Wafer polishing surface type control method, wafer polishing surface type model training method and related device

Abstract

The invention discloses a wafer polishing surface type control method, a model training method thereof and a related device, a wafer polishing surface type control model is a generative adversarial network model comprising a generator and a discriminator, and the training method of the wafer polishing surface type control model comprises the steps of obtaining historical processing data; wherein the historical processing data at least comprise actual process control parameters of different stages, before-polishing profile curve data and expected profile curve data; before-polishing profile curve data and expected profile curve data are input into a generator, and predicted process control parameters are output; inputting the predicted process control parameters and the actual process control parameters into a discriminator, and respectively outputting true and false discrimination results of the predicted process control parameters and the actual process control parameters; and iteratively updating the parameters of the wafer polishing surface type control model based on the true and false discrimination result. In this way, the control precision and the production efficiency of the polishing surface type of the wafer can be improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor processing technology, and in particular to a wafer polishing surface control method and a model training method thereof, and related devices. Background Art

[0002] Wafer polishing is a key process in the semiconductor manufacturing process. It aims to remove tiny impurities and unevenness on the wafer surface through mechanical polishing to ensure its flatness and smoothness. The surface shape of the wafer is an important indicator reflecting the surface morphology of the wafer. Among them, GBIR (TTV) refers to the total thickness deviation of the silicon wafer surface, and SFQR refers to the local flatness of the silicon wafer, which can be used to judge the edge collapse of the silicon wafer after polishing.

[0003] However, the polishing process is a time-varying and highly nonlinear process. The product morphology is affected by many factors such as working conditions and types of raw and auxiliary materials. It is difficult for the existing solutions to describe the relationship between various process parameters and surface changes through mechanism models. In addition, during the wafer production process, the process parameters need to be manually adjusted according to the predicted subsequent shape. It is difficult to cope with changes in different materials and different production batches of products, and has poor adaptability to environmental changes, which seriously affects the control accuracy and production efficiency of the wafer polishing surface. Summary of the invention

[0004] The present application provides a wafer polishing surface shape control method and a model training method thereof, as well as related devices, to improve the control accuracy and production efficiency of the wafer polishing surface shape.

[0005] In order to solve the above technical problems, the present application provides a training method for a wafer polishing surface shape control model, wherein the wafer polishing surface shape control model is a generative adversarial network model including a generator and a discriminator, and the method includes: obtaining historical processing data; wherein the historical processing data at least includes actual process control parameters, front polishing shape curve data and expected surface shape curve data at different stages; inputting the front polishing shape curve data and the expected surface shape curve data into the generator, and outputting predicted process control parameters; inputting the predicted process control parameters and the actual process control parameters into the discriminator, and outputting true or false discrimination results of the predicted process control parameters and the actual process control parameters respectively; and iteratively updating the parameters of the wafer polishing surface shape control model based on the true or false discrimination results.

[0006] Wherein, the generator includes a connected convolutional neural network and a fully connected neural network, and the historical processing data also includes auxiliary material life data at different stages; the step of inputting the pre-polishing shape curve data and the expected surface shape curve data into the generator and outputting predicted process control parameters includes: inputting the pre-polishing shape curve data and the expected surface shape curve data into the convolutional neural network to extract polishing feature information; inputting the polishing feature information and the auxiliary material life data into the fully connected neural network to map the polishing feature information and the auxiliary material life data to the predicted process control parameters.

[0007] The convolutional neural network includes a plurality of convolutional layers and pooling layers connected alternately; the convolutional layer is used to extract features of the inputted pre-polishing profile curve data and the desired profile curve data; the pooling layer is used to perform a pooling operation on the features extracted by the convolutional layer; And / or, the fully connected neural network includes multiple first fully connected layers connected in sequence; the first fully connected layer is used to perform nonlinear mapping on the polishing feature information extracted by the convolutional neural network, and finally output the predicted process control parameters.

[0008] Among them, the discriminator includes a deep neural network, and the deep neural network includes multiple second fully connected layers connected in sequence; each neuron of each layer of the second fully connected layer is connected to all neurons of the second fully connected layer of the previous layer, and the number of neurons in adjacent second fully connected layers gradually decreases; the second fully connected layer is used to perform nonlinear mapping on the input predicted process control parameters and the actual process control parameters, and finally output the true or false discrimination result.

[0009] Among them, the step of iteratively updating the parameters of the wafer polishing surface control model based on the true or false discrimination results includes: determining the loss function of the generator according to the true or false discrimination results of the discriminator on the predicted process control parameters; and determining the loss function of the discriminator according to the true or false discrimination results of the discriminator on the predicted process control parameters and the actual process control parameters; and alternately optimizing the network parameters of the generator and the discriminator according to the loss function of the generator and the loss function of the discriminator.

[0010] Among them, the step of alternately optimizing the network parameters of the generator and the discriminator according to the loss function of the generator and the loss function of the discriminator includes: in the Nth round of training, fixing the network parameters of the generator, minimizing the loss function of the discriminator, and updating the network parameters of the discriminator by gradient descent; in the N+1th round of training, fixing the network parameters of the discriminator, minimizing the loss function of the generator, and updating the network parameters of the generator by gradient descent.

[0011] In order to solve the above technical problems, the present application also provides a wafer polishing surface shape control method, which is applied to a process control system of a wafer polishing equipment, wherein the process control system is integrated with a generator of a wafer polishing surface shape control model, and the wafer polishing surface shape control model is trained by any of the above-mentioned wafer polishing surface shape control model training methods; the wafer polishing surface shape control method comprises: when polishing a current batch of wafers, obtaining processing data corresponding to the current batch of wafers; wherein the processing data corresponding to the current batch of wafers at least includes front polishing shape curve data and expected shape curve data of the current batch of wafers; the front polishing shape curve data and expected shape curve data of the current batch of wafers are input into the generator, and the predicted process control parameters of the current batch of wafers are output; based on the predicted process control parameters of the current batch of wafers, an optimized control instruction is generated and sent to the process control system; the process control system adjusts the processing parameters of the wafer polishing equipment according to the optimized control instruction to polish the current batch of wafers.

[0012] Wherein, the generator includes a connected convolutional neural network and a fully connected neural network, and the processing data corresponding to the current batch of wafers also includes the auxiliary material life data corresponding to the current batch of wafers; the step of inputting the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers into the generator and outputting the predicted process control parameters of the current batch of wafers includes: inputting the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers into the convolutional neural network to extract polishing feature information; inputting the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers into the fully connected neural network to map the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers to the predicted process control parameters of the current batch of wafers.

[0013] Among them, the wafer polishing surface shape control method also includes: before polishing the next batch of wafers, obtaining the historical processing data corresponding to the previous batch of wafers; updating the parameters of the wafer polishing surface shape control model according to the historical processing data corresponding to the previous batch of wafers; the process control system calls the generator with updated parameters to control the polishing process of the next batch of wafers.

[0014] In order to solve the above technical problems, the present application also provides a training device for a wafer polishing surface shape control model, wherein the wafer polishing surface shape control model is a generative adversarial network model including a generator and a discriminator, and the training device includes: a first acquisition module, wherein the first acquisition module is used to acquire historical processing data; wherein the historical processing data at least includes actual process control parameters, front polishing shape curve data and expected surface shape curve data at different stages; a first processing module, wherein the first processing module is used to input the front polishing shape curve data and the expected surface shape curve data into the generator, and output predicted process control parameters; input the predicted process control parameters and the actual process control parameters into the discriminator, and output true or false discrimination results of the predicted process control parameters and the actual process control parameters respectively; and a first update module, wherein the first update module is used to iteratively update the parameters of the wafer polishing surface shape control model based on the true or false discrimination results.

[0015] In order to solve the above technical problems, the present application also provides a wafer polishing surface type control device, the wafer polishing surface type control device is used to control the wafer polishing equipment to polish the wafer, the wafer polishing surface type control device is integrated with a generator of a wafer polishing surface type control model, and the wafer polishing surface type control model is obtained by training the training method of the wafer polishing surface type control model described in any of the above items; the wafer polishing surface type control device includes: a second acquisition module, the second acquisition module is used to obtain the processing data corresponding to the current batch of wafers when polishing the current batch of wafers; wherein, the current batch The processing data corresponding to the secondary wafer at least includes the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers; a second processing module, the second processing module is used to input the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers into the generator, and output the predicted process control parameters of the current batch of wafers; based on the predicted process control parameters of the current batch of wafers, generate optimized control instructions; a second update module, the second update module is used to adjust the processing parameters of the wafer polishing equipment according to the optimized control instructions, so as to perform polishing processing on the current batch of wafers.

[0016] In order to solve the above-mentioned technical problems, the present application also provides an electronic device, which includes: a memory and a processor coupled to each other, and the processor is used to execute program instructions stored in the memory to implement the training method of the wafer polishing surface shape control model described in any of the above items, and / or the wafer polishing surface shape control method.

[0017] To solve the above technical problems, the present application also provides a computer-readable storage medium, which stores program instructions, and the program instructions can be executed to implement the training method of the wafer polishing surface type control model as described in any of the above items, and / or the wafer polishing surface type control method.

[0018] The beneficial effects of the present application are as follows: different from the prior art, the wafer polishing surface shape control model of the present application is a generative adversarial network model including a generator and a discriminator. In the process of training the wafer polishing surface shape control model, historical processing data is first obtained, wherein the historical processing data at least includes actual process control parameters, front polishing shape curve data and expected surface shape curve data at different stages; then the front polishing shape curve data and the expected surface shape curve data are input into the generator, and the predicted process control parameters are output; then the predicted process control parameters and the actual process control parameters are input into the discriminator, and the true or false discrimination results of the predicted process control parameters and the actual process control parameters are respectively output; therefore, the parameters of the wafer polishing surface shape control model can be iteratively updated based on the true or false discrimination results. A generative adversarial network model including a generator and a discriminator is constructed as a wafer polishing surface control model, and the generative adversarial network model is trained using historical processing data. Since the historical processing data at least includes actual process control parameters at different stages, front-polishing profile curve data, and expected surface profile curve data, the generator will generate predicted process control parameters based on the front-polishing profile curve data and the expected surface profile curve data at different stages, and the discriminator is used to distinguish the gap between the generated predicted process control parameters and the actual process control parameters at the corresponding stage. Through adversarial training between the generator and the discriminator, the discriminator can maximize its classification accuracy of the actual process control parameters and the generated predicted process control parameters, while the generator can minimize the gap between its generated predicted process control parameters and the actual The difference between the actual process control parameters; therefore, it can be understood that after the wafer polishing surface shape control model is trained, when polishing any batch of wafers, the pre-polishing surface shape curve data of the batch of wafers, the expected surface shape curve data and the generator of the wafer polishing surface shape control model can be used to generate the process control parameters corresponding to the batch of wafers, and after processing using the process control parameters, the batch of wafers can obtain processing results close to their expected surface shape curve data. Therefore, in the wafer polishing process, the trained wafer polishing surface shape control model of the present application can be used to accurately adjust its process control parameters according to the surface shape data of any batch of wafers, ensuring that the final surface shape of the wafer meets the expected target, thereby improving the accuracy and production efficiency of the polishing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the first embodiment of the training method of the wafer polishing surface control model provided by the present application; Figure 2 It is a flow chart of the second embodiment of the training method of the wafer polishing surface control model provided by the present application; Figure 3 It is a structural schematic diagram of a wafer polishing surface shape control model in an application scenario of the present application; Figure 4 yes Figure 2 A schematic diagram of a flow chart of an embodiment of step S25; Figure 5 It is a schematic diagram of a training method for a wafer polishing surface control model in an application scenario of the present application; Figure 6 It is a flow chart of the first embodiment of the wafer polishing surface shape control method provided by the present application; Figure 7 It is a flow chart of the second embodiment of the wafer polishing surface shape control method provided by the present application; Figure 8 It is a schematic diagram of a method for controlling a wafer polishing surface shape in an application scenario of the present application; Fig. 9 It is a structural schematic diagram of an embodiment of a training device for a wafer polishing surface control model provided by the present application; Fig.10 It is a structural schematic diagram of an embodiment of a wafer polishing surface type control device provided by the present application; Fig.11 It is a structural schematic diagram of an embodiment of an electronic device provided by the present application; Fig.12 It is a structural schematic diagram of an embodiment of a computer-readable storage medium provided by the present application. DETAILED DESCRIPTION

[0020] The scheme of the embodiment of the present application is described in detail below in conjunction with the drawings of the specification.

[0021] In the following description, for the purpose of explanation rather than limitation, specific details such as specific system structures, interfaces, and technologies are provided to facilitate a thorough understanding of the present application.

[0022] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0023] See also Figure 1 , Figure 1 1 is a flow chart of the first embodiment of the training method of the wafer polishing surface control model provided by the present application. The wafer polishing surface control model in the embodiment of the present application is a generative adversarial network model including a generator and a discriminator. The training method of the wafer polishing surface control model in the present embodiment includes the following steps: Step S11: Acquire historical processing data; wherein the historical processing data at least includes actual process control parameters at different stages, surface profile curve data before polishing, and expected surface profile curve data.

[0024] The wafer in the embodiment of the present application may include workpieces such as silicon wafers, sapphire wafers, and silicon carbide wafers. The wafer polishing surface control model can be used as a training sample by collecting historical processing data of wafer polishing equipment. The historical processing data includes process control parameters and wafer surface curve data at different stages of the wafer processing process, wherein the process control parameters may include the rotation speed and feed speed of the grinding and polishing disk, the inclination angle of the grinding and polishing disk, the temperature parameters during the grinding process, the sensor thickness measurement parameters, the pressure parameters applied to the wafer, and the polishing time, etc. The process control parameters will affect the surface shape of the wafer after polishing. The wafer surface curve data includes the surface curve data and the expected surface curve data of the wafer before polishing; wherein the surface curve data of the wafer before polishing refers to the surface shape of the wafer before polishing, which is expressed as a two-dimensional curve image, and the main information contained is the change in the height or shape of the wafer surface, which can reflect the surface characteristics of the wafer before polishing; the expected surface curve data of the wafer refers to the target surface shape of the wafer, that is, the surface shape expected to be achieved after the wafer is ideally polished, which is expressed as a two-dimensional curve image.

[0025] Step S12: inputting the pre-polishing profile curve data and the desired profile curve data into the generator, and outputting predicted process control parameters.

[0026] It can be understood that after the wafer's front polishing shape curve data and the corresponding expected surface shape curve data are input into the generator, the generator can simultaneously process the front polishing shape curve data and the expected surface shape curve data, and generate predicted process control parameters based on these inputs, that is, the generator infers that after polishing the wafer with the front polishing shape curve data using the predicted process control parameters it generates, the polished wafer can have the corresponding expected surface shape curve data.

[0027] Step S13: inputting the predicted process control parameter and the actual process control parameter into the discriminator, and outputting true or false discrimination results of the predicted process control parameter and the actual process control parameter respectively.

[0028] It can be understood that the discriminator needs to learn the features extracted from the real data and the generated data of the generator, and finally give a probability of distinguishing between the real data and the generated data; therefore, after the predicted process control parameters and the actual process control parameters are input into the discriminator, the predicted process control parameters are used as the generated data of the generator, and the discriminator can give a true or false judgment result of whether the predicted process control parameters are the real data. Similarly, the actual process control parameters are used as the real data, and the discriminator can also give a true or false judgment result of whether the actual process control parameters are the real data.

[0029] Step S14: iteratively updating the parameters of the wafer polishing surface control model based on the true or false discrimination result.

[0030] It is understandable that the predicted process control parameters generated by the generator of the generative adversarial network should confuse the discriminator as much as possible, while the discriminator should distinguish the predicted process control parameters generated by the generator from the actual process control parameters as much as possible. When conducting adversarial optimization training, it is necessary to continuously update and optimize the parameters of the wafer polishing surface control model based on the true or false judgment results of the discriminator on whether the predicted process control parameters and the actual process control parameters are real data, so as to minimize the loss between the predicted value of the model and the true output value. The smaller the loss value between the predicted value and the true output value, the closer the model prediction is to the true value.

[0031] The above scheme builds a generative adversarial network model including a generator and a discriminator as a wafer polishing surface shape control model, and uses historical processing data to train the generative adversarial network model. Since the historical processing data at least includes actual process control parameters, front-polishing shape curve data and expected surface shape curve data at different stages, the generator will generate predicted process control parameters according to the front-polishing shape curve data and the expected surface shape curve data at different stages, and the discriminator is used to distinguish the difference between the generated predicted process control parameters and the actual process control parameters obtained in the corresponding stage. Through the adversarial training of the generator and the discriminator, the discriminator can maximize its classification accuracy of the actual process control parameters and the generated predicted process control parameters, while the generator can minimize the difference between the generated predicted process control parameters and the actual process control parameters. Therefore, it can be understood that after the training of the wafer polishing surface shape control model is completed, the batch of wafers can be used when polishing any batch of wafers. The generator of the front polishing shape curve data, the expected shape curve data and the wafer polishing surface shape control model of the batch of wafers generates the process control parameters corresponding to the batch of wafers, and after processing with the process control parameters, the batch of wafers can obtain processing results close to their expected shape curve data; the wafer polishing surface shape control model of the present application adopts an end-to-end design scheme, inputs the front polishing shape curve data and the expected shape curve data, and outputs the process control parameters that can obtain the expected shape curve data, that is, the control quantity is directly obtained from the current quantity and the expected quantity. Since the whole process is a unified model, the nonlinear relationship between the input and output can be better captured, thereby improving the accuracy of the model; therefore, in the wafer polishing process, the trained wafer polishing surface shape control model of the present application can be used to accurately adjust its process control parameters according to the shape data of any batch of wafers, ensuring that the final surface shape of the wafer meets the expected target, thereby improving the accuracy and production efficiency of the polishing process.

[0032] See also Figure 2 , Figure 2 1 is a flow chart of a second embodiment of a method for training a wafer polishing surface control model provided by the present application. The method for training a wafer polishing surface control model in this embodiment comprises the following steps: Step S21: Acquire historical processing data; wherein the historical processing data at least includes actual process control parameters at different stages, surface curve data before polishing, expected surface curve data and auxiliary material life data.

[0033] It is understandable that the life of auxiliary materials such as polishing liquid and polishing pad also has a certain influence on the wafer polishing surface shape. Therefore, the difference from the previous embodiment is that the historical processing data in this embodiment also includes auxiliary material life data at different stages. The auxiliary material life data is the life information about the polishing liquid, polishing pad and other auxiliary materials, which contains time series data, such as the usage and wear life of each auxiliary material, and the auxiliary material life data is expressed in the form of a one-dimensional vector.

[0034] In other embodiments, the historical processing data may also include randomly generated noise data having the same dimension as the surface curve. It is understandable that adding noise appropriately to the model can make the model more challenging and generalizable.

[0035] Step S22: inputting the surface profile curve data before polishing and the desired surface profile curve data into the convolutional neural network to extract polishing feature information.

[0036] Step S23: inputting the polishing characteristic information and the auxiliary material life data into the fully connected neural network to map the polishing characteristic information and the auxiliary material life data to the predicted process control parameters.

[0037] Please combine Figure 3 The generator of the wafer polishing surface shape control model in this embodiment includes a connected convolutional neural network and a fully connected neural network. The structure combining the convolutional neural network and the fully connected neural network can simultaneously process the front polishing surface shape curve data, the expected surface shape curve data and the auxiliary material life data, and generate the predicted process control parameters based on these inputs.

[0038] In one embodiment, the convolutional neural network includes multiple convolutional layers and pooling layers that are alternately connected; the convolutional layer is used to extract features of the input front polishing surface curve data and the expected surface curve data; the pooling layer is used to perform pooling operations on the features extracted by the convolutional layer.

[0039] Specifically, the convolution layer extracts features from the input image data (i.e., the surface curve data before polishing and the desired surface curve data in this application) through a sliding window. Each convolution layer performs a convolution operation on the input image data through a set of convolution kernels (filters) to extract local features. Suppose the input image data of the lth layer is , the convolution kernel is , the bias term is , then the convolution operation can be expressed as: .

[0040] Among them, * represents the convolution operation, ReLU is the activation function, and the form of the ReLU function is .

[0041] The pooling layer is used to reduce the dimension and ensure the stability of the extracted local features. In the embodiment of the present application, the pooling methods of Max Pooling and Average Pooling can be used. For example, the pooling operation can slide the window to take the maximum value or average value of each local area to reduce the size of the extracted feature map. The pooling operation can be expressed as: .

[0042] It can be understood that after multiple convolution and pooling operations on the surface curve data before polishing and the expected surface curve data, high-dimensional polishing feature information can be extracted, such as Figure 3 As shown, a flattening layer can be set as a transition layer between the convolutional neural network and the fully connected neural network. The flattening layer can flatten the high-dimensional features extracted by the convolutional neural network into one-dimensional data, that is, obtain one-dimensional polishing feature information, and then input the one-dimensional polishing feature information together with the auxiliary material life data represented as a one-dimensional vector form into the fully connected neural network. The fully connected neural network maps the polishing feature information extracted by the convolutional neural network and the directly input auxiliary material life data to the generated predicted process control parameters.

[0043] In one embodiment, the fully connected neural network includes a plurality of first fully connected layers connected in sequence; the first fully connected layer is used to perform nonlinear mapping on the polishing feature information extracted by the convolutional neural network, and finally output the predicted process control parameters.

[0044] The fully connected neural network performs nonlinear mapping on the high-dimensional polishing feature information extracted by the convolutional neural network through the first fully connected layer, and finally outputs the generated predicted process control parameters. Figure 3 As shown in Figure 1, each first fully connected layer of a fully connected neural network consists of multiple neurons, each of which is connected to all neurons in the previous layer. Assume that the input of the first fully connected layer of the kth layer is , the weight is , the bias is , then the output of the kth layer The calculation formula is: .

[0045] The final output layer of the fully connected neural network is also a first fully connected layer, which is used to generate predicted process control parameters. Assume that the final output is , represents the generated predicted process control parameters: .

[0046] in, is the weight of the output layer, is the output of the previous layer.

[0047] Step S24: input the predicted process control parameter and the actual process control parameter into the discriminator, and output true or false discrimination results of the predicted process control parameter and the actual process control parameter respectively.

[0048] Please combine Figure 3 In one embodiment, the discriminator includes a deep neural network, which includes multiple second fully connected layers connected in sequence; each neuron in each second fully connected layer is connected to all neurons in the previous second fully connected layer, and the number of neurons in adjacent second fully connected layers gradually decreases; the second fully connected layer is used to perform nonlinear mapping on the input predicted process control parameters and actual process control parameters, and finally output a true or false discrimination result.

[0049] It can be understood that the deep neural network is the core part of the discriminator. Each neuron in each second fully connected layer of the deep neural network is connected to all neurons in the previous layer. The discriminator gradually extracts high-level features of the input data through multiple second fully connected layers, and finally classifies true and false. Similar to the convolutional neural network in the generator above, let the input of the lth layer of the deep neural network be , the weight is , the bias term is , then the output of layer l is Calculated by the following formula: .

[0050] Among them, the ReLU activation function is used to introduce nonlinear transformation to enhance the expressive power of the network.

[0051] Therefore, the discriminator includes multiple second fully connected layers, and after each second fully connected layer, a ReLU activation function is applied to further capture the nonlinear relationship in the input data. The number of neurons in each second fully connected layer gradually decreases until the last layer outputs a single probability value. Assuming that the discriminator has L layers, the output of the last layer is , the value of the output layer is obtained through linear transformation: .

[0052] The output layer of the discriminator uses the sigmoid activation function to map the final output to the probability space, indicating the probability value of the input data being the real data. Assume that the final output is , then the output formula is: .

[0053] in, It is a sigmoid activation function, which compresses the output value to the interval [0,1], indicating the probability that the data is true.

[0054] Step S25: iteratively updating the parameters of the wafer polishing surface control model based on the true or false discrimination result.

[0055] Please combine Figure 4 , Figure 4 yes Figure 2 Schematic diagram of a flow chart of an embodiment of step S25. In one embodiment, the above step S25 specifically includes: Step S251: Determine the loss function of the generator based on the true or false discrimination result of the discriminator on the predicted process control parameters; and determine the loss function of the discriminator based on the true or false discrimination result of the discriminator on the predicted process control parameters and the actual process control parameters.

[0056] It is understandable that the core of the generative adversarial network model lies in its loss function, which usually includes two parts: the generator loss function and the discriminator loss function. Suppose the predicted process control parameter generated by the generator is G(z), the actual process control parameter is x, and the output of the discriminator is D(x) and D(G(z)), where D(x) represents the output probability of the discriminator's judgment on the actual process control parameter, and D(G(z)) represents the output probability of the discriminator's judgment on the predicted process control parameter generated by the generator. Among them, the generator loss function LG can be expressed as: .

[0057] The optimization goal of the discriminator is to maximize its classification accuracy of actual process control parameters and predicted process control parameters. To this end, the discriminator can use the cross entropy loss function (Binary Cross-Entropy Loss) to calculate the loss. The formula of the discriminator's loss function is as follows: .

[0058] Among them, pdata represents the true data distribution and pz is the noise distribution of the generator input.

[0059] Step S252: According to the loss function of the generator and the loss function of the discriminator, alternately optimize the network parameters of the generator and the discriminator.

[0060] The training of the generative adversarial network model in this application is the process of adversarial training of the generator and the discriminator. Its optimization goal is to optimize the loss function of the generator and the loss function of the discriminator through gradient descent. During the training process, the generator and the discriminator continuously adjust their respective network parameters through alternating optimization.

[0061] In one embodiment, during the Nth round of training, the network parameters of the generator are fixed, the loss function of the discriminator is minimized, and the network parameters of the discriminator are updated by gradient descent; during the N+1th round of training, the network parameters of the discriminator are fixed, the loss function of the generator is minimized, and the network parameters of the generator are updated by gradient descent.

[0062] Please combine Figure 5 , Figure 5 It is a schematic diagram of the training method of the wafer polishing surface control model in an application scenario of the present application. As shown in the figure, in one round of training, after the historical processing data is input into the generator, the generator can generate predicted process control parameters, and then the predicted process control parameters and the actual process control parameters are input into the discriminator for training. When training the discriminator, it is necessary to fix the parameters of the generator, and use the back-propagation algorithm to update the parameters of the discriminator to minimize the loss function of the discriminator; in the next round of training, it is necessary to fix the parameters of the discriminator, and also use the back-propagation algorithm to update the parameters of the generator to minimize the loss function of the generator. In the optimization process of the discriminator, it is necessary to minimize the loss function LD of the discriminator, and update the parameters of the discriminator by gradient descent: In the optimization process of the generator, it is necessary to minimize the loss function LG of the generator and update the parameters of the generator through gradient descent: Through adversarial training between the generator and the discriminator, the predicted process control parameters generated by the generator can enable the wafer to obtain a processing result close to its expected surface curve data.

[0063] It can be understood that the training goal of the wafer polishing surface control model is to minimize the difference between the generated predicted process control parameters and the actual process control parameters. This can be measured by the following loss function: .

[0064] in, is the generated predicted process control parameter, ytarget is the real data, and the loss function Lfinal evaluates the difference between the generated result and the expected result by minimizing the Euclidean distance.

[0065] See also Figure 6 , Figure 6 1 is a flow chart of the first embodiment of the wafer polishing surface shape control method provided by the present application. The wafer polishing surface shape control method of the embodiment of the present application is applied to the process control system of the wafer polishing equipment, and the process control system is integrated with a generator of the wafer polishing surface shape control model, and the wafer polishing surface shape control model is obtained by training the wafer polishing surface shape control model training method in any of the above embodiments. Specifically, the wafer polishing surface shape control method in this embodiment includes the following steps: Step S61: when performing polishing processing on the current batch of wafers, obtaining processing data corresponding to the current batch of wafers; wherein the processing data corresponding to the current batch of wafers at least includes the front-polishing profile curve data and the expected profile curve data of the current batch of wafers.

[0066] Step S62: inputting the pre-polishing profile curve data and the expected profile curve data of the current batch of wafers into the generator, and outputting the predicted process control parameters of the current batch of wafers.

[0067] Step S63: Based on the predicted process control parameters of the current batch of wafers, generate and send optimized control instructions to the process control system.

[0068] Step S64: the process control system adjusts the processing parameters of the wafer polishing equipment according to the optimized control instructions to perform polishing processing on the current batch of wafers.

[0069] It can be understood that the wafer polishing surface control model after training can be used for the control of the wafer polishing surface, wherein the generator is integrated into the process control system of the wafer polishing equipment. At this time, the generator no longer relies solely on historical processing data, but can generate dynamically optimized control instructions based on the processing data corresponding to the latest input current batch of wafers. The generator can receive input signals from the process control system in real time, and the input signals of the process control system include processing data corresponding to the current batch of wafers, such as the front polishing profile curve data and the expected profile curve data of the current batch of wafers, so the generator can output the predicted process control parameters of the current batch of wafers in real time according to these signals, generate optimized control instructions and send them to the process control system, for example, the optimized control instructions can be to adjust the parameters of the wafer polishing equipment to the predicted process control parameters of the current batch of wafers in real time, and the process control system can adjust the processing parameters of the wafer polishing equipment according to the optimized control instructions, and polish the current batch of wafers using the predicted process control parameters to achieve the surface profile of the current batch of wafers close to the expected profile curve data, in line with the expected target.

[0070] Therefore, deploying the trained wafer polishing surface control model in the process control system of the wafer polishing equipment can realize real-time control of the polishing process control parameters.

[0071] See also Figure 7 , Figure 7 1 is a flow chart of a second embodiment of a wafer polishing surface shape control method provided by the present application. The wafer polishing surface shape control method in this embodiment comprises the following steps: Step S71: when polishing the current batch of wafers, obtaining processing data corresponding to the current batch of wafers; wherein the processing data corresponding to the current batch of wafers at least includes the front-polishing shape curve data, the expected shape curve data and the auxiliary material life data of the current batch of wafers.

[0072] It is understandable that the life of auxiliary materials such as polishing liquid and polishing pad also has a certain impact on the polishing surface shape of the wafer. Therefore, the difference from the previous embodiment is that the processing data corresponding to the current batch of wafers in this embodiment also includes the auxiliary material life data corresponding to the current batch of wafers.

[0073] Step S72: inputting the front surface curve data and the expected surface curve data of the current batch of wafers before polishing into the convolutional neural network to extract polishing feature information.

[0074] Step S73: inputting the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers into the fully connected neural network to map the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers to the predicted process control parameters of the current batch of wafers.

[0075] The generator of the wafer polishing surface shape control model in this embodiment includes a connected convolutional neural network and a fully connected neural network. The structure combining the convolutional neural network and the fully connected neural network can simultaneously process the front polishing shape curve data, expected surface shape curve data and auxiliary material life data corresponding to the current batch of wafers, and generate predicted process control parameters for the current batch of wafers based on these inputs.

[0076] Step S74: Based on the predicted process control parameters of the current batch of wafers, generate and send optimized control instructions to the process control system.

[0077] Step S75: the process control system adjusts the processing parameters of the wafer polishing equipment according to the optimized control instructions to perform polishing processing on the current batch of wafers.

[0078] Step S74 and step S75 in this embodiment are substantially the same as step S63 and step S64 in the above embodiment, and are not described in detail here.

[0079] In one embodiment, the above-mentioned wafer polishing surface shape control method further includes the following steps: Step S76: Before polishing the next batch of wafers, obtain the historical processing data corresponding to the previous batch of wafers.

[0080] Step S77: updating the parameters of the wafer polishing surface control model according to the historical processing data corresponding to the previous batch of wafers.

[0081] Step S78: The process control system calls the generator with updated parameters to control the polishing process of the next batch of wafers.

[0082] Please combine Figure 8 , Figure 8 It is a schematic diagram of a wafer polishing surface control method in an application scenario of the present application. When the trained wafer polishing surface control model is deployed in the process control system of the wafer polishing equipment, before each polishing, the wafer polishing surface control model can learn the processing data of the previous polishing; for example, before the K+1th polishing, the wafer polishing surface control model can learn the latest Kth processing data and update the parameters of the wafer polishing surface control model; then when performing the K+1th polishing, the process control system calls the generator with updated parameters to generate the predicted process control parameters of the wafer corresponding to the K+1th polishing, generates and sends the optimized control instructions to the process control system, so as to ensure that the generated optimized control instructions can adapt to the continuous changes in the production environment and ensure the accuracy of the process control parameters of the K+1th polishing. This closed-loop feedback mechanism enables the polishing control strategy of each production process to be continuously optimized, and realizes efficient and accurate control strategy updates in a dynamic and complex production environment.

[0083] In the embodiment of the present application, in order to ensure continuous optimization of the generator, the feedback of the discriminator needs to be promptly transmitted to the generator for gradient update to form R2R control, that is, to adjust and optimize future control strategies through historical and real-time feedback information. The generator not only relies on static process control parameters, but can flexibly adjust the control signal and adapt to fluctuations in the production environment through learning from historical and real-time data. This method can achieve precise regulation of the production process and improve the consistency and stability of product quality.

[0084] See also Fig. 9 , Fig. 9It is a structural schematic diagram of an embodiment of a training device for a wafer polishing surface control model provided by the present application. The wafer polishing surface control model is a generative adversarial network model including a generator and a discriminator. The training device 90 for the wafer polishing surface control model of this embodiment includes a first acquisition module 900, a first processing module 902 and a first update module 904 connected to each other. The first acquisition module 900 is used to acquire historical processing data; wherein the historical processing data at least includes actual process control parameters, front-polishing profile curve data and expected profile curve data at different stages. The first processing module 902 is used to input the front-polishing profile curve data and the expected profile curve data into the generator and output predicted process control parameters; input the predicted process control parameters and the actual process control parameters into the discriminator and output true and false discrimination results of the predicted process control parameters and the actual process control parameters respectively. The first update module 904 is used to iteratively update the parameters of the wafer polishing surface control model based on the true and false discrimination results.

[0085] In one embodiment, the generator includes a connected convolutional neural network and a fully connected neural network, and the historical processing data also includes auxiliary material life data at different stages. The first processing module 902 performs the step of inputting the pre-polishing profile curve data and the desired profile curve data into the generator and outputting the predicted process control parameters, including: inputting the pre-polishing profile curve data and the desired profile curve data into the convolutional neural network to extract polishing feature information; inputting the polishing feature information and the auxiliary material life data into the fully connected neural network to map the polishing feature information and the auxiliary material life data to the predicted process control parameters.

[0086] In one embodiment, the first update module 904 performs the step of iteratively updating the parameters of the wafer polishing surface control model based on the true or false discrimination results, including: determining the loss function of the generator according to the true or false discrimination results of the discriminator on the predicted process control parameters; and determining the loss function of the discriminator according to the true or false discrimination results of the discriminator on the predicted process control parameters and the actual process control parameters; and alternately optimizing the network parameters of the generator and the discriminator according to the loss function of the generator and the loss function of the discriminator.

[0087] In one embodiment, the first update module 904 executes the step of alternately optimizing the network parameters of the generator and the discriminator according to the loss function of the generator and the loss function of the discriminator, specifically including: in the Nth round of training, fixing the network parameters of the generator, minimizing the loss function of the discriminator, and updating the network parameters of the discriminator by gradient descent; in the N+1th round of training, fixing the network parameters of the discriminator, minimizing the loss function of the generator, and updating the network parameters of the generator by gradient descent.

[0088] See also Fig.10 , Fig.10 1 is a schematic diagram of the structure of an embodiment of a wafer polishing surface shape control device provided by the present application. The wafer polishing surface shape control device 100 in this embodiment is used to control a wafer polishing device to perform a polishing process on a wafer, and the wafer polishing surface shape control device 100 is integrated with a generator of a wafer polishing surface shape control model, and the wafer polishing surface shape control model is trained by the training method of the wafer polishing surface shape control model of any of the above embodiments.

[0089] The wafer polishing surface shape control device 100 includes a second acquisition module 1000, a second processing module 1002 and a second update module 1004 which are interconnected. The second acquisition module 1000 is used to acquire the processing data corresponding to the current batch of wafers when the current batch of wafers is polished; wherein the processing data corresponding to the current batch of wafers at least includes the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers. The second processing module 1002 is used to input the front polishing shape curve data and the expected surface shape curve data of the current batch of wafers into the generator, output the predicted process control parameters of the current batch of wafers; and generate optimized control instructions based on the predicted process control parameters of the current batch of wafers. The second update module 1004 is used to adjust the processing parameters of the wafer polishing equipment according to the optimized control instructions to perform polishing on the current batch of wafers.

[0090] In one embodiment, the generator includes a connected convolutional neural network and a fully connected neural network, and the processing data corresponding to the current batch of wafers also includes the auxiliary material life data corresponding to the current batch of wafers. The second processing module 1002 performs the steps of inputting the front-polishing profile curve data and the expected profile curve data of the current batch of wafers into the generator and outputting the predicted process control parameters of the current batch of wafers, including: inputting the front-polishing profile curve data and the expected profile curve data of the current batch of wafers into the convolutional neural network to extract polishing feature information; inputting the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers into the fully connected neural network to map the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers to the predicted process control parameters of the current batch of wafers.

[0091] In one embodiment, the second processing module 1002 is also used to obtain historical processing data corresponding to a previous batch of wafers before polishing the next batch of wafers; update the parameters of the wafer polishing surface control model according to the historical processing data corresponding to the previous batch of wafers; and the process control system calls the generator with updated parameters to control the polishing process of the next batch of wafers.

[0092] See also Fig.11 , Fig.11 It is a schematic diagram of the structure of an embodiment of an electronic device of the present application. The electronic device 110 in this embodiment includes a processor 1102 and a memory 1101 connected to each other; the memory 1101 is used to store program instructions, and the processor 1102 is used to execute the program instructions stored in the memory 1101 to implement the steps of any of the above-mentioned training method embodiments of the wafer polishing surface control model, and / or the steps of any of the above-mentioned wafer polishing surface control method embodiments. In a specific implementation scenario, the electronic device 110 may include but is not limited to: a microcomputer, a server.

[0093] Specifically, the processor 1102 is used to control itself and the memory 1101 to implement the steps of the training method embodiment of any of the above-mentioned wafer polishing surface type control models, and / or the steps of any of the above-mentioned wafer polishing surface type control method embodiments. The processor 1102 can also be called a CPU (Central Processing Unit). The processor 1102 may be an integrated circuit chip with signal processing capabilities. The processor 1102 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 1102 can be implemented by an integrated circuit chip.

[0094] See also Fig.12 , Fig.12 The computer-readable storage medium 120 of the present application stores program instructions 1200 thereon, which, when executed by a processor, implement the steps of any of the above-mentioned training methods for the wafer polishing surface control model and / or the steps of any of the above-mentioned methods for controlling the wafer polishing surface.

[0095] The computer-readable storage medium 120 may specifically be a medium that can store the program instructions 1200, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or may be a server that stores the program instructions 1200. The server may send the stored program instructions 1200 to other devices for execution, or may also execute the stored program instructions 1200 by itself.

[0096] In the several embodiments provided in the present application, it should be understood that the disclosed methods, devices and apparatuses can be implemented in other ways. For example, the above-described device and apparatus implementation methods are merely schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0097] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0098] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0099] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0100] If the technical solution of this application involves personal information, the product using the technical solution of this application has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using the technical solution of this application has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, the personal information processing rules are notified by obvious signs / information, and the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A method for training a wafer polishing surface control model, characterized in that: The wafer polishing surface control model is a generative adversarial network model including a generator and a discriminator, and the method includes: Acquire historical processing data; wherein the historical processing data at least includes actual process control parameters at different stages, surface profile curve data before polishing, and expected surface profile curve data; Input the pre-polishing profile curve data and the desired profile curve data into the generator, and output the predicted process control parameters; Inputting the predicted process control parameter and the actual process control parameter into the discriminator, and outputting true or false discrimination results of the predicted process control parameter and the actual process control parameter respectively; The parameters of the wafer polishing surface control model are iteratively updated based on the true and false discrimination results.

2. The training method of the wafer polishing surface control model according to claim 1, characterized in that: The generator includes a connected convolutional neural network and a fully connected neural network, and the historical processing data also includes auxiliary material life data at different stages; The step of inputting the pre-polishing profile curve data and the expected profile curve data into the generator and outputting the predicted process control parameters comprises: Inputting the pre-polishing profile curve data and the desired profile curve data into the convolutional neural network to extract polishing feature information; The polishing characteristic information and the auxiliary material life data are input into the fully connected neural network to map the polishing characteristic information and the auxiliary material life data to the predicted process control parameters.

3. The training method of the wafer polishing surface control model according to claim 2, characterized in that: The convolutional neural network includes a plurality of convolutional layers and pooling layers connected alternately; the convolutional layer is used to extract features from the inputted pre-polishing profile curve data and the desired profile curve data; the pooling layer is used to perform a pooling operation on the features extracted by the convolutional layer; And / or, the fully connected neural network includes multiple first fully connected layers connected in sequence; the first fully connected layer is used to perform nonlinear mapping on the polishing feature information extracted by the convolutional neural network, and finally output the predicted process control parameters.

4. The training method of the wafer polishing surface control model according to claim 1, characterized in that: The discriminator includes a deep neural network, which includes multiple second fully connected layers connected in sequence; each neuron of each second fully connected layer is connected to all neurons of the second fully connected layer of the previous layer, and the number of neurons in adjacent second fully connected layers gradually decreases; the second fully connected layer is used to perform nonlinear mapping on the input predicted process control parameters and the actual process control parameters, and finally output the true or false discrimination result.

5. The method for training a wafer polishing surface control model according to claim 1, characterized in that: The step of iteratively updating the parameters of the wafer polishing surface control model based on the true and false discrimination result comprises: Determining the loss function of the generator according to the true or false discrimination result of the discriminator on the predicted process control parameter; and determining the loss function of the discriminator according to the true or false discrimination result of the discriminator on the predicted process control parameter and the actual process control parameter; According to the loss function of the generator and the loss function of the discriminator, the network parameters of the generator and the discriminator are alternately optimized.

6. The method for training a wafer polishing surface control model according to claim 5, characterized in that: The step of alternately optimizing the network parameters of the generator and the discriminator according to the loss function of the generator and the loss function of the discriminator comprises: During the Nth round of training, the network parameters of the generator are fixed, the loss function of the discriminator is minimized, and the network parameters of the discriminator are updated by gradient descent; During the N+1th round of training, the network parameters of the discriminator are fixed, the loss function of the generator is minimized, and the network parameters of the generator are updated by gradient descent.

7. A method for controlling a wafer polishing surface, characterized in that: A process control system applied to a wafer polishing device, the process control system integrating a generator of a wafer polishing surface shape control model, the wafer polishing surface shape control model being obtained by training using the wafer polishing surface shape control model training method according to any one of claims 1 to 6; The wafer polishing surface shape control method comprises: When performing polishing processing on the current batch of wafers, obtaining processing data corresponding to the current batch of wafers; wherein the processing data corresponding to the current batch of wafers at least includes the front-polishing profile curve data and the expected profile curve data of the current batch of wafers; Inputting the front-polishing profile curve data and the expected profile curve data of the current batch of wafers into the generator, and outputting the predicted process control parameters of the current batch of wafers; Based on the predicted process control parameters of the current batch of wafers, generating and sending optimized control instructions to the process control system; The process control system adjusts the processing parameters of the wafer polishing equipment according to the optimized control instructions to perform polishing processing on the current batch of wafers.

8. The wafer polishing surface shape control method according to claim 7, characterized in that: The generator includes a connected convolutional neural network and a fully connected neural network, and the processing data corresponding to the current batch of wafers also includes auxiliary material life data corresponding to the current batch of wafers; The step of inputting the pre-polishing profile curve data and the expected profile curve data of the current batch of wafers into the generator and outputting the predicted process control parameters of the current batch of wafers comprises: Inputting the front-end profile curve data and the expected profile curve data of the current batch of wafers into the convolutional neural network to extract polishing feature information; The polishing feature information and the auxiliary material life data corresponding to the current batch of wafers are input into the fully connected neural network to map the polishing feature information and the auxiliary material life data corresponding to the current batch of wafers to the predicted process control parameters of the current batch of wafers.

9. The wafer polishing surface shape control method according to claim 7 or 8, characterized in that: The wafer polishing surface shape control method also includes: Before polishing the next batch of wafers, obtain the historical processing data corresponding to the previous batch of wafers; According to the historical processing data corresponding to the previous batch of wafers, updating the parameters of the wafer polishing surface control model; The process control system calls the generator with updated parameters to control the polishing process of the next batch of wafers.

10. A training device for a wafer polishing surface control model, characterized in that: The wafer polishing surface control model is a generative adversarial network model including a generator and a discriminator, and the training device includes: A first acquisition module, the first acquisition module is used to acquire historical processing data; wherein the historical processing data at least includes actual process control parameters at different stages, surface profile curve data before polishing, and expected surface profile curve data; A first processing module, the first processing module is used to input the pre-polishing profile curve data and the expected profile curve data into the generator, and output the predicted process control parameters; input the predicted process control parameters and the actual process control parameters into the discriminator, and output the true and false discrimination results of the predicted process control parameters and the actual process control parameters respectively; A first updating module is used for iteratively updating the parameters of the wafer polishing surface control model based on the true or false discrimination result.

11. A wafer polishing surface control device, characterized in that: The wafer polishing surface shape control device is used to control the wafer polishing equipment to perform polishing processing on the wafer, and the wafer polishing surface shape control device is integrated with a generator of a wafer polishing surface shape control model, and the wafer polishing surface shape control model is trained by the training method of the wafer polishing surface shape control model according to any one of claims 1 to 6; The wafer polishing surface type control device comprises: A second acquisition module, wherein the second acquisition module is used to acquire processing data corresponding to the current batch of wafers when the current batch of wafers is polished; wherein the processing data corresponding to the current batch of wafers at least includes the front-polishing profile curve data and the expected profile curve data of the current batch of wafers; A second processing module, the second processing module is used to input the front-polishing profile curve data and the expected profile curve data of the current batch of wafers into the generator, output the predicted process control parameters of the current batch of wafers; and generate optimized control instructions based on the predicted process control parameters of the current batch of wafers; The second updating module is used to adjust the processing parameters of the wafer polishing equipment according to the optimized control instructions so as to perform polishing processing on the current batch of wafers.

12. An electronic device, characterized in that: The electronic device includes: a memory and a processor coupled to each other, the processor being used to execute program instructions stored in the memory to implement the training method of the wafer polishing surface control model as described in any one of claims 1 to 6, and / or the wafer polishing surface control method as described in any one of claims 7 to 9.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, which can be executed to implement the training method of the wafer polishing surface type control model as described in any one of claims 1 to 6, and / or the wafer polishing surface type control method as described in any one of claims 7 to 9.

Images