Process parameter optimization method and related equipment
By optimizing the process parameters of semiconductor devices through deep neural networks and reinforcement learning algorithms, the problems of low accuracy and efficiency in existing technologies are solved, and automatic optimization of process parameters is achieved.
Patent Information
- Application Number
- CN202510770009.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-10
- Publication Date
- 2025-09-12
AI Technical Summary
In the prior art, the process parameter optimization method for semiconductor devices has problems of low accuracy and efficiency.
A deep neural network system is used to predict the input process parameters, obtain the electrical parameter prediction results, and calculate the device quality score based on the electrical parameter prediction results. The process parameters are optimized through a reinforcement learning algorithm until the device quality score meets the preset conditions.
The accuracy and efficiency of process parameter optimization are improved, and automatic optimization of process parameters is achieved.
Smart Images
Figure CN120633434A_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the field of semiconductor manufacturing, and in particular to a process parameter optimization method and related equipment. Background Art
[0002] In recent years, with the rapid development of deep neural networks, academia and industry have witnessed significant breakthroughs in deep learning across numerous fields. The rapid development of fields such as information science, energy, and national defense has placed diverse demands on semiconductor devices.
[0003] Currently, the method of manually optimizing the process parameters of semiconductor devices based on different requirements for electrical parameters of semiconductor devices has problems of low accuracy and efficiency. Summary of the Invention
[0004] The problem solved by the embodiments of the present invention is to provide a process parameter optimization method and related equipment, which can improve the accuracy and efficiency of process parameter optimization.
[0005] To solve the above problems, an embodiment of the present invention provides a process parameter optimization method, including:
[0006] Obtain input process parameters;
[0007] The input process parameters are predicted and processed through a preset deep neural network system to obtain electrical parameter prediction results;
[0008] Based on the electrical parameter prediction results, obtaining a corresponding device quality score;
[0009] If the device quality score meets the preset conditions, the input process parameters are used as the optimal process parameters;
[0010] If the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, the optimized process parameters are obtained as the input process parameters, and the step of predicting the input process parameters through a preset deep neural network system and obtaining the electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
[0011] Optionally, based on the electrical parameter prediction results, the following formula is used to obtain the corresponding device quality score:
[0012]
[0013] Wherein, S represents the device quality score, N represents the number of electrical parameters in the electrical parameter prediction result, and x irepresents the predicted value of the ith electrical parameter in the electrical parameter prediction result, f i represents the target value of the i-th electrical parameter in the electrical parameter prediction result, α i Represents the weight coefficient of the i-th electrical parameter in the electrical parameter prediction result.
[0014] Optionally, optimizing the input process parameters based on the corresponding device quality scores includes optimizing the input process parameters using a reinforcement learning algorithm based on the corresponding device quality scores.
[0015] Optionally, the reinforcement learning algorithm includes a DDPG algorithm.
[0016] Optionally, the preset condition includes any one of the device quality score converges and the device quality score is less than or equal to a preset threshold.
[0017] Optionally, the deep neural network system includes a fully connected neural network system.
[0018] Accordingly, an embodiment of the present invention further provides a process parameter optimization device, comprising:
[0019] a parameter acquisition unit adapted to acquire input process parameters;
[0020] A parameter optimization unit is suitable for predicting and processing input process parameters through a preset deep neural network system to obtain electrical parameter prediction results; based on the electrical parameter prediction results, obtaining corresponding device quality scores; if the device quality scores meet preset conditions, the input process parameters are used as optimal process parameters; if the device quality scores do not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality scores, and the optimized process parameters are obtained as the input process parameters, and the step of predicting and processing the input process parameters through the preset deep neural network system to obtain electrical parameter prediction results is restarted until the device quality scores meet the preset conditions.
[0021] Optionally, the parameter optimization unit is adapted to calculate a corresponding device quality score based on the electrical parameter prediction result using the following formula:
[0022]
[0023] Wherein, S represents the device quality score, N represents the number of electrical parameters in the electrical parameter prediction result, and x i represents the predicted value of the ith electrical parameter in the electrical parameter prediction result, f i represents the target value of the i-th electrical parameter in the electrical parameter prediction result, αi Represents the weight coefficient of the i-th electrical parameter in the electrical parameter prediction result.
[0024] Optionally, the parameter optimization unit is adapted to optimize the input process parameters using a reinforcement learning algorithm based on the corresponding device quality scores.
[0025] Optionally, the reinforcement learning algorithm includes a DDPG algorithm.
[0026] Optionally, the preset condition includes any one of the device quality score converges and the device quality score is less than or equal to a preset threshold.
[0027] Optionally, the deep neural network system includes a fully connected neural network system.
[0028] Correspondingly, an embodiment of the present invention also provides a computer device, characterized in that it includes at least one memory and at least one processor, the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the process parameter optimization method as described in any one of the above items.
[0029] Accordingly, an embodiment of the present invention further provides a computer program product, comprising a computer program / instruction, which is suitable for implementing any of the process parameter optimization methods described above when executed by a processor.
[0030] Correspondingly, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, and the one or more computer instructions are suitable for implementing any of the process parameter optimization methods described above.
[0031] Compared with the prior art, the technical solution of the embodiment of the present invention has the following advantages:
[0032] The process parameter optimization method provided by an embodiment of the present invention includes: obtaining input process parameters; predicting and processing the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results; obtaining corresponding device quality scores based on the electrical parameter prediction results; if the device quality score meets preset conditions, the input process parameters are used as optimal process parameters; if the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, and the optimized process parameters are obtained as the input process parameters, and the step of predicting and processing the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
[0033] The process parameter optimization method provided in an embodiment of the present invention first predicts and processes the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results, and based on the electrical parameter prediction results, obtains corresponding device quality scores, and then optimizes the input process parameters based on the corresponding device quality scores until the device quality scores meet preset conditions. Compared with the method of manually optimizing the input process parameters according to the electrical parameter prediction results, the accuracy and efficiency of process parameter optimization can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 This is a flow chart of an embodiment of a process parameter optimization method provided by the technical solution of the present invention;
[0035] Figure 2 It is a structural diagram of an embodiment of a fully connected neural network system in the process parameter optimization method provided by the technical solution of the present invention;
[0036] Figure 3 It is a structural diagram of an embodiment of an intermediate layer of a fully connected neural network system in the process parameter optimization method provided by the technical solution of the present invention;
[0037] Figure 4 It is a structural schematic diagram of an embodiment of a process parameter optimization device provided by the technical solution of the present invention;
[0038] Figure 5 This is a schematic diagram of an optional hardware structure of a computer device provided by the technical solution of the present invention. DETAILED DESCRIPTION
[0039] As can be seen from the background art, the method of manually optimizing the process parameters of semiconductor devices based on the requirements for the electrical parameters of semiconductor devices has problems of low accuracy and efficiency.
[0040] In order to solve the above technical problems, an embodiment of the present invention provides a process parameter optimization method, including: obtaining input process parameters; predicting and processing the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results; obtaining corresponding device quality scores based on the electrical parameter prediction results; if the device quality score meets the preset conditions, the input process parameters are used as the optimal process parameters; if the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, and the optimized process parameters are obtained as the input process parameters, and the step of predicting and processing the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
[0041] The process parameter optimization method provided in an embodiment of the present invention first predicts and processes the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results, and based on the electrical parameter prediction results, obtains corresponding device quality scores, and then optimizes the input process parameters based on the corresponding device quality scores until the device quality scores meet preset conditions. Compared with the method of manually optimizing the input process parameters according to the electrical parameter prediction results, the accuracy and efficiency of process parameter optimization can be improved.
[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0043] Figure 1 The following is a flow chart showing an embodiment of a process parameter optimization method provided by the technical solution of the present invention. Figure 1 A process parameter optimization method may include the following steps:
[0044] Step S110: obtaining input process parameters;
[0045] Step S120: performing prediction processing on the input process parameters via a preset deep neural network system to obtain electrical parameter prediction results;
[0046] Step S130: obtaining a corresponding device quality score based on the electrical parameter prediction result;
[0047] Step S140: Determine whether the device quality score meets a preset condition; if the determination result is no, then execute step S150; if the determination result is yes, then execute step S160;
[0048] Step S150: Optimizing the input process parameters based on the corresponding device quality scores, obtaining the optimized process parameters as the input process parameters, and restarting from step S120;
[0049] Step S160: taking the input process parameters as the optimal process parameters.
[0050] Please continue to refer to Figure 1 , execute step S110 to obtain input process parameters.
[0051] Obtaining input process parameters provides a basis for subsequent prediction processing of the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results.
[0052] The input process parameters refer to parameters related to the manufacturing process of the semiconductor device, which can be set by the designer of the semiconductor device according to actual needs and are not limited here.
[0053] In an exemplary embodiment, the input process parameters are process parameters of a complementary metal oxide semiconductor (CMOS) device. Accordingly, the parameters related to the manufacturing process of the CMOS device may include one or more of process parameters of a gate oxide layer formation process, process parameters of an N-type polysilicon gate pre-doping process, process parameters of a polysilicon gate etching process, process parameters of an offset spacer formation process, process parameters of a pocket implantation process, process parameters of a spacer formation process, process parameters of a source / drain implantation process, and process parameters of an annealing process.
[0054] Please continue to refer to Figure 1 , executing step S120, performing prediction processing on the input process parameters through a preset neural network system to obtain electrical parameter prediction results.
[0055] The input process parameters are predicted and processed through a preset neural network system to obtain electrical parameter prediction results, which provide a basis for subsequently obtaining corresponding device quality scores based on the electrical parameter prediction results.
[0056] In one exemplary embodiment, the deep neural network system is a fully connected neural network system. A fully connected neural network is a classic feedforward neural network structure, typically composed of multiple stacked layers of neurons, with each layer of neurons connected to all neurons in the previous layer. This allows the fully connected neural network to perform complex feature extraction and nonlinear transformations, giving the model stronger expressive power.
[0057] Reference Figure 2 The fully connected neural network system described includes an input layer 210, an intermediate layer 220, and an output layer 230. The input layer 210 is adapted to obtain the input process parameters, the intermediate layer 220 is adapted to extract the features of the input process parameters, and the output layer 230 is adapted to output electrical parameter prediction results for the input process parameters.
[0058] Accordingly, the intermediate layer 220 includes at least one fully connected neural network layer. Specifically, the at least one fully connected neural network layer includes at least one network block.
[0059] In an exemplary embodiment, referring to Figure 3The network block 30 described herein includes a cascaded fully connected layer 310 and an activation function layer 320. The fully connected layer 310 receives multiple input features and performs a fully connected operation on the received multiple input features to obtain corresponding multiple fully connected features; the activation function layer 320 performs a nonlinear transformation on the multiple fully connected features to obtain corresponding transformed features.
[0060] The fully connected layer 310 performs a weighted sum operation on the input from the previous layer of neurons and adds a bias term, and then the activation function layer 320 performs a nonlinear transformation operation on the fully connected features from the fully connected layer 310, so that at least one fully connected neural network layer can capture the nonlinear relationship in the data and improve the expressive power of the neural network.
[0061] Common activation functions include Sigmoid activation function, ReLU (Rectified Linear Unit), Tanh activation function, etc. In an exemplary embodiment, the activation function layer 320 is a ReLU activation function.
[0062] In an exemplary embodiment, the input process parameters are process parameters of a CMOS device. Accordingly, the output layer 230 outputs electrical parameter prediction results for the input process parameters, including one or more of threshold voltage, on-state current, off-state leakage current, transconductance, on-resistance, parasitic capacitance, gate oxide breakdown voltage, and source-drain breakdown voltage.
[0063] Generally speaking, data processing using deep neural network systems can be divided into two phases: training and data processing. The training phase involves first training the neural network using training data to adjust the neural network's weights (also known as parameters). The data processing phase involves subsequently using the trained neural network to extract features from the input data and perform other tasks such as identifying and classifying objects within the data.
[0064] In an exemplary embodiment, the training process of a deep neural network system for electrical parameter prediction includes: a data collection phase, dividing the sample data set obtained from the production line into a training set and a validation set according to a preset ratio; a data training phase, using the training set to train the fully connected neural network system until the loss value of the fully connected neural network system on the preset validation set reaches convergence.
[0065] During the data collection phase, the process parameter dimension can be the union of the process parameters under all process categories. Furthermore, the dimension of the process parameters under each process category can be different. That is, the fully connected neural network system in this application can be used in situations where the process parameter dimension is not fixed, which is conducive to expanding the scope of application of the fully connected neural network system and meeting the diverse needs of users.
[0066] In an exemplary embodiment, a preset number of training data sets are used each time to perform an iterative training on the neural network system with initial weights, thereby completing an adjustment of the weights of the neural network system.
[0067] Specifically, the process of each iterative training includes: using a preset number of training data to train the deep neural network system to be trained for electrical parameter prediction, and obtaining a preset number of training results; according to the difference between the training results and the corresponding correct results, respectively using a preset loss function to calculate the loss value; performing network back propagation derivation based on the calculated loss value to obtain the gradient value; according to the gradient value obtained by back propagation derivation, adjusting the weight of the deep neural network system for electrical parameter prediction.
[0068] Accordingly, multiple iterations of training are performed, and the weights of the deep neural network system for electrical parameter prediction are continuously iteratively updated until the loss value of the deep neural network system for electrical parameter prediction on the validation set converges. The convergence of the loss value of the deep neural network system for electrical parameter prediction on the validation set means that the loss value of the deep neural network system for electrical parameter prediction on the validation set reaches a minimum value.
[0069] For more detailed information about the multiple iterative training processes of the deep neural network system for electrical parameter prediction, please refer to the prior art on the iterative training process of the neural network system, which will not be repeated here.
[0070] Please continue to refer to Figure 1 , executing step S130, obtaining a corresponding device quality score based on the electrical parameter prediction result.
[0071] Based on the electrical parameter prediction results, a corresponding device quality score is obtained, which provides a basis for subsequent judgment on whether the device quality score meets preset conditions.
[0072] In an exemplary embodiment, based on the electrical parameter prediction results, the corresponding device quality score is calculated using the following formula:
[0073]
[0074] Wherein, S represents the device quality score, N represents the number of electrical parameters in the electrical parameter prediction result, and x i represents the predicted value of the ith electrical parameter in the electrical parameter prediction result, f i represents the target value of the i-th electrical parameter in the electrical parameter prediction result, α i Represents the weight coefficient of the i-th electrical parameter in the electrical parameter prediction result.
[0075] The number N of electrical parameters in the electrical parameter prediction result in the above formula (1) can have different values depending on the device. In other words, the dimensions of the electrical parameter prediction result can be different for different devices, which can be achieved by training the deep neural network.
[0076] The weight coefficient α of the i-th electrical parameter in the electrical parameter prediction result described in the above formula (1) is i , can be set according to actual needs. For example, a larger weight coefficient is set for the more important electrical parameters in the electrical parameter prediction results, and a smaller weight coefficient is set for the electrical parameters with relatively less importance.
[0077] It can be seen from the above formula (1) that the smaller the difference between the predicted value and the target value of each electrical parameter in the electrical parameter prediction result, the lower the device quality score; the larger the difference between the predicted value and the target value of each electrical parameter in the electrical parameter prediction result, the higher the device quality score.
[0078] Please continue to refer to Figure 1 , execute step S140 to determine whether the device quality score meets the preset conditions; if the judgment result is no, execute step S150; if the judgment result is yes, execute step S160.
[0079] In an exemplary embodiment, the step of determining whether the device quality score meets a preset condition includes determining whether the device quality score converges. Determining whether the device quality score converges refers to determining whether the device quality score reaches a minimum extreme value.
[0080] In other embodiments, it is determined whether the device quality score is less than or equal to a preset threshold, wherein the threshold can be set by a person skilled in the art according to actual needs and is not limited here.
[0081] Please continue to refer to Figure 1, execute step S150, optimize the input process parameters based on the corresponding device quality score, obtain the optimized process parameters as the input process parameters, and restart execution from step S120.
[0082] The input process parameters are optimized based on the corresponding device quality score, the optimized process parameters are obtained as the input process parameters, and the step of predicting the input process parameters through a preset deep neural network system and obtaining the electrical parameter prediction results is restarted until the device quality score meets the preset conditions. The corresponding input process parameters are used as the optimal process parameters, which can realize automatic optimization of the process parameters. Compared with the manual process parameter optimization method, it is beneficial to improve the accuracy and efficiency of process parameter optimization.
[0083] In an exemplary embodiment, the step of optimizing the input process parameters based on the corresponding device quality scores includes optimizing the input process parameters using a reinforcement learning (RL) algorithm based on the corresponding device quality scores.
[0084] Reinforcement learning, also known as reinforcement learning, evaluation learning or enhanced learning, is used to describe and solve the problem of how an agent can maximize rewards or achieve specific goals by learning strategies during its interaction with the environment.
[0085] Generally speaking, reinforcement learning doesn't require a training dataset. Instead, it relies on reinforcement signals (i.e., rewards) provided by the environment to evaluate the quality of actions, rather than telling the reinforcement learning system how to generate the correct actions. In other words, reinforcement learning is a machine learning paradigm used to solve the problem of how an intelligent agent learns strategies to maximize rewards or complete specific machine learning tasks during its interaction with the environment. The agent learns through trial and error, receiving rewards through interaction with the environment, which guide its action selection. It finds the optimal strategy for the current state and selects appropriate actions based on that strategy to maximize its reward.
[0086] In an exemplary embodiment, the reinforcement learning algorithm is a Deep Deterministic Policy Gradient (DDPG) algorithm.
[0087] The DDPG algorithm is a deep reinforcement learning-based algorithm within the Actor Critic framework. It maintains two networks: a policy network and a Q network. The policy network outputs the action to be taken in the current state, while the Q network outputs the Q value of the action taken in the current state. Because it combines both Deterministic Policy Gradient (DPG) and Deep Q Network (DQN), it offers the advantages of both DPG and DQN.
[0088] Please continue to refer to Figure 1 , execute step S160 and use the input process parameters as the optimal process parameters.
[0089] When the device quality score meets a preset condition, such as when the device quality score converges to a minimum extreme value, it indicates that the corresponding input process parameters have reached an optimum. In this case, the corresponding input process parameters can be used as the optimal process parameters. The corresponding input process parameters can be initial input process parameters or input process parameters obtained by optimizing the initial input process parameters.
[0090] Correspondingly, an embodiment of the present invention also provides a process parameter optimization device.
[0091] Figure 4 The schematic diagram of the structure of an embodiment of the process parameter optimization device provided by the technical solution of the present invention is shown. Figure 4 A process parameter optimization device 40 may include: a parameter acquisition unit 401, suitable for acquiring input process parameters; a parameter optimization unit 402, suitable for predicting and processing the input process parameters through a preset deep neural network system to obtain electrical parameter prediction results; based on the electrical parameter prediction results, obtaining a corresponding device quality score; if the device quality score meets the preset conditions, the input process parameters are used as the optimal process parameters; if the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, and the optimized process parameters are obtained as the input process parameters, and the step of predicting and processing the input process parameters through the preset deep neural network system to obtain electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
[0092] In an exemplary embodiment, the parameter optimization unit 402 is adapted to calculate the corresponding device quality score based on the electrical parameter prediction result using the above formula (1). The specific content of formula (1) is described above and will not be repeated here.
[0093] In an exemplary embodiment, the parameter optimization unit 402 is adapted to optimize the input process parameters using a reinforcement learning algorithm based on the corresponding device quality scores.
[0094] In an exemplary embodiment, the reinforcement learning algorithm comprises a DDPG algorithm.
[0095] In an exemplary embodiment, the preset condition includes any one of the device quality score converging and being less than or equal to a preset threshold.
[0096] In an exemplary embodiment, the deep neural network system comprises a fully connected neural network system.
[0097] The process parameter optimization device in the embodiment of the present invention can be used to execute the aforementioned process parameter optimization method, or other functional modules can be used to execute the aforementioned process parameter optimization method. For the process parameter optimization method provided by the embodiment of the present invention, please refer to the detailed description of the aforementioned part, which will not be repeated here.
[0098] Correspondingly, an embodiment of the present invention further provides a computer device, which can implement the process parameter optimization method provided by the embodiment of the present invention in the form of a loaded program.
[0099] refer to Figure 5 , which shows an optional hardware structure diagram of a computer device provided in an embodiment of the present invention. The computer device in the embodiment of the present invention includes: at least one processor 01, at least one communication interface 02, at least one memory 03 and at least one communication bus 04.
[0100] In this embodiment, the number of each of the processor 01 , the communication interface 02 , the memory 03 and the communication bus 04 is at least one, and the processor 01 , the communication interface 02 and the memory 03 communicate with each other via the communication bus 04 .
[0101] The communication interface 02 may be an interface of a communication module suitable for network communication, such as an interface of a GSM module.
[0102] The processor 01 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the process parameter optimization method of this embodiment.
[0103] The memory 03 may include a high-speed RAM memory, or may also include a non-volatile memory, such as at least one disk storage. The memory 03 stores one or more computer instructions, which are executed by the processor 01 to implement the process parameter optimization method provided in the aforementioned embodiment.
[0104] It should be noted that the above-mentioned computer device may also include other devices (not shown) that may not be necessary for understanding the contents disclosed in the embodiments of the present invention; since these other devices may not be necessary for understanding the contents disclosed in the embodiments of the present invention, the embodiments of the present invention will not introduce them one by one.
[0105] Accordingly, an embodiment of the present invention further provides a computer program product, including a computer program / instruction, which is suitable for implementing the process parameter optimization method described in the embodiment of the present invention when executed by a processor.
[0106] An embodiment of the present invention further provides a storage medium, wherein the storage medium stores one or more computer instructions, and the one or more computer instructions are suitable for implementing the process parameter optimization method provided in the above embodiment.
[0107] The embodiments of the present invention described above are combinations of elements and features of the present invention. Unless otherwise mentioned, elements or features may be considered as optional. Each element or feature may be put into practice without being combined with other elements or features. In addition, embodiments of the present invention may be constructed by combining some elements and / or features. The order of operations described in the embodiments of the present invention may be rearranged. Some configurations of any one embodiment may be included in another embodiment and may be replaced by the corresponding configuration of another embodiment. It is obvious to those skilled in the art that claims that do not have a clear reference relationship to each other in the appended claims may be combined into embodiments of the present invention, or may be included as new claims in amendments after submitting this application.
[0108] The embodiments of the present invention may be implemented by various means such as hardware, firmware, software, or a combination thereof. In a hardware configuration, the method according to the exemplary embodiment of the present invention may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, etc.
[0109] In a firmware or software configuration, the embodiments of the present invention may be implemented in the form of modules, procedures, functions, and the like. Software codes may be stored in a memory unit and executed by a processor. The memory unit may be located inside or outside the processor and may send and receive data to and from the processor via various known means.
[0110] The above description of the disclosed embodiments will enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.
[0111] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope defined by the claims.
Claims
1. A process parameter optimization method, characterized in that: include: Obtain input process parameters; The input process parameters are predicted and processed through a preset deep neural network system to obtain electrical parameter prediction results; Based on the electrical parameter prediction results, obtaining a corresponding device quality score; If the device quality score meets the preset conditions, the input process parameters are used as the optimal process parameters; If the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, the optimized process parameters are obtained as the input process parameters, and the step of predicting the input process parameters through a preset deep neural network system and obtaining the electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
2. The process parameter optimization method according to claim 1, wherein: Based on the electrical parameter prediction results, the following formula is used to obtain the corresponding device quality score: Wherein, S represents the device quality score, N represents the number of electrical parameters in the electrical parameter prediction result, and x i represents the predicted value of the ith electrical parameter in the electrical parameter prediction result, f i represents the target value of the i-th electrical parameter in the electrical parameter prediction result, α i Represents the weight coefficient corresponding to the i-th electrical parameter in the electrical parameter prediction result.
3. The process parameter optimization method according to claim 1, wherein: The optimizing the input process parameters based on the corresponding device quality scores includes optimizing the input process parameters using a reinforcement learning algorithm based on the corresponding device quality scores.
4. The process parameter optimization method according to claim 3, wherein: The reinforcement learning algorithm includes the DDPG algorithm.
5. The process parameter optimization method according to claim 1, wherein: The preset condition includes any one of the following: the device quality score converges and the device quality score is less than or equal to a preset threshold.
6. The process parameter optimization method according to claim 1, wherein: The deep neural network system includes a fully connected neural network system.
7. A process parameter optimization device, characterized in that: include: a parameter acquisition unit adapted to acquire input process parameters; A parameter optimization unit, adapted to predict and process input process parameters via a preset deep neural network system to obtain electrical parameter prediction results; Based on the electrical parameter prediction results, obtaining a corresponding device quality score; If the device quality score meets the preset conditions, the input process parameters are used as the optimal process parameters; If the device quality score does not meet the preset conditions, the input process parameters are optimized based on the corresponding device quality score, the optimized process parameters are obtained as the input process parameters, and the step of predicting the input process parameters through a preset deep neural network system and obtaining the electrical parameter prediction results is restarted until the device quality score meets the preset conditions.
8. The process parameter optimization device according to claim 7, characterized in that: The parameter optimization unit is adapted to calculate the corresponding device quality score based on the electrical parameter prediction result using the following formula: Wherein, S represents the device quality score, N represents the number of electrical parameters in the electrical parameter prediction result, and x i represents the predicted value of the ith electrical parameter in the electrical parameter prediction result, f i represents the target value of the i-th electrical parameter in the electrical parameter prediction result, α i Represents the weight coefficient of the i-th electrical parameter in the electrical parameter prediction result.
9. The process parameter optimization device according to claim 7, characterized in that: The parameter optimization unit is adapted to optimize the input process parameters using a reinforcement learning algorithm based on the corresponding device quality scores.
10. The process parameter optimization device according to claim 9, characterized in that: The reinforcement learning algorithm includes the DDPG algorithm.
11. The process parameter optimization device according to claim 7, characterized in that: The preset condition includes any one of the following: the device quality score converges and the device quality score is less than or equal to a preset threshold.
12. The process parameter optimization device according to claim 7, characterized in that: The deep neural network system includes a fully connected neural network system.
13. A computer device, characterized in that: The method comprises at least one memory and at least one processor, wherein the memory stores one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the process parameter optimization method according to any one of claims 1 to 6.
14. A computer program product comprising a computer program / instructions, characterized in that The computer program / instructions, when executed by a processor, are suitable for implementing the process parameter optimization method according to any one of claims 1 to 6.
15. A storage medium, characterized in that: The storage medium stores one or more computer instructions, and the one or more computer instructions are suitable for implementing the process parameter optimization method according to any one of claims 1 to 6.