Optical system assembly precision prediction method and device based on thermal coupling simulation

Abstract

The invention provides an optical system assembly precision prediction method and device based on thermal-mechanical coupling simulation, and relates to the technical field of data processing, and the method comprises the steps: obtaining an assembly parameter and a variable temperature parameter of a fixed part of an optical system; the assembly parameters and the variable temperature parameters are analyzed based on a thermal coupling simulation system, first surface shape data are obtained, and the thermal coupling simulation system is used for analyzing the morphology change condition of the surface of the fixed part; performing feature extraction on the first surface shape data to obtain a first surface shape feature; the first surface shape feature is predicted based on a preset prediction model, a first energy concentration ratio is obtained, the first energy concentration ratio is used for representing the assembly precision of the optical system, and the preset prediction model is a model trained for predicting the energy concentration ratio. The efficiency of evaluating the assembly precision of the optical system can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for predicting the assembly accuracy of an optical system based on thermomechanical coupling simulation. Background Art

[0002] The optical system is a key component of observation instruments and equipment, and is widely used in aerospace, weaponry, ground and air detection and other fields. The optical system uses nuts to connect different structures. Since the temperature changes during use, the expansion or contraction degrees between different connecting parts are different, resulting in deformation of the connection position, which in turn affects the imaging of the optical system. In the existing technology, the deformation of the optical system under different temperature conditions is tested by high-precision laser interferometers or digital image correlation technology (Digital Image Correlation, DIC), and then it is determined whether the deformation is within a reasonable range. However, the existing technology requires long-term measurement of each optical system, which makes it difficult to quickly determine the assembly accuracy of the optical system, and there is a problem of low efficiency in evaluating the assembly accuracy of the optical system.

[0003] It can be seen that the existing technology has the problem of low efficiency in evaluating the assembly accuracy of optical systems. Summary of the Invention

[0004] The embodiments of the present invention provide a method and device for predicting the assembly accuracy of an optical system based on thermomechanical coupling simulation, so as to solve the problem of low efficiency in evaluating the assembly accuracy of an optical system in the prior art.

[0005] To solve the above problems, the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting the assembly accuracy of an optical system based on thermomechanical coupling simulation, comprising:

[0007] Obtaining assembly parameters and temperature variation parameters of fixed components of the optical system;

[0008] Analyzing the assembly parameters and the temperature-varying parameters based on a thermal-mechanical coupling simulation system to obtain first surface shape data, wherein the thermal-mechanical coupling simulation system is used to analyze the topography change of the surface of the fixed component;

[0009] Performing feature extraction on the first surface shape data to obtain a first surface shape feature;

[0010] The first surface feature is predicted based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

[0011] In one embodiment, the preset prediction model is obtained by:

[0012] Acquire multiple sample surface features and multiple sample energy concentrations;

[0013] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0014] Training each of a plurality of first initial models based on the training set data to obtain a plurality of first intermediate training models, wherein the plurality of first initial models are models constructed to perform energy concentration prediction, and different first initial models include different prediction model parameters;

[0015] Calculate a first error corresponding to each of the plurality of first intermediate training models based on the test set data;

[0016] The first intermediate training model corresponding to the minimum first error is set as the preset prediction model.

[0017] In one embodiment, each of the first initial models further includes first hidden layer parameters, and the first hidden layer parameters are obtained by:

[0018] Acquire multiple initial hidden layer parameters, and construct multiple second initial models based on each initial hidden layer parameter and other parameters, wherein the other parameters include prediction model parameters;

[0019] Acquire multiple sample surface features and multiple sample energy concentrations;

[0020] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0021] Training each of the plurality of second initial models based on the training set data to obtain a second intermediate training model;

[0022] Calculate a second error corresponding to each second intermediate training model in the plurality of second intermediate training models based on the test set data;

[0023] The initial hidden layer parameters of the second intermediate training model corresponding to the second minimum error are set as the first hidden layer parameters.

[0024] In one embodiment, each of the first initial models further includes a first activation function, and the first activation function is obtained as follows:

[0025] Obtaining multiple initial activation functions, and constructing multiple third initial models based on each initial activation function and other parameters, wherein the other parameters include prediction model parameters;

[0026] Acquire multiple sample surface features and multiple sample energy concentrations;

[0027] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0028] Training each of the plurality of third initial models based on the training set data to obtain a third intermediate training model;

[0029] Calculating a third error corresponding to each third intermediate training model in the plurality of third intermediate training models based on the test set data;

[0030] The initial activation function of the third intermediate training model corresponding to the smallest third error is set to the first activation function.

[0031] In one embodiment, the obtaining of multiple sample surface features and multiple sample energy concentrations includes:

[0032] Acquire a plurality of sample parameter data, wherein each sample parameter data includes a sample assembly parameter and a sample temperature change parameter;

[0033] Analyzing each sample parameter data based on the thermal-mechanical coupling simulation system to obtain a plurality of sample surface shape data;

[0034] Performing feature extraction on each sample face shape data to obtain the plurality of sample face shape features;

[0035] The surface features of each sample are simulated to obtain multiple sample energy concentrations.

[0036] In one embodiment, before extracting features from each sample face shape data to obtain the plurality of sample face shape features, the training process of the preset prediction model further includes:

[0037] Acquire a posture error feature, where the posture error feature is used to characterize an error condition existing during the assembly process of the optical system;

[0038] The step of extracting features from each sample face shape data to obtain the plurality of sample face shape features comprises:

[0039] Performing feature extraction on each sample face shape data to obtain a plurality of first intermediate features;

[0040] The multiple first intermediate features are combined with the pose error features to obtain the multiple sample surface features.

[0041] In a second aspect, an embodiment of the present invention further provides an optical system assembly accuracy prediction device based on thermomechanical coupling simulation, comprising:

[0042] An acquisition module, used to acquire assembly parameters and temperature variation parameters of fixed components of the optical system;

[0043] an analysis module, configured to analyze the assembly parameters and the temperature variation parameters based on a thermal coupling simulation system to obtain first surface shape data, wherein the thermal coupling simulation system is configured to analyze changes in the surface shape of the fixed component;

[0044] An extraction module, configured to perform feature extraction on the first surface shape data to obtain first surface shape features;

[0045] A prediction module is used to predict the first surface feature based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

[0046] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation as described in the first aspect above.

[0047] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation as described in the first aspect above.

[0048] In a fifth aspect, the present invention further provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation as described in the first aspect above.

[0049] In an embodiment of the present invention, assembly parameters and temperature-varying parameters of a fixed component of an optical system are obtained; the assembly parameters and temperature-varying parameters are analyzed based on a thermomechanical coupling simulation system to obtain first surface shape data, wherein the thermomechanical coupling simulation system is used to analyze the topographical changes of the surface of the fixed component; feature extraction is performed on the first surface shape data to obtain a first surface shape feature; the first surface shape feature is predicted based on a preset prediction model to obtain a first energy concentration, which is used to characterize the assembly accuracy of the optical system, wherein the preset prediction model is a model trained to predict energy concentration. In this way, the first surface shape data of the fixed component of the optical system under temperature-varying parameters is obtained by analyzing the thermomechanical coupling simulation system, and the first surface shape feature is extracted from the first surface shape data, which is then predicted using the preset prediction model to obtain the first energy concentration. The assembly accuracy of the optical system can be determined based on the first energy concentration, thereby achieving rapid determination of the assembly accuracy of the optical system without requiring long-term measurement of the optical system, thereby improving the efficiency of evaluating the assembly accuracy of the optical system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0051] Figure 1 This is a flow chart of a method for predicting optical system assembly accuracy based on thermal-mechanical coupling simulation provided by an embodiment of the present invention;

[0052] Figure 2 is a schematic diagram of a fixing component provided by an embodiment of the present invention;

[0053] Figure 3 Schematic diagram of the structure of a CNN provided by an embodiment of the present invention;

[0054] Figure 4 This is one of the stress distribution diagrams provided by the embodiment of the present invention;

[0055] Figure 5 This is the second stress distribution diagram provided by an embodiment of the present invention;

[0056] Figure 6 is a schematic diagram of average errors corresponding to different regression parameters provided by an embodiment of the present invention;

[0057] Figure 7 Schematic diagram of the structure of the MLP model provided by an embodiment of the present invention;

[0058] Figure 8 is a schematic diagram of average errors corresponding to different initial hidden layer parameters provided by an embodiment of the present invention;

[0059] Figure 9 Schematic diagram of average errors corresponding to different initial activation functions provided by an embodiment of the present invention;

[0060] Figure 10 is a flow chart of calculating sample energy concentration provided by an embodiment of the present invention;

[0061] Figure 11 is a schematic diagram of sample temperature variation parameters provided by an embodiment of the present invention;

[0062] Figure 12 is a schematic diagram of sample surface features provided by an embodiment of the present invention;

[0063] Figure 13 is a schematic diagram of the prediction results of the preset prediction model provided by an embodiment of the present invention;

[0064] Figure 14 is a schematic diagram of the prediction error of the preset prediction model provided by an embodiment of the present invention;

[0065] Figure 15 1 is a structural diagram of an optical system assembly accuracy prediction device based on thermal-mechanical coupling simulation provided by an embodiment of the present invention;

[0066] Figure 16 This is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0068] See Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for predicting optical system assembly accuracy based on thermal-mechanical coupling simulation provided by an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0069] Step 101: Obtain assembly parameters and temperature variation parameters of fixed components of an optical system.

[0070] The above-mentioned fixing parts are parts used to fix different components in the optical system, and can usually be a structure with bolts and nuts. Figure 2As shown, plate one and plate two are fixed by bolts and nuts.

[0071] The aforementioned assembly parameters are those of the fixing components, which enable the fixing components to secure the components of the optical system. For example, if the fixing components are bolts and nuts, the assembly parameters are the tightening angles of the nuts, which can range from 2° to 6°.

[0072] The above temperature variation parameters are used to simulate the temperature variation of the use environment of the optical system. The temperature variation parameters may be a temperature variation curve. The temperature variation curve may be set based on experience or obtained by collecting the use environment temperature of the optical system.

[0073] Step 102: Analyze the assembly parameters and the temperature-varying parameters based on a thermomechanical coupling simulation system to obtain first surface shape data. The thermomechanical coupling simulation system is used to analyze the topography changes of the surface of the fixed component.

[0074] The above-mentioned thermal coupling simulation system is used to simulate the morphological changes of the surface of the fixed component according to the input parameters (i.e., assembly parameters and temperature change parameters). In the present invention, simulation is performed through the thermal coupling simulation system to quickly obtain the first surface shape data of the fixed component.

[0075] The first surface shape data is the surface shape data of the contact surface of the fixing component when it is fixed. For example, if the fixing component is a bolt or nut, the first surface shape data is the surface shape data of the threaded surface where the bolt or nut contacts. It should be noted that when the temperature of the optical system changes, the fixing component deforms, and the surface topography of the fixing component changes. The first surface shape data is obtained through simulation to determine the topographic changes of the fixing component under different assembly parameters and different temperature conditions.

[0076] Step 103: extract features from the first surface shape data to obtain first surface shape features.

[0077] The first surface feature can characterize the topography of the surface of the fixing component. The power concentration of the fixing component can be predicted through the first surface feature, and the assembly accuracy of the optical system can be determined through the power concentration.

[0078] The first face shape data is subjected to feature extraction to obtain the first face shape feature. Specifically, a convolutional neural network (CNN) is used to extract the first face shape feature. The structure of CNN is as follows: Figure 3 As shown, it includes sequentially connected convolution layers, maximum pooling layers, residual blocks, average pooling layers and full connection layers. The residual block includes multiple convolution layers. By inputting the first face shape data into the Figure 3The CNN shown in Figure 1 is used to obtain the first face shape feature.

[0079] Step 104 : Predict the first surface feature based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

[0080] The above-mentioned preset prediction model is a model trained to predict energy concentration based on surface features. The first surface feature is input into the preset prediction model, and the first energy concentration corresponding to the first surface feature is predicted by the preset prediction model.

[0081] It should be noted that the energy concentration can reflect the deformation of the fixed components, such as Figure 4 and Figure 5 As shown, Figure 4 and Figure 5 is the stress distribution diagram of the fixed parts under different temperature conditions. Figure 4 The corresponding energy concentration ratio Figure 5 The corresponding energy concentration is low, through Figure 4 and Figure 5 It can be determined that the deformation degrees exhibited by the two are different. In this way, the energy concentration is calculated by calculating the assembly parameters and temperature change parameters of the fixed components. The energy concentration can be used to quantify the deformation of the fixed components under different temperature conditions, so that the assembly accuracy of the optical system can be determined by the energy concentration.

[0082] In an embodiment of the present invention, assembly parameters and temperature-varying parameters of a fixed component of an optical system are obtained; the assembly parameters and temperature-varying parameters are analyzed based on a thermomechanical coupling simulation system to obtain first surface shape data, wherein the thermomechanical coupling simulation system is used to analyze the topographical changes of the surface of the fixed component; feature extraction is performed on the first surface shape data to obtain a first surface shape feature; the first surface shape feature is predicted based on a preset prediction model to obtain a first energy concentration, which is used to characterize the assembly accuracy of the optical system, wherein the preset prediction model is a model trained to predict energy concentration. In this way, the first surface shape data of the fixed component of the optical system under temperature-varying parameters is obtained by analyzing the thermomechanical coupling simulation system, and the first surface shape feature is extracted from the first surface shape data, which is then predicted using the preset prediction model to obtain the first energy concentration. The assembly accuracy of the optical system can be determined based on the first energy concentration, thereby achieving rapid determination of the assembly accuracy of the optical system without requiring long-term measurement of the optical system, thereby improving the efficiency of evaluating the assembly accuracy of the optical system.

[0083] In one embodiment, the preset prediction model is obtained by:

[0084] Acquire multiple sample surface features and multiple sample energy concentrations;

[0085] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0086] Training each of a plurality of first initial models based on the training set data to obtain a plurality of first intermediate training models, wherein the plurality of first initial models are models constructed to perform energy concentration prediction, and different first initial models include different prediction model parameters;

[0087] Calculate a first error corresponding to each of the plurality of first intermediate training models based on the test set data;

[0088] The first intermediate training model corresponding to the minimum first error is set as the preset prediction model.

[0089] The aforementioned multiple sample surface features and multiple sample energy concentrations correspond one-to-one. The multiple sample surface features are measured or simulated based on preset assembly parameters and temperature parameters, and the multiple sample energy concentrations are measured or simulated under different special parameters and temperature parameters of the optical system. It can be understood that the multiple sample surface features and the multiple sample energy concentrations are real values, and the preset prediction model can be trained using the multiple sample surface features and the multiple sample energy concentrations.

[0090] The training set data and test set data are obtained by dividing multiple sample face features and multiple sample energy concentrations based on a preset ratio. For example, the preset ratio is 8:2, which means that 80% of the multiple sample face features and multiple sample energy concentrations are divided into the training set data, and the remaining 20% ​​of the multiple sample face features and multiple sample energy concentrations are divided into the test set data. The training set data includes at least one sample face feature and at least one sample energy concentration, and the test set data includes at least one sample face feature and at least one sample energy concentration.

[0091] Each of the multiple first initial models can be used to predict energy concentration, but each first initial model includes different prediction model parameters, and the prediction accuracy of energy concentration using different prediction model parameters also varies. In the present invention, different first initial models are trained using training set data to obtain a first intermediate training model corresponding to each first initial model. The first error is then calculated using test set data to determine the model with the best prediction effect from the multiple first intermediate models as the preset prediction model.

[0092] The prediction model parameters may be model parameters of a multi-layer perceptron (MLP), model parameters of a support vector regression (SVR), model parameters of a random forest regression (RFR), model parameters of a Gaussian process regression (GPR), and model parameters of a lightweight gradient boosting machine (LightGBM). The first error of the first intermediate training model including different prediction model parameters is calculated by the method of an embodiment of the present invention.

[0093] In some embodiments, multiple sample facial features and multiple sample energy concentrations can be divided multiple times based on a preset ratio to obtain multiple groups of training set data and test set data. Model training and error calculation are performed on multiple first initial models through each group of training set data and test set data to obtain the first error corresponding to each first intermediate training model under the conditions of each group of training set data and test set data. Finally, the first errors corresponding to each first intermediate training model are averaged to obtain an average error. The optimal first intermediate training model is determined through the average error, and then the preset prediction model is obtained.

[0094] For example, based on a preset ratio, multiple sample face features and multiple sample energy concentrations are divided 10 times to obtain 10 sets of training set data and test set data, and then 10 first errors corresponding to each first intermediate training model are calculated using the 10 sets of training set data and test set data. The 10 first errors are then averaged to obtain the average error corresponding to each first intermediate training model, such as Figure 6 As shown. Figure 6 From the calculation results, it can be seen from the figure that the random forest regression model shows the worst prediction accuracy, with an average prediction error of 17.6%; the support vector regression model is second, with an average prediction error of 10.3%; then the lightweight gradient boosting machine regression model, with an average prediction error of 6.64%; the average prediction error of the Gaussian process regression model is 5.09%, and the average error corresponding to the first intermediate training model including the MLP model parameters is 3.26%, which is the lowest. At this time, the first intermediate training model including the MLP model parameters can be used as the preset prediction model.

[0095] In an embodiment of the present invention, a plurality of sample face features and a plurality of sample energy concentrations are obtained; the plurality of sample face features and the plurality of sample energy concentrations are divided into training set data and test set data based on a preset ratio; each of the plurality of first initial models is trained based on the training set data to obtain a plurality of first intermediate training models, wherein the plurality of first initial models are models constructed to perform energy concentration prediction, and different first initial models include different prediction model parameters; a first error corresponding to each of the plurality of first intermediate training models is calculated based on the test set data; and the first intermediate training model corresponding to the minimum first error is set as the preset prediction model. In this way, multiple first intermediate training models are obtained through training using the training set data and the test set data, and the first error corresponding to each first intermediate training model is calculated, thereby obtaining a preset prediction model with the minimum first error, so that the preset prediction model can achieve a better prediction effect.

[0096] It should be noted that when the first initial model is an MLP model, the MLP model usually includes an input layer, a hidden layer, and an output layer. The first face shape feature is input from the input layer to the first initial model, and after being processed by the hidden layer and the output layer, the first energy concentration is output. The hidden layer can be set with different parameters, and the output layer can be set with different activation functions. Therefore, different hidden layer parameters and output layer activation functions can be set first, such as Figure 7 As shown, the optimal hidden layer parameters and activation function are screened.

[0097] Specifically, in one embodiment, each of the first initial models further includes first hidden layer parameters, and the first hidden layer parameters are obtained by:

[0098] Acquire multiple initial hidden layer parameters, and construct multiple second initial models based on each initial hidden layer parameter and other parameters, wherein the other parameters include prediction model parameters;

[0099] Acquire multiple sample surface features and multiple sample energy concentrations;

[0100] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0101] Training each of the plurality of second initial models based on the training set data to obtain a second intermediate training model;

[0102] Calculate a second error corresponding to each second intermediate training model in the plurality of second intermediate training models based on the test set data;

[0103] The initial hidden layer parameters of the second intermediate training model corresponding to the second minimum error are set as the first hidden layer parameters.

[0104] In an embodiment of the present invention, multiple initial hidden layer parameters are obtained, and multiple second initial models are constructed based on each initial hidden layer parameter and other parameters, wherein the other parameters include prediction model parameters; multiple sample face features and multiple sample energy concentrations are obtained; the multiple sample face features and the multiple sample energy concentrations are divided into training set data and test set data based on a preset ratio; each of the multiple second initial models is trained based on the training set data to obtain a second intermediate training model; a second error corresponding to each of the multiple second intermediate training models is calculated based on the test set data; and the initial hidden layer parameters of the second intermediate training model corresponding to the minimum second error are set as the first hidden layer parameters. In this way, the first hidden layer parameters with the minimum error are determined by the second error, thereby improving the accuracy of the model prediction of the preset prediction model.

[0105] The initial hidden layer parameters in the above-mentioned multiple second initial models are different, but other parameters are the same, so as to avoid the influence of different other parameters on the second error.

[0106] In some embodiments, multiple sample facial features and multiple sample energy concentrations can be divided multiple times based on a preset ratio to obtain multiple groups of training set data and test set data. Model training and error calculation are performed on multiple second initial models through each group of training set data and test set data to obtain the second error corresponding to each second intermediate training model under the conditions of each group of training set data and test set data. Finally, the second errors corresponding to each second intermediate training model are averaged to obtain the average error. The optimal second intermediate training model is determined through the average error, and the first hidden layer parameters are obtained.

[0107] Exemplarily, 6 groups of different initial hidden layer parameters can be set to obtain 6 second initial models. Among them, the 6 groups of different initial hidden layer parameters are [10 5], [16 8], [20 10 5], [32 16 8], [40 20 10 5], and [64 3216 8], which are named "H1" to "H6" in sequence. The multiple sample face features and the multiple sample energy concentrations are randomly divided into the training set data and the test set data according to the preset ratio of 8:2, and a total of 10 divisions are performed to obtain 10 groups of training set data and test set data. The second error corresponding to each second initial model under each group of training set data and test set data is calculated by the method of the embodiment of the present invention to obtain 10 second errors corresponding to each second initial model, and then the 10 second errors corresponding to each second initial model are averaged to obtain an average error, such as Figure 8As shown in Figure 8, the relative errors of predictions using different initial hidden layer parameters are relatively small, all less than 6%, indicating that the MLP model can accurately capture the true distribution characteristics of the data. Among these hidden layers, H1 (i.e., [10 5]) exhibits the lowest average relative error of only 3.26%, followed by H5 and H3, with average relative errors of 4.25% and 4.37% respectively, followed by H2 and H6, with average relative errors of 4.4% and 4.54% respectively, and H4 has the highest average relative error of 4.56%. Ultimately, selecting H1 as the first hidden layer parameter included in the MLP model can achieve better prediction performance than other initial hidden layer parameters.

[0108] In one embodiment, each of the first initial models further includes a first activation function, and the first activation function is obtained as follows:

[0109] Obtaining multiple initial activation functions, and constructing multiple third initial models based on each initial activation function and other parameters, wherein the other parameters include prediction model parameters;

[0110] Acquire multiple sample surface features and multiple sample energy concentrations;

[0111] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0112] Training each of the plurality of third initial models based on the training set data to obtain a third intermediate training model;

[0113] Calculating a third error corresponding to each third intermediate training model in the plurality of third intermediate training models based on the test set data;

[0114] The initial activation function of the third intermediate training model corresponding to the smallest third error is set to the first activation function.

[0115] In an embodiment of the present invention, multiple initial activation functions are obtained, and multiple third initial models are constructed based on each initial activation function and other parameters, including prediction model parameters; multiple sample face features and multiple sample energy concentrations are obtained; the multiple sample face features and the multiple sample energy concentrations are divided into training set data and test set data based on a preset ratio; each of the multiple third initial models is trained based on the training set data to obtain a third intermediate training model; a third error corresponding to each of the multiple third intermediate training models is calculated based on the test set data; and the initial activation function corresponding to the third intermediate training model with the smallest third error is set as the first activation function. In this way, the activation function with the smallest error is determined by the third error, thereby improving the accuracy of model prediction of the preset prediction model.

[0116] The initial hidden layer parameters in the above-mentioned multiple third initial models are different, but other parameters are the same, so as to avoid the influence of different other parameters on the third error. For example, the first hidden layer parameters are all [10 5].

[0117] In some embodiments, multiple sample facial features and multiple sample energy concentrations can be divided multiple times based on a preset ratio to obtain multiple groups of training set data and test set data. Model training and error calculation are performed on multiple third initial models through each group of training set data and test set data to obtain the third error corresponding to each third intermediate training model under the conditions of each group of training set data and test set data. Finally, the third errors corresponding to each third intermediate training model are averaged to obtain the average error. The optimal third intermediate training model is determined through the average error to obtain the first activation function.

[0118] Exemplarily, three initial activation functions can be set, namely, Sigmoid function, Tanh function and ReLU function. The multiple sample face features and the multiple sample energy concentrations are randomly divided into the training set data and the test set data according to the preset ratio of 8:2, and a total of 10 divisions are performed to obtain 10 groups of training set data and test set data. The third error corresponding to each third initial model under each group of training set data and test set data is calculated by the embodiment of the present invention to obtain 10 third errors corresponding to each third initial model, and then the 10 third errors corresponding to each third initial model are averaged to obtain an average error, such as Figure 9 As shown. Among them, Figure 9 The first activation function is Sigmoid, the second activation function is Tanh, and the third activation function is ReLU. Figure 9As can be seen, the average error of the ReLU activation function was greater than 9%, averaging around 11%, demonstrating poor prediction performance. Using the Tanh activation function, the prediction error exceeded 4% in both test groups, with an average error of 4.02%. In contrast, the Sigmoid activation function had an error of less than 4% across all test cycles, with an average error of only 3.26%. Therefore, the Sigmoid activation function exhibited better prediction performance than the other activation functions, and was therefore selected as the primary activation function.

[0119] In one embodiment, the obtaining of multiple sample surface features and multiple sample energy concentrations includes:

[0120] Acquire a plurality of sample parameter data, wherein each sample parameter data includes a sample assembly parameter and a sample temperature change parameter;

[0121] Analyzing each sample parameter data based on the thermal-mechanical coupling simulation system to obtain a plurality of sample surface shape data;

[0122] Performing feature extraction on each sample face shape data to obtain the plurality of sample face shape features;

[0123] The surface features of each sample are simulated to obtain multiple sample energy concentrations.

[0124] In an embodiment of the present invention, multiple sample parameter data are acquired, each of which includes a sample assembly parameter and a sample temperature variation parameter; each of the sample parameter data is analyzed using the thermomechanical coupling simulation system to obtain multiple sample surface shape data; feature extraction is performed on each sample surface shape data to obtain multiple sample surface shape features; and simulation is performed on each sample surface shape feature to obtain multiple sample energy concentrations. Thus, by setting different sample assembly parameters and sample temperature variation parameters, and then performing analysis and feature extraction, multiple sample surface shape features and multiple sample energy concentrations can be obtained.

[0125] Among them, such as Figure 10 As shown, a structural analysis is performed on the structural model of the optical system to obtain sample surface features. The thermal-mechanical coupling simulation system analyzes each sample parameter data (i.e., the structural model of the optical system) to obtain multiple sample surface features. Feature extraction is then performed on each sample surface data to obtain the multiple sample surface features, which can be obtained by performing surface fitting using Zernike polynomials. Furthermore, each sample surface feature is simulated to obtain multiple sample energy concentrations, which can be obtained by simulating an optical system containing mirror surface errors using Zemax software.

[0126] In some embodiments, in order to enable the prediction preset model to more accurately predict the energy concentration under different temperature change parameters, when setting the sample temperature change parameters, the sample temperature change parameters are first increased and maintained, then decreased and maintained, and finally the above curve, such as Figure 11 In this way, when the prediction preset model is used in the subsequent prediction, the energy concentration under different conditions such as cooling, heating, cooling and maintaining, and heating and maintaining can be accurately predicted.

[0127] In one embodiment, before extracting features from each sample face shape data to obtain the plurality of sample face shape features, the training process of the preset prediction model further includes:

[0128] Acquire a posture error feature, where the posture error feature is used to characterize an error condition existing during the assembly process of the optical system;

[0129] The step of extracting features from each sample face shape data to obtain the plurality of sample face shape features comprises:

[0130] Performing feature extraction on each sample face shape data to obtain a plurality of first intermediate features;

[0131] The multiple first intermediate features are combined with the pose error features to obtain the multiple sample surface features.

[0132] It should be noted that in addition to considering the sample surface features and sample temperature variation parameters, the pose error of the optical system itself also needs to be considered. The pose error of the optical system itself exhibits a normal distribution. The Latin hypercube sampling (LHS) method can be used to optically analyze the optical system analysis model to obtain position error features that characterize the pose error. This is then combined with the first intermediate features corresponding to the sample surface data to obtain the optimized sample surface features.

[0133] Among them, multiple first intermediate features are combined with the posture error features to obtain multiple sample surface features, which can be Figure 12 As shown in the figure, the first intermediate feature is obtained based on CNN and is 512 bits; the pose error feature is obtained by sampling and is 8 bits; multiple first intermediate features are combined with the pose error features to obtain multiple sample face features, and the sample face features are 520 bits.

[0134] In an embodiment of the present invention, obtaining a pose error feature, which is used to characterize errors present during assembly of the optical system; extracting features from each sample surface data to obtain the multiple sample surface features, includes: extracting features from each sample surface data to obtain multiple first intermediate features; and merging the multiple first intermediate features with the pose error feature to obtain the multiple sample surface features. In this way, optimizing the sample surface features using pose error parameters can further improve the model prediction accuracy of the trained preset prediction model.

[0135] For example, when the fixing parts are bolts and nuts, different tightening angles (2° to 6°) are applied to the nuts, the sample temperature change parameter is 10°C / min (can be adjusted according to the actual value), and the holding time is 10min (can be adjusted according to the actual value). A total of 30 different process parameter combinations are designed, and 50 sets of posture deviation parameters corresponding to each process parameter combination are set, generating a total of 1500 sets of sample surface features. Then, by simulating the 1500 sets of sample surface features, 1500 sets of sample energy concentration are obtained. The 1500 sets of data are divided into training set data and test set data in a ratio of 8:2 for model training and testing.

[0136] The model is trained by the training set data to obtain the preset prediction model, and the determination coefficient (R 2 ), Root Mean Square Error (RMSE) and Mean Relative Error (MRE) of each test set data are used to measure the prediction accuracy and generalization ability of the model.

[0137] The results obtained from training are as follows Figure 13 and Figure 14 As shown. Among them, Figure 13 The predicted values ​​for most of the data coincide with or are very close to the true values, with only a few data points exhibiting certain deviations. This demonstrates that the preset prediction model has high reliability and accuracy in predicting energy concentration. Figure 14 The maximum average error is 3.73%, the minimum average error is 2.84%, and the total average error is 3.26%. This shows that under the conditions of multiple repeated tests, the average prediction accuracy of the constructed preset prediction model reaches 96.74%, which can achieve accurate prediction of energy concentration and thus realize rapid evaluation of the assembly accuracy of the optical system.

[0138] See Figure 15 , Figure 15 FIG. 1 is a structural diagram of an optical system assembly accuracy prediction device based on thermal-mechanical coupling simulation provided by an embodiment of the present invention. Figure 15 As shown, the optical system assembly accuracy prediction device 1500 based on thermal-mechanical coupling simulation includes:

[0139] An acquisition module 1501 is used to acquire assembly parameters and temperature variation parameters of fixed components of the optical system;

[0140] An analysis module 1502 is configured to analyze the assembly parameters and the temperature variation parameters based on a thermal coupling simulation system to obtain first surface shape data, wherein the thermal coupling simulation system is configured to analyze changes in the surface shape of the fixed component.

[0141] An extraction module 1503 is configured to perform feature extraction on the first surface shape data to obtain a first surface shape feature;

[0142] The prediction module 1504 is used to predict the first surface feature based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

[0143] In one embodiment, the preset prediction model is obtained by:

[0144] Acquire multiple sample surface features and multiple sample energy concentrations;

[0145] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0146] Training each of a plurality of first initial models based on the training set data to obtain a plurality of first intermediate training models, wherein the plurality of first initial models are models constructed to perform energy concentration prediction, and different first initial models include different prediction model parameters;

[0147] Calculate a first error corresponding to each of the plurality of first intermediate training models based on the test set data;

[0148] The first intermediate training model corresponding to the minimum first error is set as the preset prediction model.

[0149] In one embodiment, each of the first initial models further includes first hidden layer parameters, and the first hidden layer parameters are obtained by:

[0150] Acquire multiple initial hidden layer parameters, and construct multiple second initial models based on each initial hidden layer parameter and other parameters, wherein the other parameters include prediction model parameters;

[0151] Acquire multiple sample surface features and multiple sample energy concentrations;

[0152] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0153] Training each of the plurality of second initial models based on the training set data to obtain a second intermediate training model;

[0154] Calculate a second error corresponding to each second intermediate training model in the plurality of second intermediate training models based on the test set data;

[0155] The initial hidden layer parameters of the second intermediate training model corresponding to the second minimum error are set as the first hidden layer parameters.

[0156] In one embodiment, each of the first initial models further includes a first activation function, and the first activation function is obtained as follows:

[0157] Obtaining multiple initial activation functions, and constructing multiple third initial models based on each initial activation function and other parameters, wherein the other parameters include prediction model parameters;

[0158] Acquire multiple sample surface features and multiple sample energy concentrations;

[0159] Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio;

[0160] Training each of the plurality of third initial models based on the training set data to obtain a third intermediate training model;

[0161] Calculating a third error corresponding to each third intermediate training model in the plurality of third intermediate training models based on the test set data;

[0162] The initial activation function of the third intermediate training model corresponding to the smallest third error is set to the first activation function.

[0163] In one embodiment, the obtaining of multiple sample surface features and multiple sample energy concentrations includes:

[0164] Acquire a plurality of sample parameter data, wherein each sample parameter data includes a sample assembly parameter and a sample temperature change parameter;

[0165] Analyzing each sample parameter data based on the thermal-mechanical coupling simulation system to obtain a plurality of sample surface shape data;

[0166] Performing feature extraction on each sample face shape data to obtain the plurality of sample face shape features;

[0167] The surface features of each sample are simulated to obtain multiple sample energy concentrations.

[0168] In one embodiment, before extracting features from each sample face shape data to obtain the plurality of sample face shape features, the training process of the preset prediction model further includes:

[0169] Acquire a posture error feature, where the posture error feature is used to characterize an error condition existing during the assembly process of the optical system;

[0170] The step of extracting features from each sample face shape data to obtain the plurality of sample face shape features comprises:

[0171] Performing feature extraction on each sample face shape data to obtain a plurality of first intermediate features;

[0172] The multiple first intermediate features are combined with the pose error features to obtain the multiple sample surface features.

[0173] The optical system assembly accuracy prediction device based on thermomechanical coupling simulation provided by an embodiment of the present invention is capable of realizing the various processes of each embodiment of the above-mentioned optical system assembly accuracy prediction method based on thermomechanical coupling simulation. The technical features correspond one to one and can achieve the same technical effects. To avoid repetition, they will not be described here.

[0174] It should be noted that the optical system assembly accuracy prediction device based on thermo-mechanical coupling simulation in the embodiment of the present invention may be a device, or a component, integrated circuit, or chip in an electronic device.

[0175] The present invention also provides an electronic device, see Figure 16 , Figure 16 This is a schematic diagram of the structure of an electronic device provided by the present invention. The electronic device includes a memory 1601, a processor 1602, and a program or instruction stored in the memory 1601 and running on the memory 1601. When the program or instruction is executed by the processor 1602, Figure 1 Any steps in the corresponding embodiment of the optical system assembly accuracy prediction method based on thermal-mechanical coupling simulation and the same beneficial effects are achieved will not be repeated here.

[0176] The processor 1602 may be a CPU, an ASIC, an FPGA, or a GPU.

[0177] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned embodiment of the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation can be completed by hardware related to program instructions, and the program can be stored in a readable medium.

[0178] The present invention also provides a readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above Figure 1 Any steps in the corresponding embodiments of the method for predicting the assembly accuracy of an optical system based on thermal-mechanical coupling simulation can achieve the same technical effect and are not described here in detail to avoid repetition. The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0179] The present invention also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1 The corresponding processes of the embodiment of the optical system assembly accuracy prediction method based on thermal-mechanical coupling simulation can achieve the same technical effect, and will not be described again here to avoid repetition.

[0180] The terms "first", "second" and the like in the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. In addition, the terms "comprise" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices. In addition, "and / or" is used in this application to represent at least one of the connected objects, for example A and / or B and / or C, which means comprising seven situations including single A, single B, single C, and both A and B exist, both B and C exist, both A and C exist, and both A, B and C exist.

[0181] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0182] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) to execute the methods of each embodiment of the present application.

[0183] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. A method for predicting the assembly accuracy of an optical system based on thermal-mechanical coupling simulation, characterized in that: include: Obtaining assembly parameters and temperature variation parameters of fixed components of the optical system; Analyzing the assembly parameters and the temperature-varying parameters based on a thermal-mechanical coupling simulation system to obtain first surface shape data, wherein the thermal-mechanical coupling simulation system is used to analyze the topography change of the surface of the fixed component; Performing feature extraction on the first surface shape data to obtain a first surface shape feature; The first surface feature is predicted based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

2. The method according to claim 1, wherein The preset prediction model is obtained in the following way: Acquire multiple sample surface features and multiple sample energy concentrations; Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio; Training each of a plurality of first initial models based on the training set data to obtain a plurality of first intermediate training models, wherein the plurality of first initial models are models constructed to perform energy concentration prediction, and different first initial models include different prediction model parameters; Calculate a first error corresponding to each of the plurality of first intermediate training models based on the test set data; The first intermediate training model corresponding to the minimum first error is set as the preset prediction model.

3. The method according to claim 2, wherein Each of the first initial models further includes first hidden layer parameters, which are obtained by: Acquire multiple initial hidden layer parameters, and construct multiple second initial models based on each initial hidden layer parameter and other parameters, wherein the other parameters include prediction model parameters; Acquire multiple sample surface features and multiple sample energy concentrations; Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio; Training each of the plurality of second initial models based on the training set data to obtain a second intermediate training model; Calculate a second error corresponding to each second intermediate training model in the plurality of second intermediate training models based on the test set data; The initial hidden layer parameters of the second intermediate training model corresponding to the second minimum error are set as the first hidden layer parameters.

4. The method according to claim 2, wherein Each of the first initial models further includes a first activation function, which is obtained in the following manner: Obtaining multiple initial activation functions, and constructing multiple third initial models based on each initial activation function and other parameters, wherein the other parameters include prediction model parameters; Acquire multiple sample surface features and multiple sample energy concentrations; Dividing the plurality of sample face features and the plurality of sample energy concentrations into training set data and test set data based on a preset ratio; Training each of the plurality of third initial models based on the training set data to obtain a third intermediate training model; Calculating a third error corresponding to each third intermediate training model in the plurality of third intermediate training models based on the test set data; The initial activation function of the third intermediate training model corresponding to the smallest third error is set to the first activation function.

5. The method according to any one of claims 2 to 4, characterized in that The obtaining of multiple sample surface features and multiple sample energy concentrations includes: Acquire a plurality of sample parameter data, wherein each sample parameter data includes a sample assembly parameter and a sample temperature change parameter; Analyzing each sample parameter data based on the thermal-mechanical coupling simulation system to obtain a plurality of sample surface shape data; Performing feature extraction on each sample face shape data to obtain the plurality of sample face shape features; The surface features of each sample are simulated to obtain multiple sample energy concentrations.

6. The method according to claim 5, wherein Before extracting features from each sample face shape data to obtain the plurality of sample face shape features, the training process of the preset prediction model further includes: Acquire a posture error feature, where the posture error feature is used to characterize an error condition existing during the assembly process of the optical system; The step of extracting features from each sample face shape data to obtain the plurality of sample face shape features comprises: Performing feature extraction on each sample face shape data to obtain a plurality of first intermediate features; The multiple first intermediate features are combined with the pose error features to obtain the multiple sample surface features.

7. A method for predicting the assembly accuracy of an optical system based on thermal-mechanical coupling simulation, characterized in that: include: An acquisition module, used to acquire assembly parameters and temperature variation parameters of fixed components of the optical system; an analysis module, configured to analyze the assembly parameters and the temperature variation parameters based on a thermal coupling simulation system to obtain first surface shape data, wherein the thermal coupling simulation system is configured to analyze changes in the surface shape of the fixed component; An extraction module, configured to perform feature extraction on the first surface shape data to obtain first surface shape features; A prediction module is used to predict the first surface feature based on a preset prediction model to obtain a first energy concentration, where the first energy concentration is used to characterize the assembly accuracy of the optical system. The preset prediction model is a model trained to predict energy concentration.

8. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation are implemented as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the optical system assembly accuracy prediction method based on thermo-mechanical coupling simulation according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the optical system assembly accuracy prediction method based on thermal-mechanical coupling simulation as described in any one of claims 1 to 6.

Images