Method for deciding polishing process parameters based on image quality indexes of neural network

Abstract

The invention relates to the technical field of optical processing, in particular to a method for deciding polishing process parameters based on image quality indexes of a neural network. Comprising the steps of determining structure parameters of an optical system according to specific requirements; marking an optical element which needs to be subjected to process decision in the optical system, and recording historical processing data; an optical system is built, the marked optical element is placed in the optical system, actual image quality indexes are obtained through debugging, and the deterioration degree compared with ideal system image quality is collected; repeating the steps to obtain a data set; taking the image quality deterioration degree data set as input, taking the process parameters of the elements as output, and inputting the image quality deterioration degree data set and the process parameters into a modular convolutional neural network for training to obtain a prediction network model; and when an optical system element manufacturing task arrives, inputting the system type and the image quality deterioration degree allowance into the prediction network model to obtain polishing process parameters meeting the image quality deterioration degree allowance. The method has the advantages of quick decision, less calculation time and high understandability.

Description

Technical Field

[0001] The present invention relates to the technical field of optical processing, and particularly relates to a method for determining polishing process parameters based on image quality indicators of a neural network. Background Art

[0002] For a manufacturing process, manufacturing errors will inevitably remain on the surface of the component. An optical component with manufacturing errors installed in the system affects the imaging quality of the optical system. The paper "Analysis and Prediction of Image Quality Degradation Caused by Diffraction of Infrared Optical System Turning Marks" published in the journal "Photonics" expounds on the influence of manufacturing errors generated by single-point diamond turning on diffraction. The period of the manufacturing errors generated by this deterministic process is in the micron order of magnitude, and obvious diffraction effects can be generated for visible light, which can be explained by diffraction theory. Thus, by reducing the diffraction intensity, it can be deduced which characteristic indicators of manufacturing errors are optional, and then the process parameters can be selected.

[0003] The paper "Surface variation analysis of freeform optical systems over surface frequency bands for prescribed wavefront errors" published in the journal "Optics & Laser Technology" expounds on the influence of optical components with toroidal manufacturing errors, and deduces the range in which the manufacturing error level should be in a geometric optics manner.

[0004] The characteristic period of the manufacturing errors generated by the magnetorheological process is in the millimeter order of magnitude, and the characteristic period of the manufacturing errors generated by the small grinding head process is in the centimeter order of magnitude. Moreover, most of the manufacturing errors generated by this process are grating-like errors, and their period scale is much larger than the wavelength. According to the diffraction theory, obvious diffraction effects cannot be generated. In the absence of a model to accurately describe this process, the neural network, such a blind-box fitting tool, is relatively suitable. Therefore, a method for determining polishing process parameters based on image quality indicators of a neural network should be developed. Summary of the Invention

[0005] The present invention provides a method for determining polishing process parameters based on image quality indicators of a neural network to solve the above problems.

[0006] The purpose of the present invention is to provide a method for determining polishing process parameters based on image quality indicators of a neural network, which specifically includes the following steps: S1. Determine the structural parameters of the optical system according to specific requirements; S2. Mark the optical elements in the optical system that require process decision-making, and record the historical processing data of the optical elements; S3. Build the optical system, place the marked optical elements in step S2 in the optical system, debug to obtain the actual image quality index, and collect the data on the deterioration degree compared with the image quality of the ideal system; S4. Repeat steps S2 - S3 to obtain the process parameter dataset and the image quality deterioration degree dataset of the optical elements; S5. Use the image quality deterioration degree dataset as the input dataset and the process parameters of the elements as the output dataset, and input them into the modular convolutional neural network for training to obtain the prediction network model of the optical element polishing process parameters; S6. According to the requirements of the process decision-making task, obtain the tolerance of the image quality deterioration degree of the optical element, and use the maximum and minimum values of the tolerance of the image quality deterioration degree as the image quality deterioration degree data to input into the prediction network model of the optical element polishing process parameters to obtain the polishing process parameters that meet the tolerance of the image quality deterioration degree.

[0007] Preferably, the structural parameters of the optical system include the type of the optical system, and the relative positions, placement angles, element surface shapes, and substrate materials of all optical elements in the optical system.

[0008] Preferably, the historical processing data in step S2 includes process parameters, environmental parameters, and manufacturing error data corresponding to the process parameters; The process parameter data includes the processing gauge h, the removal function RF, the processing angle theta, and the processing dwell time dt.

[0009] Preferably, in step S4, repeat steps S2 - S3 at least N groups, where N ≥ 50.

[0010] Preferably, use the image quality deterioration degree dataset as the input dataset and the process parameter dataset as the output dataset; The modular convolutional neural network includes a backbone convolutional layer and a dynamic weight layer; the backbone convolutional layer is used to extract the features of the image quality deterioration degree data; the dynamic weight layer dynamically loads different weights according to the optical system category and is used to map the features to the process parameter space.

[0011] Preferably, the training optimization process in step S5 is as follows: S501. Perform weight retrieval according to the optical system feature vector, and the weight index k is determined by the following formula:

[0012] where x is the optical system feature vector; The function is used to find the input parameter value that makes the function reach the maximum value, or return the index of the maximum value element in the dataset; S502. Input the image quality deterioration degree data into the modular convolutional neural network; the backbone convolutional layer extracts the feature F of the image quality deterioration data:

[0013] Among them, F is the feature of the image quality deterioration degree data extracted by the backbone convolutional layer and is also the input of the dynamic weight layer; Conv represents the convolutional function; Flatten the feature F of the image quality deterioration data output by the backbone convolutional layer into a vector f, and map it to the process parameter space through the dynamically loaded weight PT k to calculate the process parameter prediction value; the calculation formula is as follows:

[0014] Among them, is the ReLU activation function, the bias term, is the process parameter prediction value; S503. Use the difference between the predicted process parameter and the actual process parameter as the loss function, which is expressed as follows:

[0015] In the formula, h i represents the actual processing gauge, represents the predicted processing trajectory, represents the actual processing angle, represents the predicted processing angle,... represents other process parameters.

[0016] Preferably, in the training and optimization process of step S5, the adam algorithm is used as the optimization algorithm to fit the relationship between the input dataset and the output dataset under different optical systems, and the weight files trained for different optical systems are numbered and recorded as PT k , ; where k is the number of the optical system and W is the number of types of optical systems.

[0017] Compared with the prior art, the present invention can achieve the following beneficial effects: The present invention provides a method for determining polishing process parameters based on image quality indicators of a neural network, that is, a method for determining optical processing process parameters associated with the imaging indicators of an optical system. Although the amount of data set collection in the early stage of the method of the present invention is large, once the neural network calculation model is built, the acquisition of decision-making process parameters is very fast and objective. The method for determining deterministic polishing process parameters based on image quality indicators of a neural network proposed by the present invention has obvious advantages in terms of calculation time consumption and understandability. Traditional debugging determines process parameters by continuously adjusting parameters through experiments. Compared with the traditional debugging method, the process development efficiency of the present invention can be increased by 5-6 times, and it is particularly suitable for precision machining scenarios of complex optical elements such as aspherical surfaces. Description of the Drawings

[0018] Figure 1 is a flowchart of a method for determining polishing process parameters based on image quality indicators of a neural network according to an embodiment of the present invention.

[0019] Figure 2 is a removal function diagram according to an embodiment of the present invention.

[0020] Figure 3 is a simulated tool mark error according to an embodiment of the present invention.

[0021] Figure 4 is a flowchart for selecting modular neural network weights according to an embodiment of the present invention.

[0022] Figure 5 is a flowchart for running a modular neural network according to an embodiment of the present invention.

[0023] Figure 6 is an image quality evaluation result of an optical system according to an embodiment of the present invention. Detailed Embodiment

[0024] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, the same modules are denoted by the same reference numerals. In the case of the same reference numerals, their names and functions are also the same. Therefore, their detailed descriptions will not be repeated.

[0025] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation to the present invention.

[0026] The present invention provides a method for determining polishing process parameters based on image quality indicators of a neural network (the flowchart is as Figure 1 ) aiming to efficiently and accurately determine the polishing process parameters of optical elements to meet the strict requirements of the optical system for image quality; specifically, it includes the following steps: S1. Obtaining basic system information: According to specific requirements, determine the structure of the optical system and the structural parameters of the optical system for which the network is to be established. In a specific embodiment, the structural parameters of the optical system include the type of the optical system (such as off-axis three-mirror, Cassegrain, etc.), as well as the relative positions, placement angles, surface profiles, substrate materials, etc. of all optical elements within the optical system; the above information will provide an important reference basis for subsequent process parameter decisions, ensuring that the constructed neural network can accurately reflect the actual characteristics of the optical system.

[0027] S2. Mark the optical elements that require process decisions in the optical system and record the historical processing data of the optical elements. The historical processing data includes process parameters and manufacturing errors corresponding to the process parameters. The process parameter data includes processing gauge h, removal function RF, processing angle theta, processing dwell time dt, etc.; the removal function is as Figure 2 shown; for the convenience of display, a manufacturing error of a magnetorheological polishing feature is shown in Figure 3 .

[0028] S3. Set up the optical system, place the optical elements that require decisions in the optical system, debug to obtain the actual image quality indicators, and collect the data on the deterioration degree compared with the image quality of the ideal system. The deterioration degree of image quality is a key indicator to measure the gap in imaging quality between the actual optical system and the ideal optical system. By collecting this data, it can provide a clear input target for the neural network, enabling it to learn how to adjust process parameters according to image quality requirements.

[0029] S4. Repeat steps S2 - S3 at least N groups (N≥50) to obtain the process parameter data set and the image quality deterioration degree data set of the optical elements; use the image quality deterioration degree data set as the input data set and the process parameter data set as the output data set. Through the iterative verification of N≥50 groups of experimental data, ensure the generalization ability of the neural network in the high-dimensional space of process parameters. S5. Input the dataset of image quality deterioration degree into a Modular Neural Network (MNN) for training to obtain an optical element polishing process parameter prediction network model; specifically, the Modular Neural Network is a modular network that can select different weight data for network prediction according to different optical system structure parameters; specifically, in the training part, the input-output dataset obtained under different optical system structures is trained, and then the network weights obtained from the training are saved. In the next prediction process, without adjusting the network structure, only by inputting the feature vector of the optical system structure, modular applications for different optical systems can be realized; The Modular Neural Network includes a backbone convolutional layer and a dynamic weight layer; the backbone convolutional layer, also known as the shared convolutional layer, uses the same convolutional layer for different categories of optical systems to extract the feature F of the image quality deterioration degree data J:

[0030] Among them, F is the feature of the image quality deterioration degree data extracted by the backbone convolutional layer and is also the input of the dynamic weight layer; Conv represents the convolutional function; The dynamic weight layer dynamically loads different weights according to the optical system category and is used to map the features to the process parameter space; specifically, the output F of the backbone convolutional layer is flattened into a vector f, and through the dynamically loaded weight PT k it is mapped to the process parameter space:

[0031] Among them, is the activation function (ReLU function), the bias term, is the process parameter prediction value.

[0032] The training optimization process is as follows: S501. Perform weight retrieval according to the optical system feature vector, and the weight index k is determined by the following formula:

[0033] Among them, x is the optical system feature vector, is a function commonly used in mathematics and computer science, and its core function is to find the input parameter value that makes a certain function reach the maximum value, or to return the index of the maximum value element in the dataset; S502. Input the image quality deterioration degree data into the Modular Neural Network; the backbone convolutional layer extracts the feature F of the image quality deterioration data:

[0034] Among them, F is the data feature of the image quality deterioration degree extracted by the backbone convolutional layer, and is also the input of the subsequent dynamic weight layer. Conv represents the convolutional function; Flatten the feature F of the image quality deterioration data output by the backbone convolutional layer into a vector f, and map it to the process parameter space through the dynamically loaded weight PT k to calculate the predicted value of the process parameter; the calculation formula is as follows:

[0035] Among them, is the activation function (ReLU function), bias term, is the predicted value of the process parameter; S503. Use the difference between the predicted process parameter and the actual process parameter as the loss function, which is expressed as follows:

[0036] In the formula, h i represents the actual processed gauge, represents the predicted processing trajectory, represents the actual processing angle, represents the predicted processing angle,... represents other process parameters, which will not be elaborated one by one; Use the adam algorithm as the optimization algorithm to fit the relationship between the input data set and the output data set under different optical systems. Number the trained weight files of different optical systems and denote them as PT k , ; where k is the number of the optical system, and W is the number of types of optical systems; Adopt one-hot encoding to convert the optical system category into a computable vector form; for different optical systems, in a specific embodiment, taking the Cassegrain and off-axis three-mirror systems as examples, use one-hot encoding to process them to obtain the optical system category features in vector form. For example, [1,0] represents the off-axis three-mirror system, and [0,1] represents the Cassegrain system. Correlate the encoded optical system feature vector with the corresponding modular neural network weight file to ensure that the network can distinguish the process characteristics of different systems. The flowchart of modular neural network weight selection is shown in Figure 4 as shown; the network operation process is shown in Figure 5 .

[0037] Considering that the influence of the element placement angles of different optical systems on the performance indicators of the optical system varies, and at the same time the types of common optical systems are limited, therefore, in this application, different optical systems are classified, modular neural networks are trained for different types of optical systems, and the network weights are dynamically selected or adjusted according to different optical systems to ensure the prediction accuracy of the neural network.

[0038] In this application, a deep neural network (DNN) can be used as a module in the modular neural network; it should be noted that it can also be replaced with other types of neural networks to achieve the same training effect.

[0039] S6. According to the requirements of the process decision task, obtain the tolerance of the image quality deterioration degree of the optical element, and use the maximum and minimum values of the tolerance of the image quality deterioration degree as the image quality deterioration degree data to input into the optical element polishing process parameter prediction network model to obtain the polishing process parameters that meet the tolerance of the image quality deterioration degree; The polishing process is not limited to a single deterministic process, such as deterministic polishing processes like magnetorheological polishing, small grinding head polishing, ion beam polishing, etc.; The process parameters are not limited to one type. For example, the processing gauge h, the removal function RF, the processing angle theta, the processing dwell time dt, etc., that is, P = (RF, h, theta, dt); The geometric optical evaluation indicators (image quality evaluation indicators) are not limited to one type, such as the root mean square value of the spot diagram, the modulation value at the cut-off frequency of the amplitude modulation function, etc. For the sake of illustration, for one of the optical elements of a certain system, Figure 2 Three process parameters of the removal function are used for simulation manufacturing calculation to obtain the corresponding virtual elements and input them into the optical design software to simulate the image quality of the optical system corresponding to each virtual element. The evaluation result of the clarity of the image quality of one system is shown in Figure 6 .

[0040] The advantage of the present invention is that through systematic data-driven modeling, it solves the technical problem of traditional process parameter dependence on empirical tuning, and adopts a dynamic weight switching mechanism triggered by the optical system category to construct a modular neural network. Although the amount of data set collection in the early stage of the method of the present invention is large, once the neural network calculation model is built, the acquisition of the decision-making process parameters is very fast and objective. The method for determining the deterministic polishing process parameters based on the image quality index of the neural network proposed by the present invention has obvious advantages in terms of calculation time consumption and comprehensibility. Compared with the traditional debugging method, the process development efficiency can be increased by 5-6 times, and it is particularly suitable for the precision machining scenarios of complex optical elements such as aspheric surfaces.

[0041] It should be understood that the various forms of processes shown above can be used, with steps reordered, added or deleted. For example, the steps described in the disclosure of the present invention can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and no limitation is imposed herein.

[0042] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub - combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for determining polishing process parameters based on image quality indicators of a neural network, characterized in that: The specific steps include: S1. Determine the structural parameters of the optical system according to specific needs; S2. marking the optical elements that require process decisions in the optical system and recording the historical processing data of the optical elements; S3. Build the optical system, place the optical element marked in step S2 in the optical system, debug to obtain the actual image quality index, and collect the deterioration data compared with the ideal system image quality; S4. Repeat steps S2 to S3 to obtain a process parameter data set and an image quality deterioration degree data set of the optical element; S5. Using the image quality deterioration degree dataset as the input dataset and the process parameters of the component as the output dataset, inputting them into the modular convolutional neural network for training, and obtaining a network model for predicting the polishing process parameters of the optical component; S6. According to the requirements of the process decision task, the image quality deterioration tolerance of the optical element is obtained, and the maximum value of the image quality deterioration tolerance is input as the image quality deterioration data into the optical element polishing process parameter prediction network model to obtain the polishing process parameters that meet the image quality deterioration tolerance.

2. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 1, characterized in that: The optical system structural parameters include the type of the optical system, and the relative positions, placement angles, component surface shapes and substrate materials of all optical components in the optical system.

3. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 1, characterized in that: The historical processing data in step S2 includes process parameters, environmental parameters, and manufacturing error data corresponding to the process parameters; The process parameter data include machining track gauge h, removal function RF, machining angle theta, and machining dwell time dt.

4. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 1, characterized in that: In the step S4, steps S2 to S3 are repeated for at least N groups, where N≥50.

5. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 1, characterized in that: The image quality deterioration degree dataset is used as the input dataset, and the process parameter dataset is used as the output dataset; The modular convolutional neural network includes a backbone convolution layer and a dynamic weight layer; the backbone convolution layer is used to extract features of image quality deterioration degree data; The dynamic weight layer dynamically loads different weights according to the optical system category, so as to map the features to the process parameter space.

6. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 5, characterized in that: The training optimization process in step S5 is as follows: S501. Perform weighted search based on the optical system feature vector, and the weight index k is determined by the following formula: Where x is the characteristic vector of the optical system; The function is used to find the input parameter value that makes the function reach the maximum value, or return the index of the maximum value element in the data set; S502. Input the image quality deterioration degree data into the modular convolutional neural network; the backbone convolution layer extracts the feature F of the image quality deterioration data: Among them, F is the image quality deterioration data feature extracted by the backbone convolution layer, and is also the input of the dynamic weight layer; Conv represents the convolution function; The feature F of the image quality deterioration data output by the backbone convolutional layer is flattened into a vector f, and the dynamically loaded weights PT k Map to the process parameter space and calculate the predicted value of the process parameter; the calculation formula is as follows: in, is the ReLU activation function, Bias term, is the predicted value of process parameters; S503. Use the difference between the predicted process parameters and the actual process parameters as the loss function, expressed as follows: In the formula, h i Indicates the actual processing gauge, represents the predicted processing trajectory, Indicates the actual processing angle, represents the predicted processing angle, … represents other process parameters.

7. The method for determining polishing process parameters based on image quality indicators based on a neural network according to claim 6, characterized in that: In the training optimization process of step S5, the adam algorithm is used as the optimization algorithm to fit the relationship between the input data set and the output data set under different optical systems, and the weight files trained by different optical systems are numbered and recorded as PT k , ; where k is the number of the optical system and W is the number of types of optical systems.

Images