is the predicted historical device model;

[0047] The polynomial satisfies the following formula:

[0048]

[0049] In the above formula: β ii is the coefficient of the bivariate square term, and β ij is the coefficient of the bivariate cross term, x j is the cross term of the historical sample data, and j is the j-th dimension of the cross term of the historical sample data.

[0050] Furthermore, the data sample satisfies the following formula:

[0051]

[0052] In the above formula: d is the dimension of the sample data, and N sampling is the number of sample data.

[0053] The above polynomial response surface is easy to construct, and its characteristics of continuous and smooth function contribute to the rapid convergence of noisy optimization problems. However, due to its simple characteristics, it is difficult to accurately predict and express in nonlinear problems. It has a wide range of applications, including robust optimization, multidisciplinary optimization, adaptive strategies for global optimization, and manufacturing analysis, etc.;

[0054] The intelligent optimization model is mainly used for the rapid intelligent design of equipment. By inputting boundary conditions and constraint conditions, a three-dimensional geometric model of the equipment can be quickly generated. The intelligent optimization model is mainly trained by small sample test data and simulation data to provide data sources. By performing high-dimensional space sparse sampling on the data of the data sources, a surrogate model of the data is constructed, and by performing data-driven optimization on the surrogate model, the intelligent optimization model is realized. The following figure shows the design process using the intelligent optimization model. Designers only need to input performance constraint indicators to quickly obtain a three-dimensional geometric model.

[0055] The present invention improves the structural design efficiency of complex products through artificial intelligence means. Traditional complex product design starts from CAD modeling, uses CAE simulation means to verify product performance, and through manual parameter adjustment or by means of optimization algorithms for parameter optimization, and performs cyclic simulation calculations, and finally obtains a CAD model that meets the product performance indicators. The intelligent optimization model, based on artificial intelligence technology, uses historical test data and simulation data as data sources for model training to construct an intelligent optimization model. Designers only need to input multiple model test data to quickly generate a CAD model that meets the product performance indicators.

[0056] Compared with the traditional R & D and design process of equipment, which is time-consuming and laborious, with the rapid development of current artificial intelligence technology, when exploring the integration of artificial intelligence technology and simulation technology, it is tried to use historical test data and simulation data as big data to construct a big model, which can quickly perform simulation calculations, and then integrate data-driven optimization algorithms to perform optimization while performing simulation calculations to quickly obtain the optimal simulation result. Through the method of this patent, the R & D and design of equipment can be carried out simply and quickly.

[0057] Embodiment 2:

[0058] Based on the same concept, the present invention provides an intelligent optimization system for a device model, including:

[0059] An acquisition module, configured to acquire a plurality of model test data of the device;

[0060] A data sample acquisition module, configured to perform sparse sampling on the plurality of test data by using a central composite design algorithm or a Box-Behnken design algorithm to obtain data samples;

[0061] A predicted device model acquisition module, configured to input the data samples into a pre-constructed intelligent optimization model to obtain a predicted device model; wherein, the intelligent optimization model is trained by using historical sample data for a polynomial response surface model.

[0062] The above embodiments are only used to illustrate the present invention, rather than limiting the present invention. Those of ordinary skill in the relevant technical field can also make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the present invention. The patent protection scope of the present invention is defined by the claims.

[0063] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent optimization method for a device model, characterized in that Including: Obtaining a plurality of model test data of the device; Performing sparse sampling on the plurality of test data by using a central composite design algorithm or a Box-Behnken design algorithm to obtain data samples; Inputting the data samples into a pre-constructed intelligent optimization model to obtain a predicted device model; Wherein, the intelligent optimization model is obtained by training a polynomial response surface model with historical sample data.

2. The method according to claim 1, wherein The construction process of the intelligent optimization model includes: Obtaining historical sample data of the device; using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical device model; based on the historical sample data, obtaining a historical measured device model; taking the minimum error between the predicted historical device model and the historical measured device model as the target, training the polynomial response surface model, and iterating repeatedly until a preset iteration stop condition is reached, and outputting the model with the corresponding parameters reaching the iteration stop condition as the intelligent optimization model.

3. The method according to claim 2, wherein Using the historical sample data as the input of the polynomial response surface model to obtain a predicted historical device model, which satisfies the following formula: a first-order formula or a polynomial; The first-order formula satisfies the following formula: In the above formula: x i is the i-th dimension of the historical sample data, n is the total number of dimensions, β i represents the univariate polynomial coefficient, β i is the univariate polynomial coefficient, is the predicted historical device model; The polynomial satisfies the following formula: In the above formula: β ii is the coefficient of the bivariate square term, β ij is the coefficient of the bivariate cross term, x j is the cross term of the historical sample data, and j is the j-th dimension of the cross term of the historical sample data.

4. The method according to claim 2, wherein The data samples satisfy the following formula: In the above formula: d is the dimension of the sample data, and N sampling is the number of the sample data.

5. The method according to claim 1, wherein The obtaining of the plurality of model test data of the device includes: Based on the plurality of model test data, obtaining the data volume of the model test data; When the data volume is less than one hundred, performing multiple simulation calculations on the device to obtain multiple sets of simulated model test data.

6. An intelligent optimization system for a device model, characterized in that, Including: An obtaining module, configured to obtain a plurality of model test data of the device; A data sample obtaining module, configured to perform sparse sampling on the plurality of test data by using a central composite design algorithm or a Box-Behnken design algorithm to obtain data samples; A predicted device model obtaining module, configured to input the data samples into a pre-constructed intelligent optimization model to obtain a predicted device model; wherein, the intelligent optimization model is obtained by training a polynomial response surface model with historical sample data.

Images