CN115392138A - Optical-mechanical-thermal coupling analysis model based on machine learning - Google Patents

Optical-mechanical-thermal coupling analysis model based on machine learning Download PDF

Info

Publication number
CN115392138A
CN115392138A CN202211322033.9A CN202211322033A CN115392138A CN 115392138 A CN115392138 A CN 115392138A CN 202211322033 A CN202211322033 A CN 202211322033A CN 115392138 A CN115392138 A CN 115392138A
Authority
CN
China
Prior art keywords
model
mechanical
feedforward neural
data set
neural network
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202211322033.9A
Other languages
Chinese (zh)
Other versions
CN115392138B (en
Inventor
李强
高政旺
武春风
姜永亮
胡黎明
韩西萌
刘利民
朱梦楠
王旭锋
胡金萌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Space Sanjiang Group Co Ltd
Original Assignee
China Space Sanjiang Group Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Space Sanjiang Group Co Ltd filed Critical China Space Sanjiang Group Co Ltd
Priority to CN202211322033.9A priority Critical patent/CN115392138B/en
Publication of CN115392138A publication Critical patent/CN115392138A/en
Application granted granted Critical
Publication of CN115392138B publication Critical patent/CN115392138B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Exposure And Positioning Against Photoresist Photosensitive Materials (AREA)

Abstract

The invention provides an optical-mechanical-thermal coupling analysis model based on machine learning, which is characterized in that a feedforward neural network model is constructed by acquiring an optical-mechanical-thermal coupling analysis related data set of an optical system and forming an input/output matrix by using the data set, and then training, evaluating and testing the feedforward neural network model; by the method, machine learning and optics can be combined, the purpose of predicting the optical-mechanical-thermal coupling analysis of the optical system by using data driving is achieved, the establishment of a complex optical system optical-mechanical-thermal coupling simulation mathematical equation is avoided, the difficulty of the optical-mechanical-thermal coupling analysis is reduced, and the accuracy of coupling simulation is improved.

Description

Optical-mechanical-thermal coupling analysis model based on machine learning
Technical Field
The invention relates to the technical field of optical-mechanical-thermal multi-physical field coupling, in particular to an optical-mechanical-thermal coupling analysis model based on machine learning.
Background
Uneven distribution of heat and temperature across the optical element can cause structural distortion of the optical element or create wavefront aberrations, significantly limiting the performance of the optical system. In order to analyze, predict and correct thermally induced wavefront aberrations, heat flux or other factors causing temperature gradients can be linked to the temperature, distortion and Optical aberrations of the system by creating transient models, called "Structural-Thermal-mechanical-Performance-analysis" (STOP) models, which present many challenges to develop accurate STOP models due to the coupling of Thermal, structural mechanical and Optical equations, as well as different boundary conditions and physical parameters.
However, to accurately predict wavefront aberrations, a lower dimensional STOP model is required. The currently common method for reducing dimensionality is based on Model Order Reduction, book Model Order Reduction by Schilders: theory, research, aspects and Application proposes that the sparsity of a system matrix and a model reduction algorithm can be utilized to realize the significant reduction of the dimension of a system model and capture the relevant dynamics of a high-order system. However, before the model can be applied, it is necessary to ensure that the STOP model is validated through extensive experimentation, and to adjust the parameters of the STOP model to ensure that the model predictions match the observed experimental data. Then, the parameters of the STOP model cannot be directly determined, and can only be continuously estimated according to experimental data, and in addition, a plurality of state variables need to be introduced into the STOP model in a complex optical system. Therefore, when the optical-mechanical thermal coupling analysis of the optical system is performed according to the existing STOP model, the mathematical calculation process is complex, the simulation time is long, and the method is not suitable for practical application.
In view of the above, there is a need to design an improved optical-mechanical-thermal coupling analysis model based on machine learning to solve the above problems.
Disclosure of Invention
The invention aims to provide an optical-mechanical-thermal coupling analysis model based on machine learning.
In order to achieve the above object, the present invention provides an optical-mechanical-thermal coupling analysis model based on machine learning, comprising the following steps:
s1, carrying out optical-mechanical-thermal coupling analysis on an optical system in an experiment or numerical simulation mode, and obtaining data sets under different time windows p, wherein the data sets comprise a training data set, a verification data set and a test data set which are mutually independent;
s2, converting the data set in the step S1 into an input and output matrix to construct a feedforward neural network model;
and S3, training, evaluating and testing the feedforward neural network model.
Preferably, in step S2, the input and output of the feedforward neural network model are the same as the input and output of the VARX model.
Preferably, the expression of the VARX model is as follows:
Figure 702892DEST_PATH_IMAGE001
wherein, subscript
Figure 682350DEST_PATH_IMAGE002
Which represents a discrete point in time of the time,
Figure 275136DEST_PATH_IMAGE003
in order to output the vector, the vector is output,
Figure 575405DEST_PATH_IMAGE004
in order to input the vector, the vector is input,
Figure 733985DEST_PATH_IMAGE005
and
Figure 884344DEST_PATH_IMAGE006
the matrix product of the output vector and the input vector, respectively, p is the past time window point,
Figure 462962DEST_PATH_IMAGE007
is a zero mean white noise sequence.
Preferably, the output vector is a wavefront aberration observed at an image plane of the optical system, and the input vector is thermal power of a heater in the optical system and external thermal disturbance.
Preferably, the input and output data matrix formats learned by the VARX model are respectively as follows:
Figure 458600DEST_PATH_IMAGE008
and
Figure 409369DEST_PATH_IMAGE009
wherein, in the step (A),
Figure 933891DEST_PATH_IMAGE010
is a matrix of machine-learned input data,
Figure 16117DEST_PATH_IMAGE011
is a machine learning output data matrix, each row of matrix U is used to predict a corresponding row of matrix Y.
Preferably, the input layer of the feedforward neural mesh model is composed of
Figure 267976DEST_PATH_IMAGE012
The output layer of the feedforward neural grid model consists of r nodes.
Preferably, in step S3, the feedforward neural lattice model is trained, evaluated and tested by using open source TensorFlow, and the specific process of training, evaluating and testing is as follows: firstly, training the feedforward neural network model by using the training data set and the verification data set obtained in the step S1, wherein the training data set is used for fitting parameters of the feedforward neural network model, the verification data set is used for carrying out hyper-parameter optimization on the fitted feedforward neural network model, and in the training process, a mean square error is used as a loss function, and an adopted training method is a gradient descent algorithm; in the training process, the training data set and the verification data set under each time window p can correspondingly obtain a trained feedforward neural network model; and finally, evaluating all the trained feedforward neural network models by using the test data set, and taking the feedforward neural network model corresponding to the minimum relative error between the prediction result output by the evaluated feedforward neural network model and the output result in the test data set as the optimal feedforward neural network model.
Preferably, the gradient descent algorithm is a root mean square propagation optimization algorithm.
Preferably, the wavefront aberration of the optical system can be predicted by using the machine learning-based optical-mechanical-thermal coupling analysis model, wherein the wavefront aberration is expressed by Zernike coefficients, and the Zernike coefficients are mainly a translation term, a tilt term and a defocus term.
Preferably, in step S1, the optical system includes a newton telescope system.
The invention has the beneficial effects that:
the optical-mechanical-thermal coupling analysis model based on machine learning and the verification framework are constructed by combining the neural network in the machine learning with optics, the wave aberration of the optical system can be rapidly and accurately predicted by adopting the method, and the important reference value is achieved for performing feedback compensation on the aberration caused by heat in the optical system. In addition, the model provided by the invention can realize the prediction of the wave aberration of the optical system only by data driving, avoids the establishment of a mathematical equation of an optical-mechanical-thermal coupling analysis model of a complex optical system, reduces the difficulty of calculating the wave aberration of the optical system, can be applied to the fields of adaptive optics and the like, and has the characteristics of wide application range, low prediction time cost, high accuracy and the like.
Drawings
FIG. 1 is a flow chart of an optical-mechanical-thermal coupling analysis model based on machine learning of the present invention;
FIG. 2 is a schematic view showing a configuration of a Newtonian telescope system according to embodiment 1 of the present invention;
FIG. 3 is a graph of the relative error between the model prediction output and the true output of the system for the test data set trained at different time windows in example 1 of the present invention;
FIG. 4 is a comparison graph of defocus terms of a predicted and real system based on a feedforward neural network model in embodiment 1 of the present invention;
the reference numbers are as follows:
1. a primary mirror; 2. a mirror support structure; 3. an image plane; 4. a secondary mirror; 5. light transmission obstacles; 6. external thermal interference; 7. an array of wells; 8. the heater location.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
It should be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and/or processing steps closely related to the aspects of the present invention are shown in the drawings, and other details not closely related to the present invention are omitted.
In addition, it is also to be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Referring to fig. 1, the optical-mechanical-thermal coupling analysis model based on machine learning according to the present invention includes the following steps:
s1, obtaining a data set;
performing optical-mechanical-thermal coupling analysis on the optical system in an experiment or numerical simulation mode, and acquiring data sets under different time windows p, wherein the data sets comprise a training data set, a verification data set and a test data set which are mutually independent;
s2, forming an input and output matrix and constructing a feedforward neural network model;
converting the data set in the step S1 into an input-output matrix to construct a feedforward neural network model; the inputs and outputs of the feedforward neural mesh model are the same as those of the VARX model, which is expressed as follows:
Figure 322519DEST_PATH_IMAGE013
wherein, in the step (A),
Figure 814680DEST_PATH_IMAGE014
which represents a discrete point in time of the time,
Figure 338197DEST_PATH_IMAGE015
the output vector, here the wavefront aberration observed at the image plane of the optical system,
Figure 675637DEST_PATH_IMAGE016
is an input vector, which is the thermal power of the heater in the optical system and the external thermal disturbance,
Figure 427430DEST_PATH_IMAGE017
and
Figure 824914DEST_PATH_IMAGE018
respectively, the matrix product of the output vector and the input vector, p being the time window in the pastThe mouth point is arranged at the position of the mouth,
Figure 101305DEST_PATH_IMAGE019
is a white noise sequence with zero mean, and the VARX model is a Vector AutoRegressive eXogenous model (Vector AutoRegressive eXogenous).
For predicting system output
Figure 242437DEST_PATH_IMAGE020
Using the already acquired system input data set
Figure 317578DEST_PATH_IMAGE021
And system output data set
Figure 964591DEST_PATH_IMAGE022
The input and output matrix formats of the feedforward neural network model learning can be expressed as:
Figure 977546DEST_PATH_IMAGE023
and
Figure 171636DEST_PATH_IMAGE024
wherein, in the step (A),
Figure 789699DEST_PATH_IMAGE010
is the input data matrix of the feedforward neural network model,
Figure 76455DEST_PATH_IMAGE025
is the output data matrix of the feedforward neural network model, and each row of the matrix U is used for a corresponding row of the prediction matrix Y. That is, the rows of the matrix U should be used as feed-forward neural network inputs, while Y is used as the output of the feed-forward neural network. Therefore, the input layer of the feedforward neural net model for training should be composed of
Figure 45548DEST_PATH_IMAGE026
Each node is composed of r nodes, the output layer is composed of r nodes, and the number of internal layers is determinedAnd adjusting the optimization variables.
S3, training, evaluating and testing a feedforward neural network model:
training, evaluating and testing the neural network model by using an open source TensorFlow platform, wherein the training, evaluating and testing process comprises the following steps: firstly, carrying out training of a feedforward neural network model by using a training data set and a verification data set obtained in the step S1, wherein the training data set is used for fitting parameters of the feedforward neural network model, the verification data set is used for carrying out hyperparametric optimization on the fitted model, in the training process, a mean square error is used as a loss function, and an adopted training method is a root mean square propagation optimization algorithm in a gradient descent algorithm; in the model training process, a trained feedforward neural network model can be obtained by corresponding a training data set and a verification data set under each time window p; and finally, evaluating the prediction effects of all the trained feedforward neural network models by using the test data set, and taking the feedforward neural network model corresponding to the time when the relative error between the prediction result output by the feedforward neural network model and the output result in the test data set is minimum as the optimal feedforward neural network model.
It should be noted that the reason why the feedforward neural network model and the VARX model are selected as the model structures in the present invention is as follows: once the parameters of the neural network are learned, a VARX model can be easily constructed from the learned parameters, and the constructed VARX model can also be used for designing advanced estimation and feedback control algorithms; second, the feedforward neural network structure is easier to learn than the recurrent neural network structure.
The optical-mechanical-thermal coupling analysis model based on machine learning of the present invention is further illustrated below with reference to specific examples:
example 1
The embodiment provides an optical-mechanical-thermal coupling analysis model based on machine learning, which comprises the following steps:
s1, obtaining a data set:
performing optical-mechanical-thermal coupling analysis on the optical system in an experimental or numerical simulation mode, and acquiring a data set under each time window of which the time window p =0 to 24, wherein the data set comprises a training data set, a verification data set and a test data set which are independent of each other:
referring to fig. 2, the optical system adopted in this embodiment is a newton telescope system, and includes a primary mirror 1, and an image plane 3 and a secondary mirror 4 disposed on the front side of the primary mirror 1, the primary mirror 1 is supported and fixed by a mirror support structure 2, a light propagation barrier 5 is disposed on the secondary mirror 4 on the side far from the primary mirror 1, a hole array 7 is disposed on the primary mirror 1 on the side far from the secondary mirror 4, the hole array 7 is used for placing a heater or a cooler, and a thermocouple for measuring a temperature field, in this embodiment, 9 heaters 8 are disposed in the hole array 7, and the heaters 8 can be used for developing a feedback temperature control system to achieve thermal stability in space and time, thereby ensuring that wavefront aberration and system structural deformation caused by external thermal interference are kept below a maximum allowable value. With the heater 8 and the external thermal disturbance 6 as inputs to the optical system and the Zernike coefficients of the wavefront aberrations observed at the image plane 3 as outputs from the optical system, only the main three terms of the Zernike coefficients, the translation, tilt and defocus terms, respectively, are selected here. Three statistically independent sets of input data were simulated by COMSOL software to generate statistically independent training, validation and testing data sets, yielding 301 data samples in total. It should be noted that, in other embodiments, the number of the heaters 8 may also be adjusted according to actual needs, and the disclosure is not limited thereto.
S2, forming an input and output matrix, and constructing a feedforward neural network model:
converting any data set in the step S1 into an input-output matrix to construct a feedforward neural network model; the inputs and outputs of the feedforward neural mesh model are the same as those of the VARX model, which is expressed as follows:
Figure 777750DEST_PATH_IMAGE027
wherein, subscript
Figure 578216DEST_PATH_IMAGE014
Which represents a discrete point in time of the time,
Figure 504714DEST_PATH_IMAGE015
the output vectors, in particular the three main Zernike coefficients of the wavefront aberration observed at the image plane 3, i.e., the translation term, the tilt term and the defocus term,
Figure 492262DEST_PATH_IMAGE028
as input vectors, specifically heater 8 and external thermal disturbance 6,
Figure 762575DEST_PATH_IMAGE017
and
Figure 151968DEST_PATH_IMAGE029
the matrix product of the output vector and the input vector, respectively, p is the past time window point,
Figure 701898DEST_PATH_IMAGE007
is a white noise sequence with zero mean value, and the simulation of the embodiment
Figure 193054DEST_PATH_IMAGE030
Is 0.
To predict system output
Figure 126112DEST_PATH_IMAGE020
Input of a data set using a system in which a simulation has been developed is required
Figure 651903DEST_PATH_IMAGE021
And the simulated system output data set
Figure 294105DEST_PATH_IMAGE031
The input and output matrix formats of the feedforward neural network model learning can be expressed as:
Figure 256245DEST_PATH_IMAGE032
and
Figure 930678DEST_PATH_IMAGE033
，
Figure 294663DEST_PATH_IMAGE010
is the input data matrix of the feedforward neural network model,
Figure 937128DEST_PATH_IMAGE025
is the output data matrix of the feedforward neural network model, and each row of the matrix U is used for a corresponding row of the prediction matrix Y. That is, the rows of the matrix U are used as inputs to the feedforward neural network model, and Y is used as an output of the feedforward neural network model. The training results of this example show that only the nodes of the input layer of the trained feedforward neural lattice model are
Figure 386564DEST_PATH_IMAGE026
When the number of nodes of the output layer is r, the feedforward neural network model can accurately predict the performance of the system, and the feedforward neural network model constructed in the embodiment only comprises the input layer and the output layer.
S3, training, evaluating and testing a feedforward neural network model:
training, evaluating and testing the feedforward neural network model by using an open-source TensorFlow platform, wherein the training, evaluating and testing process comprises the following steps: firstly, training a feedforward neural network model by using a training data set and a verification data set obtained in the step S1, wherein the training data set is used for fitting parameters of the feedforward neural network model, and the verification data set is used for carrying out hyper-parameter optimization on the fitted model; in the training process, the mean square error is used as a loss function, and the adopted training method is a root mean square propagation optimization algorithm in a gradient descent algorithm; in the training process, a trained feedforward neural network model can be obtained by corresponding the training data set and the verification data set in each time window p, and the embodiment has 24 training models; finally, the prediction effects of the 24 training models are evaluated by using the test data set, the relative error between the prediction results output by the training models in different time windows and the output data of the test data is used as an evaluation index, the relative error between the prediction results and the output data of the test data is shown in fig. 3, and it can be seen from fig. 3 that the relative error is the smallest when the time window p =10, so that the training model corresponding to p is selected as the optimal feedforward neural network model.
Further, a set of data is arbitrarily extracted from the test data set to verify the prediction result, the distribution of the predicted defocus term of the Zernike coefficient and the real output result is shown in fig. 4, and it can be seen from fig. 4 that the defocus term of the Zernike coefficient output by prediction is substantially consistent with the real result, and the result shows that the wavefront aberration of the optical system can be accurately predicted by using the analysis model of the embodiment.
In summary, according to the optical-mechanical-thermal coupling analysis model based on machine learning provided by the invention, the relevant data set is firstly obtained, then the feedforward neural network model is constructed and optimized, and then the optimized feedforward neural network model is used for predicting the wavefront aberration of the optical system. The optical-mechanical-thermal coupling analysis model based on machine learning has the characteristics of wide application range, low prediction time cost, high accuracy and the like, and has an important reference value for performing feedback compensation on aberration caused by heat in an optical system.
Although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the spirit and scope of the present invention.

Claims (10)

1. An optical-mechanical-thermal coupling analysis model based on machine learning, comprising the steps of:
s1, carrying out optical-mechanical-thermal coupling analysis on an optical system in an experiment or numerical simulation mode, and obtaining data sets under different time windows p, wherein the data sets comprise a training data set, a verification data set and a test data set which are mutually independent;
s2, converting the data set in the step S1 into an input and output matrix to construct a feedforward neural network model;
and S3, training, evaluating and testing the feedforward neural network model.
2. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 1, wherein in step S2, the inputs and outputs of the feedforward neural network model are the same as the inputs and outputs of the VARX model.
3. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 2, wherein the expression of the VARX model is as follows:
Figure 583239DEST_PATH_IMAGE001
wherein, subscript
Figure 270573DEST_PATH_IMAGE002
Which represents a discrete point in time of the time,
Figure 942994DEST_PATH_IMAGE003
is the output vector of the output vector,
Figure 660152DEST_PATH_IMAGE004
is the input vector of the input vector,
Figure 99223DEST_PATH_IMAGE005
and
Figure 250850DEST_PATH_IMAGE006
the product of the input and output matrices, respectively, p is the past time window point,
Figure 530390DEST_PATH_IMAGE007
is a zero mean white noise sequence.
4. The machine-learning-based opto-mechanical-thermal coupling analysis model according to claim 3, wherein the output vector is a wavefront aberration observed at an image plane of the optical system, and the input vector is thermal power of a heater in the optical system and external thermal disturbance.
5. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 3, wherein the input and output data matrix formats of the VARX model learning are respectively:
Figure 970730DEST_PATH_IMAGE008
and
Figure 213493DEST_PATH_IMAGE009
wherein, in the step (A),
Figure 344260DEST_PATH_IMAGE010
is a matrix of machine-learned input data,
Figure 60281DEST_PATH_IMAGE011
is a machine learning output data matrix, each row of matrix U is used to predict a corresponding row of matrix Y.
6. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 1, wherein the input layer of the feedforward neural mesh model is defined by
Figure 706026DEST_PATH_IMAGE012
The output layer of the feedforward neural grid model consists of r nodes.
7. The machine-learning-based opto-mechanical-thermal coupling analysis model according to claim 1, wherein in step S3, the feedforward neural net model is trained, evaluated and tested by using open-source TensorFlow, and the specific process of training, evaluating and testing is as follows: firstly, training the feedforward neural network model by using the training data set and the verification data set obtained in the step S1, wherein the training data set is used for fitting parameters of the feedforward neural network model, the verification data set is used for carrying out hyper-parameter optimization on the fitted feedforward neural network model, and in the training process, a mean square error is used as a loss function, and an adopted training method is a gradient descent algorithm; in the training process, the training data set and the verification data set under each time window p can correspondingly obtain a trained feedforward neural network model; and finally, evaluating all the trained feedforward neural network models by using the test data set, and taking the feedforward neural network model corresponding to the condition that the relative error between the prediction result output by the evaluated feedforward neural network model and the output result in the test data set is minimum as the optimal feedforward neural network model.
8. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 7, wherein the gradient descent algorithm is a root mean square propagation optimization algorithm.
9. The machine-learning based optical-mechanical-thermal coupling analysis model of claim 1, wherein the machine-learning based optical-mechanical-thermal coupling analysis model is utilized to predict wavefront aberration of the optical system, the wavefront aberration is expressed as Zernike coefficients, and the Zernike coefficients are mainly a translation term, a tilt term and a defocus term.
10. The machine-learning based opto-mechanical-thermal coupling analytical model of claim 1, wherein in step S1 the optical system comprises a newtonian telescope system.
CN202211322033.9A 2022-10-27 2022-10-27 Optical-mechanical-thermal coupling analysis model based on machine learning Active CN115392138B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211322033.9A CN115392138B (en) 2022-10-27 2022-10-27 Optical-mechanical-thermal coupling analysis model based on machine learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211322033.9A CN115392138B (en) 2022-10-27 2022-10-27 Optical-mechanical-thermal coupling analysis model based on machine learning

Publications (2)

Publication Number Publication Date
CN115392138A true CN115392138A (en) 2022-11-25
CN115392138B CN115392138B (en) 2023-01-24

Family

ID=84128557

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211322033.9A Active CN115392138B (en) 2022-10-27 2022-10-27 Optical-mechanical-thermal coupling analysis model based on machine learning

Country Status (1)

Country Link
CN (1) CN115392138B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180341859A1 (en) * 2017-05-24 2018-11-29 Southwest Research Institute Detection of Hazardous Leaks from Pipelines Using Optical Imaging and Neural Network
CN113239615A (en) * 2021-04-21 2021-08-10 西北工业大学 Flight visual display system design method based on machine learning
WO2022031982A1 (en) * 2020-08-05 2022-02-10 The George Washington University Massively parallel amplitude-only optical processing system and methods for machine learning
CN114334041A (en) * 2021-12-31 2022-04-12 厦门大学 Transformer-based electromagnetic metamaterial complex spectrum high-precision prediction method
CN114418287A (en) * 2021-12-10 2022-04-29 国网浙江省电力有限公司平阳县供电公司 Micro-grid frequency safety evaluation method containing electricity-gas-thermal coupling system
US20220138549A1 (en) * 2019-07-18 2022-05-05 Huawei Technologies Co., Ltd. Photonic Signal Processing

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180341859A1 (en) * 2017-05-24 2018-11-29 Southwest Research Institute Detection of Hazardous Leaks from Pipelines Using Optical Imaging and Neural Network
US20220138549A1 (en) * 2019-07-18 2022-05-05 Huawei Technologies Co., Ltd. Photonic Signal Processing
WO2022031982A1 (en) * 2020-08-05 2022-02-10 The George Washington University Massively parallel amplitude-only optical processing system and methods for machine learning
CN113239615A (en) * 2021-04-21 2021-08-10 西北工业大学 Flight visual display system design method based on machine learning
CN114418287A (en) * 2021-12-10 2022-04-29 国网浙江省电力有限公司平阳县供电公司 Micro-grid frequency safety evaluation method containing electricity-gas-thermal coupling system
CN114334041A (en) * 2021-12-31 2022-04-12 厦门大学 Transformer-based electromagnetic metamaterial complex spectrum high-precision prediction method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
吴明长等: "Hα与白光望远镜的光机热一体化分析", 《红外与激光工程》 *
肖阳等: "光机系统的一体化仿真分析", 《光学学报》 *

Also Published As

Publication number Publication date
CN115392138B (en) 2023-01-24

Similar Documents

Publication Publication Date Title
US8346693B2 (en) Method for hammerstein modeling of steam generator plant
JP5885831B2 (en) Method of acquiring data from a non-linear dynamic real system during a test run implemented on a machine
Hinnen et al. A Data-Driven ${\cal H} _ {2} $-Optimal Control Approach for Adaptive Optics
Salimifard et al. Identification of nonlinear MIMO block-oriented systems with moving average noises using gradient based and least squares based iterative algorithms
Poyneer et al. Kalman filtering to suppress spurious signals in adaptive optics control
Yucesan et al. Adjusting a torsional vibration damper model with physics-informed neural networks
CN116306248B (en) Lithium battery temperature field prediction method based on space-time nonlinear error compensation model
Sun et al. Identification and adaptive control of a high-contrast focal plane wavefront correction system
Haber et al. General approach to precise deformable mirror control
Agathos et al. Parametric reduced order models for output-only vibration-based crack detection in shell structures
Gosea et al. Data-driven modeling and control of large-scale dynamical systems in the Loewner framework: Methodology and applications
Haffert et al. Data-driven subspace predictive control of adaptive optics for high-contrast imaging
Waseem et al. Data-driven reduced homogenization for transient diffusion problems with emergent history effects
Morishima et al. Thermal displacement error compensation in temperature domain
CN115392138B (en) Optical-mechanical-thermal coupling analysis model based on machine learning
Schindlbeck et al. Predictor-corrector framework for the sequential assembly of optical systems based on wavefront sensing
Piscaer et al. Predictive wavefront sensorless adaptive optics for time-varying aberrations
Wang et al. Identification of LPV model for superheated steam temperature system using A-QPSO
Haber Subspace identification of temperature dynamics
Zobeiry et al. Theory-guided machine learning composites processing modelling for manufacturability assessment in preliminary design
Yucesan et al. Estimating parameters and discrepancy of computer models with graphs and neural networks
Verhaegen et al. Data-Driven Identification of Networks of Dynamic Systems
Newberry et al. Bi-fidelity reduced polynomial chaos expansion for uncertainty quantification
Haber et al. Dual-update data-driven control of deformable mirrors using Walsh basis functions
Li et al. A comprehensive framework for model validation and reliability assessment of container crane structures

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant