Process control parameter optimization method and device, electronic equipment and storage medium

Abstract

The application provides a process control parameter optimization method and device, electronic equipment and a storage medium. The method comprises obtaining each prediction state data corresponding to each detection point according to each process control parameter corresponding to each detection point in a process control sequence, determining target state data of a target detection point in each prediction state data according to a target detection point and an adjustment detection point determined in each detection point, and obtaining actual state data of the target detection point. The target state data and the actual state data are analyzed to obtain state analysis data of the target detection point, and the process control parameter of the adjustment detection point is optimized according to the state analysis data of the target detection point. Therefore, the application can provide an optimization scheme of the process control parameter in real time and dynamically during the process treatment, so as to improve the production efficiency and the product yield.

Description

Technical Field

[0001] This application relates to the field of industrial automation technology, and in particular to a method, apparatus, electronic device and storage medium for optimizing process control parameters. Background Technology

[0002] In industrial processes, real-time production optimization of dynamic systems has always been a topic of great interest. Process engineers need to observe measurement data from the process section and production environment data in real time, and adjust process control parameters accordingly to improve production efficiency or ensure yield.

[0003] Currently, the optimization of process control parameters is broadly categorized into two types: dynamic optimization and steady-state optimization. Specifically, for industrial processes, if the trajectory shape of multiple process control parameters as they adjust from their initial values ​​to target values ​​affects the final production result, then dynamic models are needed for optimization. In this case, the optimization algorithm needs to provide control suggestions at each time step to achieve the desired trajectory. If, in actual operation, the stable values ​​of the control parameters are the main factors affecting the production result, then steady-state optimization performed by steady-state models has greater applicability.

[0004] Generally speaking, in the dynamic optimization of process control parameters, the current state of the process control system is greatly affected by the historical state, and the calculation time of the optimization scheme cannot exceed the given time step, which makes the calculation speed more demanding and the dynamic optimization of process control parameters more complicated. Summary of the Invention

[0005] In view of this, the process control parameter optimization method, apparatus, electronic equipment and storage medium provided in this application can provide optimization schemes for process control parameters in real time and dynamically during the process, so as to improve production efficiency and product yield.

[0006] According to a first aspect of the present application, a method for optimizing process control parameters is provided, comprising: predicting the states of a process system under test corresponding to each detection point based on process control parameters corresponding to each detection point in a process control sequence, and obtaining predicted state data corresponding to each detection point; determining at least one target state data corresponding to the at least one target detection point from the predicted state data based on at least one target detection point and at least one detection point to be adjusted determined in each detection point, and detecting the state of the process system under test corresponding to the at least one target detection point, and obtaining at least one actual state data of the at least one target detection point; analyzing the at least one target state data and the at least one actual state data to obtain state analysis data of the at least one target detection point; and optimizing at least one process control parameter of the at least one detection point to be adjusted based on the state analysis data of the at least one target detection point.

[0007] According to a second aspect of the embodiments of this application, a process control parameter optimization apparatus is provided, comprising: a prediction unit, configured to predict each state of a process system under test corresponding to each detection point based on each process control parameter corresponding to each detection point in a process control sequence, and obtain each predicted state data corresponding to each detection point; a determination unit, configured to determine at least one target state data corresponding to the at least one target detection point from the predicted state data based on at least one target detection point and at least one detection point to be adjusted determined in each detection point, and detect the state of the process system under test corresponding to the at least one target detection point, and obtain at least one actual state data of the at least one target detection point; an analysis unit, configured to analyze the at least one target state data and the at least one actual state data to obtain state analysis data of the at least one target detection point; and an optimization unit, configured to optimize at least one process control parameter of the at least one detection point to be adjusted based on the state analysis data of the at least one target detection point.

[0008] According to a third aspect of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a bus, wherein the processor, the memory, and the communication interface communicate with each other via the bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the process control parameter optimization method described in the first aspect above.

[0009] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided, wherein computer instructions are stored on the computer-readable storage medium, and when executed by a processor, the computer instructions cause the processor to perform the process control parameter optimization method as described in the second aspect above.

[0010] The process control parameter optimization schemes provided in this application optimize the process control parameters of each detection point by predicting the predicted state data corresponding to each detection point in the process control sequence, analyzing the target state data and actual state data of the target detection point, and so on. This allows for the real-time dynamic provision of the optimal combination of process control parameters, which not only improves production efficiency but also increases product yield. Attached Figure Description

[0011] Figure 1 This is a flowchart of a process control parameter optimization method according to an exemplary embodiment of this application.

[0012] Figure 2 This is a flowchart of a process control parameter optimization method according to another exemplary embodiment of this application.

[0013] Figure 3 This is a flowchart of a process control parameter optimization method according to another exemplary embodiment of this application.

[0014] Figure 4 This is a schematic diagram of the process control parameter optimization device in an exemplary embodiment of this application.

[0015] Figure 5 This is a schematic diagram of an electronic device according to an exemplary embodiment of this application.

[0016] List of reference numerals in the attached diagram:

[0017] 102: Based on the process control parameters corresponding to each detection point in the process control sequence, predict the state of the process system under test corresponding to each detection point, and obtain the predicted state data corresponding to each detection point.

[0018] 104: Based on the target detection point and the detection point to be adjusted determined in each detection point, determine the target state data corresponding to the target detection point in each predicted state data, and detect the state of the process system under test corresponding to the target detection point to obtain the actual state data of the target detection point.

[0019] 106: Analyze the target state data and actual state data to obtain the state analysis data of the target detection points.

[0020] 108: Based on the status analysis data of the target detection points, optimize the process control parameters of the detection points to be adjusted.

[0021] 202: Obtain the training sample set of the process system under test

[0022] 204: Determine the current sampling point and the target sampling point among all sampling points.

[0023] 206: Using the state prediction model to be trained, based on the process control parameters and environmental influencing factor parameters of the process system under test corresponding to the current sampling point, predict the state of the process system under test corresponding to the target sampling point, and obtain the preset state data of the target sampling point.

[0024] 208: Based on the predicted state data and actual state data of the target sampling points, train the state prediction model to be trained to obtain the state prediction model.

[0025] 302: Determine the process control parameters corresponding to each detection point in the process control sequence as the process control parameters to be optimized for each detection point.

[0026] 304: Initialize parameter values ​​for the optimization algorithm.

[0027] 306: Based on the parameter values ​​of the parameter optimization algorithm, perform parameter optimization on each process control parameter to be optimized corresponding to each detection point to obtain each intermediate process control parameter corresponding to each detection point.

[0028] 308: Based on the intermediate process control parameters corresponding to each detection point, predict the states of the process system under test corresponding to each detection point, and obtain the intermediate predicted state data and prediction result evaluation data corresponding to each detection point. 310: Update the parameter values ​​of the parameter optimization algorithm based on the prediction result evaluation data corresponding to each detection point, and update the intermediate process control parameters corresponding to each detection point to the process control parameters to be optimized corresponding to each detection point.

[0029] 312: Determine whether the parameter optimization meets the preset parameter optimization termination condition.

[0030] 314: Obtain the optimized process control parameters for each detection point.

[0031] 316: Based on the optimized process control parameters corresponding to each detection point, predict the state of the process system under test corresponding to each detection point, and obtain the predicted state data corresponding to each detection point.

[0032] 400: Process control parameter optimization device; 402: Prediction unit; 404: Determination unit

[0033] 406: Analysis Unit; 408: Optimization Unit; 500: Electronic Equipment

[0034] 502: Processor; 504: Communication Interface; 506: Memory

[0035] 508: Bus 510: Program Detailed Implementation

[0036] To enable those skilled in the art to better understand the technical solutions in the embodiments of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art should fall within the protection scope of the embodiments of this application.

[0037] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Where there is no conflict between the embodiments, the following embodiments and features can be combined with each other. The steps in the following method embodiments are for illustrative purposes only and are not intended to limit the invention.

[0038] As described in the background section, in the dynamic optimization of process control parameters, the current state of the process control system is greatly affected by the historical state, and the calculation time of the optimization scheme cannot exceed the given time step, which makes the dynamic optimization of process control parameters more complicated.

[0039] Specifically, due to differences in process characteristics, the time-series nature of process control parameters, the real-time requirements of parameter optimization calculations, the influence of production environment factors, the high-dimensional characteristics of process control data, other uncertainties in actual production, and different customer requirements, the difficulty of solving the optimal combination of control parameters in real time is extremely high.

[0040] Currently, methods for achieving dynamic optimization of control parameters mainly include: Dynamic Matrix Control (DMC), Fuzzy Logic Control (FLC), mechanism simulation, mechanism simulation-based reinforcement learning, and model-based offline reinforcement learning. DMC, a type of advanced control, describes the process system using a linear step response model, then uses a rolling time-domain method to calculate the optimal combination of control parameters and adjusts the parameter change magnitude based on the residuals. FLC, also an advanced control method, describes the process system using prior experience or rules and provides the optimal combination of control parameters through defuzzification methods. Mechanism simulation describes the process system using simultaneous equations based on fundamental scientific principles (physics, biology, chemistry, etc.). Due to its slow computation speed, it often requires generating a surrogate model first, and then using an optimization solver to provide the optimal combination of control parameters. Mechanism simulation-based reinforcement learning describes the process system using simultaneous equations based on fundamental scientific principles, and then interacts with the simulation model to train an agent to determine the control strategy. Model-based offline reinforcement learning describes the characteristics of the process system through deep neural networks and then trains an agent that determines the control strategy using historical data.

[0041] The above-mentioned dynamic optimization schemes for control parameters mainly have the following problems:

[0042] First, most of these methods fail to take into account the uncertainties in the production process during the dynamic optimization of parameters. For example, neither the step response of DMC nor the rules of FLC consider uncertainties, which are precisely the uncertainties in the production process that cannot be explained by basic scientific principles and formulas.

[0043] Secondly, because the fundamental research on many parameter dynamic optimization schemes is not thorough enough, the optimization effects of these schemes fail to meet expectations in practical production applications. Furthermore, parameter dynamic optimization schemes involving simulation cannot handle confidential processes. Therefore, current parameter dynamic optimization schemes have significant limitations in application.

[0044] Furthermore, the step response model of DMC requires experimental measurement under shutdown conditions and cannot be executed in real time during production. Additionally, when the characteristics of a process segment shift as production progresses, the DMC algorithm model cannot be updated accordingly. The rule design for FLC requires extensive experience in both the methodology and the process; otherwise, highly unstable control problems and maintenance difficulties can easily arise. Simultaneously, the mechanistic simulation model also requires the experience of simulation experts for updating and maintenance. Therefore, most of the current dynamic parameter optimization schemes suffer from the problem of lagging model updates and maintenance.

[0045] Furthermore, aggressive optimization suggestions for process control parameters may damage equipment or process sections. While model-based offline reinforcement learning can select between aggressive or conservative suggestions, it does not take into account the real-world feedback from the process system. Therefore, if the model training is inadequate, distribution shifts in historical data can easily lead to inaccurate model predictions, resulting in equipment or process section damage.

[0046] Based on the various problems existing in the above-mentioned dynamic optimization schemes for various parameters, the embodiments of this application aim to provide a process control parameter optimization scheme that combines advanced control concepts and machine learning algorithms. By predicting the preset state data corresponding to each detection point in the process control sequence, analyzing the preset state data and actual state data of the target detection point in each detection point, and optimizing at least one process control parameter of at least one detection point to be adjusted in each detection point according to the state analysis results of the target detection point, the optimization scheme for process control parameters can be provided in real time and dynamically during the process.

[0047] The following detailed description, with reference to the accompanying drawings, describes the process control parameter optimization methods, apparatus, electronic devices, and storage media provided in the various embodiments of this application.

[0048] Process control parameter optimization methods

[0049] Figure 1 This is a flowchart of a process control parameter optimization method according to an exemplary embodiment of this application. Figure 1 As shown, the process control parameter optimization method in this embodiment includes the following steps:

[0050] Step 102: Based on the process control parameters corresponding to each detection point in the process control sequence, predict the state of the process system under test corresponding to each detection point, and obtain the predicted state data corresponding to each detection point.

[0051] The process system under test can be applied to various process control fields, including but not limited to: metal smelting, water treatment, power, energy, etc. In some embodiments, the process system under test may include at least one industrial control device.

[0052] In various embodiments of this application, process control parameters refer to controllable parameters that can be adjusted manually during the process control process, such as current parameters, voltage parameters, bearing speed, etc.

[0053] Optionally, the detection points arranged in time sequence can be determined according to the preset detection interval.

[0054] In some embodiments, the number of detection points included in the process control sequence can be arbitrarily adjusted according to the actual process control accuracy requirements. For example, a process control sequence may contain 10 or 20 detection points, but this is not a limitation. Those skilled in the art can arbitrarily adjust the number of detection points included in the process control sequence according to the actual parameter optimization requirements and equipment performance conditions, etc., and this application does not impose any restrictions on this.

[0055] Optionally, for each detection point (individual detection point) in the process control sequence, a process control parameter can be set for each detection point (e.g., only one process control parameter C1 can be set for detection point T1), or multiple process control parameters can be set for each detection point (e.g., multiple process control parameters C11, C12, and C13 can be set for detection point T1).

[0056] In some embodiments, process control parameters corresponding to each detection point in the process control sequence can be identified. For example, detection point T1 corresponds to process control parameter C1, detection point T2 corresponds to process control parameter C2, and so on. Based on the process control parameters corresponding to each detection point in the process control sequence, the states of the process system under test corresponding to each detection point are predicted, and the predicted state data corresponding to each detection point are obtained. For example, predicting the system state of the process system under test running process control parameter C1 at detection point T1 yields the preset state data FD1 (where FD is an abbreviation for Forecast Data) of the process system under test corresponding to detection point T1; predicting the system state of the process system under test running process control parameter C2 at detection point T2 yields the preset state data FD2 of the process system under test corresponding to detection point T2, and so on.

[0057] In some embodiments, a given parameter optimization algorithm can be used to optimize each process control parameter corresponding to each detection point in the process control sequence to obtain each optimized process control parameter corresponding to each detection point. Based on each optimized process control parameter corresponding to each detection point, the state of the process system under test corresponding to each detection point can be predicted to obtain each predicted state data corresponding to each detection point.

[0058] Optionally, the given parameter optimization algorithm may include, but is not limited to, the cross-entropy algorithm.

[0059] In some embodiments, a state prediction model can be used to predict the state of the process system under test corresponding to each detection point based on the process control parameters corresponding to each detection point in the process control sequence, thereby obtaining the predicted state data corresponding to each detection point.

[0060] The state prediction model can include any type of regression model. In some embodiments, the state prediction model is a sparse Gaussian process regression model.

[0061] Step 104: Based on the target detection point and the detection point to be adjusted determined in each detection point, determine the target state data corresponding to the target detection point in each predicted state data, and detect the state of the process system under test corresponding to the target detection point to obtain the actual state data of the target detection point.

[0062] Optionally, the number of target detection points and the number of detection points to be adjusted can be determined based on the actual prediction accuracy requirements. Generally, there are at least one target detection point and at least one detection point to be adjusted.

[0063] Specifically, based on at least one target detection point and at least one detection point to be adjusted determined in each detection point, at least one target state data corresponding to at least one target detection point can be determined in each predicted state data, and the state of the process system under test corresponding to at least one target detection point can be detected to obtain at least one actual state data of at least one target detection point.

[0064] In some embodiments, the first detection point among all detection points may be determined as the target detection point, and at least one detection point following the target detection point among all detection points may be determined as the detection point to be adjusted.

[0065] For example, if each detection point in the process control sequence is detection point T1 to detection point T10, and both the target detection point and the detection point to be adjusted are set to one, then detection point T1 can be determined as the target detection point, and detection point T2 can be determined as the detection point to be adjusted.

[0066] In some embodiments, based on the target detection point (e.g., detection point T1) determined in each detection point, the predicted state data corresponding to the target detection point (e.g., predicted state data FD1 corresponding to detection point T1) can be determined from the predicted state data corresponding to each detection point, and the system state of the process system under test running process control parameter C1 at detection point T1 can be detected to obtain the actual state data of the target detection point, such as the actual state data AD1 corresponding to detection point T1 (where AD is an abbreviation for Actual Data).

[0067] Step 106: Analyze the target state data and the actual state data to obtain the state analysis data of the target detection point.

[0068] In some embodiments, residual analysis can be performed on at least one target state data and at least one actual state data of at least one target detection point to obtain state analysis data of at least one target detection point.

[0069] Step 108: Optimize the process control parameters of the target detection point based on the status analysis data of the target detection point.

[0070] In some embodiments, at least one process control parameter of at least one detection point to be adjusted in the process control sequence can be adjusted based on the residual analysis value of at least one target detection point, so as to provide an optimization scheme for the process control parameter in real time and dynamically.

[0071] In some embodiments, the process control parameters of the target detection point in the process control sequence can be selectively adjusted based on the residual analysis value of the target detection point and a preset adjustment threshold. For example, if the residual analysis value of the target detection point is higher than the preset adjustment threshold, at least one process control parameter of at least one target detection point in the process control sequence is adjusted; if the residual analysis value of the target detection point is not higher than the preset adjustment threshold, at least one process control parameter of at least one target detection point in the process control sequence may not be adjusted.

[0072] In some embodiments, the higher the residual analysis value of the target detection point, the more conservative the proposed process control parameter adjustment scheme will be (e.g., the smaller the optimization adjustment range of the process control parameters).

[0073] In summary, the process control parameter optimization scheme provided in this embodiment analyzes the predicted and actual state data of the target detection points in the process control sequence, and dynamically optimizes and adjusts the process control parameters of the detection points to be adjusted in the process control sequence, thereby improving the production efficiency of the industrial process and increasing product yield. Furthermore, the process control parameter optimization scheme provided in this embodiment has a relatively small system computational load, and can quickly provide the optimal process control parameter optimization scheme in real time, meeting the real-time requirements of optimization schemes in complex process production environments.

[0074] Furthermore, the process control parameter optimization scheme provided in this embodiment utilizes a sparse Gaussian process regression model to predict the state of the process system under test corresponding to each detection point, which can provide more accurate state prediction results and thus improve the parameter optimization effect.

[0075] Furthermore, the process control parameter optimization scheme provided in this embodiment determines the first detection point in the process control sequence as the target detection point, and optimizes the process control parameters of at least one detection point to be adjusted after the first detection point based on the residual analysis results of the actual state and the predicted state of the first detection point. Therefore, the technical solution of this embodiment can provide a set of optimized control trajectories in each time step, but only executes the process control parameters of the first detection point in the trajectory at the current moment, which can well meet the real-time requirements of the process control parameter optimization scheme in the process.

[0076] Figure 2 This is a flowchart of a process control parameter optimization method according to another exemplary embodiment of this application. This embodiment shows an exemplary model training scheme for the state prediction model used to perform the above step 102.

[0077] like Figure 2 As shown, the process control parameter optimization method in this embodiment includes the following steps:

[0078] Step 202: Obtain the training sample set of the process system to be tested.

[0079] In some embodiments, the training sample set includes at least the process control parameters, environmental influencing factor parameters, and actual state data of the process system under test corresponding to each sampling point.

[0080] Optionally, the sampling points arranged in time sequence can be determined according to a preset detection interval.

[0081] It should be noted that in the various embodiments of this application, sampling points and detection points are substantially the same. Two different names are used to distinguish different stages of the state prediction model's application. Sampling points correspond to the training stage of the state prediction model to be trained, while detection points correspond to the actual application stage of the trained state prediction model.

[0082] In the various embodiments of this application, process control parameters refer to controllable parameters that can be manually adjusted in order to produce products that meet requirements during the process control flow, such as current parameters, voltage parameters, bearing speed, etc. Environmental impact factor parameters refer to conditional parameters that affect production results but cannot be directly controlled by the process control system.

[0083] In some embodiments, for any current sampling point among the sampling points, the process control parameters of the current sampling point can be determined based on the given control parameters and adjustable parameters of the current sampling point. For example, when the process control parameter is the drying temperature, the process control parameter can be determined based on the given drying temperature commitment (e.g., 55 degrees) and the adjustable temperature parameter (e.g., ±5 degrees).

[0084] In some embodiments, for any current sampling point among the sampling points, the environmental influencing factor parameters of the current sampling point can be sensed and obtained by at least one sensing device of the process system under test at the detection time corresponding to the current sampling point.

[0085] In some embodiments, the environmental impact factor parameters of the current sampling point may include at least one of the production environment detection parameters, production raw material detection parameters, and equipment parameters of the current sampling point.

[0086] For example, production environment monitoring parameters may include parameters such as ambient temperature, humidity, air flow rate, and water pressure of the process system under test. Raw material monitoring parameters may include parameters such as the type and composition of the raw materials. Equipment parameters refer to the various physical and chemical characteristics of industrial equipment during the production process, such as temperature, pressure, speed, and power. Generally, equipment parameters are determined during the equipment design and manufacturing process, and for the same type of equipment, these parameters remain constant.

[0087] Step 204: Determine the current sampling point and the target sampling point among all sampling points.

[0088] In this embodiment, the target sampling point is the sampling point that follows the current sampling point. For example, if the current sampling point is S1, the target sampling point is S2.

[0089] Step 206: Using the state prediction model to be trained, predict the state of the process system under test corresponding to the target sampling point based on the process control parameters and environmental influencing factor parameters of the process system under test corresponding to the current sampling point, and obtain the preset state data of the target sampling point.

[0090] In some embodiments, the state prediction model may include, but is not limited to, a sparse Gaussian process regression model.

[0091] Specifically, the state prediction model to be trained can be used to predict the state of the target sampling point (i.e. the next sampling point of the current sampling point) of the process system under test based on the process control parameters and environmental influencing factor parameters corresponding to the current sampling point of the process system under test, thereby obtaining the preset state data of the process system under test corresponding to the target sampling point.

[0092] Step 208: Based on the predicted state data and actual state data of the target sampling points, train the state prediction model to be trained to obtain the state prediction model.

[0093] In some embodiments, the predicted state data and actual state data of the target sampling point can be compared to obtain the model loss value of the state prediction model to be trained. Based on the model loss value, the model parameters of the state prediction model to be trained are updated. Using the updated state prediction model to be trained, the step of determining the current sampling point and the target sampling point in each sampling point (i.e., step 204) is re-executed until the model loss value meets the preset training termination condition, thereby obtaining the trained state prediction model.

[0094] For example, residual calculation can be performed based on the predicted state data and actual state data of the target sampling point to obtain the model loss value of the state prediction model to be trained.

[0095] In some embodiments, when the model loss value meets a preset convergence value, a judgment result can be obtained that the model loss value meets the preset training termination condition.

[0096] In some embodiments, when the model loss value (θ) is updated in the i-th iteration i ) and the model loss value (θ) updated in the (i-1)th iteration i-1 When the difference between the two values ​​is less than the preset difference threshold, the model loss value can be judged to meet the preset training termination condition.

[0097] In summary, the process control parameter optimization scheme provided in this embodiment uses the process control parameters, environmental influencing factor parameters, and actual state data of each sampling point as training sample sets to train the state prediction model. It takes into account various uncertainties in the actual production scenario and uses real data to perform model training, which can not only improve the accuracy of the model prediction results, but also meet the model's prediction needs under complex production conditions and improve the stability of the model's prediction performance.

[0098] Furthermore, the process control parameter optimization scheme provided in this embodiment utilizes a state prediction model to predict the state of the process system under test at the next sampling point (i.e., the target sampling point) corresponding to the current sampling point, based on the process control parameters and environmental influencing factor parameters at the current sampling point. This enables the trained state prediction model to be applicable to multi-condition switching application scenarios and to meet the real-time requirements for prediction results in actual prediction applications. Moreover, only the latest production data needs to be collected to perform model training and updates, significantly reducing the amount of training data and lowering model training costs.

[0099] Figure 3 This is a flowchart illustrating a process control parameter optimization method according to another exemplary embodiment of this application. This embodiment shows another implementation of step S102 described above, such as... Figure 3 As shown, this embodiment mainly includes the following steps:

[0100] Step 302: Determine the process control parameters corresponding to each detection point in the process control sequence as the process control parameters to be optimized for each detection point.

[0101] Specifically, before predicting the state of the process system under test corresponding to each detection point, the original process control parameters corresponding to each detection point in the process control sequence can be optimized to improve the final parameter optimization effect.

[0102] Step 304: Initialize the parameter values ​​of the parameter optimization algorithm.

[0103] In some embodiments, the parameter values ​​of the parameter optimization algorithm can be initialized according to a set initial parameter value. In another embodiment, a random initialization method can also be used to initialize the parameter values ​​of the parameter optimization algorithm.

[0104] Step 306: Based on the parameter values ​​of the parameter optimization algorithm, perform parameter optimization on each process control parameter to be optimized corresponding to each detection point to obtain each intermediate process control parameter corresponding to each detection point.

[0105] Specifically, this embodiment utilizes a parameter optimization algorithm to iteratively optimize process control parameters repeatedly. During this iterative optimization, the parameter values ​​of the parameter optimization algorithm are dynamically adjusted. Therefore, based on the current parameter values ​​of the parameter optimization algorithm, the process control parameters to be optimized at each detection point can be optimized and adjusted to obtain the intermediate process control parameters corresponding to each detection point.

[0106] Step 308: Based on the intermediate process control parameters corresponding to each detection point, predict the state of the process system under test corresponding to each detection point, and obtain the intermediate predicted state data and prediction result evaluation data corresponding to each detection point.

[0107] In some embodiments, the training sample set of the process system under test further includes optimization target parameters (or optimization indices) of the process system under test. These optimization target parameters characterize the expected or ideal production results of the process system under test.

[0108] In some embodiments, training the state prediction model further includes:

[0109] For any current sampling point among all sampling points, the prediction result evaluation data of the current sampling point is obtained based on the optimization target parameters, the process control parameters of the process system under test corresponding to the current sampling point, and the predicted state data.

[0110] In this embodiment, the prediction result evaluation data of the current sampling point is used to characterize the accuracy of the prediction state data of the current sampling point.

[0111] In some embodiments, the prediction result evaluation data of the state prediction model corresponding to the current sampling point can be obtained based on the optimized target parameters, the process control parameters of the process system under test corresponding to the current sampling point, the environmental influencing factor parameters, the predicted state data, and the actual state data.

[0112] Step 310: Update the parameter values ​​of the parameter optimization algorithm based on the evaluation data of the prediction results corresponding to each detection point, and update the intermediate process control parameters corresponding to each detection point to the process control parameters to be optimized corresponding to each detection point.

[0113] Specifically, the parameter values ​​of the parameter optimization algorithm (e.g., cross-entropy algorithm) can be updated based on the evaluation data of the prediction results corresponding to each detection point, and the intermediate process control parameters currently corresponding to each detection point can be updated to the process control parameters to be optimized corresponding to each detection point.

[0114] Step 312: Determine whether the parameter optimization meets the preset parameter optimization termination condition. If yes, proceed to step 314; otherwise, proceed to step 306.

[0115] In some embodiments, if the parameter optimization process performed on each process control parameter to be optimized corresponding to each detection point does not meet the preset parameter optimization termination condition, the process returns to step 306 to re-perform parameter optimization on each process control parameter to be optimized corresponding to each detection point based on the parameter value currently updated by the parameter optimization algorithm.

[0116] In some embodiments, when the number of parameter optimization processes performed on each process control parameter to be optimized corresponding to each detection point meets the preset iterative optimization threshold, a judgment result can be obtained that the parameter optimization meets the preset parameter optimization termination condition.

[0117] In some embodiments, the preset iteration optimization threshold can be set to 200 times, but it is not limited thereto. Those skilled in the art can adjust it according to actual optimization needs, and this application does not impose any restrictions on it.

[0118] Step 314: Obtain the optimized process control parameters for each detection point.

[0119] Specifically, the process control parameters to be optimized corresponding to each detection point can be determined as the optimized process control parameters for each detection point.

[0120] Step 316: Based on the optimized process control parameters corresponding to each detection point, predict the state of the process system under test corresponding to each detection point, and obtain the predicted state data corresponding to each detection point.

[0121] For specific implementation details of this embodiment, please refer to the description of step 102 above, which will not be repeated here.

[0122] In summary, the process control parameter optimization scheme provided in this embodiment, by combining parameter optimization algorithms and state prediction algorithms, performs multi-dimensional optimization on each process control parameter in the process control sequence, which can further improve the optimization effect of process control parameters and improve the production efficiency and product yield of the process system.

[0123] Process control parameter optimization device

[0124] Corresponding to the above method embodiments, Figure 4 A schematic diagram of a process control parameter optimization apparatus according to an embodiment of this application is shown. Figure 4 As shown, the process control parameter optimization device 400 includes:

[0125] Prediction unit 402 is used to predict the state of the process system under test corresponding to each detection point based on the process control parameters corresponding to each detection point in the process control sequence, and obtain the predicted state data corresponding to each detection point.

[0126] The determining unit 404 is configured to determine at least one target state data corresponding to the at least one target detection point in each predicted state data based on at least one target detection point and at least one detection point to be adjusted in each detection point, and detect the state of the process system under test corresponding to the at least one target detection point, and obtain at least one actual state data of the at least one target detection point.

[0127] Analysis unit 406 is used to analyze the at least one target state data and the at least one actual state data to obtain state analysis data of the at least one target detection point;

[0128] The optimization unit 408 is used to optimize at least one process control parameter of the at least one target detection point to be adjusted based on the status analysis data of the at least one target detection point.

[0129] In some embodiments, the prediction unit 402 is further configured to utilize a state prediction model to perform the step of predicting each state of the process system under test corresponding to each detection point based on each process control parameter in the process control sequence corresponding to each detection point, and obtaining each predicted state data corresponding to each detection point.

[0130] In some embodiments, the process control parameter optimization device 400 further includes a training module (not shown) for training the state prediction model, which includes: acquiring a training sample set of the process system under test, wherein the training sample set includes at least process control parameters, environmental influencing factor parameters, and actual state data of the process system under test corresponding to each sampling point; determining a current sampling point and a target sampling point among the sampling points, wherein the target sampling point is a sampling point following the current sampling point; using the state prediction model to be trained, predicting the state of the process system under test corresponding to the target sampling point based on the process control parameters and environmental influencing factor parameters of the process system under test corresponding to the current sampling point, and obtaining preset state data of the target sampling point; and training the state prediction model to be trained based on the predicted state data and actual state data of the target sampling point, thereby obtaining the state prediction model.

[0131] In some embodiments, the state prediction model includes a sparse Gaussian process regression model.

[0132] In some embodiments, for any current sampling point among the sampling points, the process control parameters of the current sampling point are determined based on the given control parameters and adjustable parameters of the current sampling point; the environmental impact factor parameters of the current sampling point are sensed and obtained by at least one sensing device of the process system under test at the detection time corresponding to the current sampling point; the environmental impact factor parameters of the current sampling point include at least one of the production environment detection parameters, production raw material detection parameters, and equipment parameters of the current sampling point.

[0133] In some embodiments, the training module is further configured to: compare the predicted state data and the actual state data of the target sampling point to obtain the model loss value of the state prediction model to be trained; update the model parameters of the state prediction model to be trained based on the model loss value; use the updated state prediction model to be trained to execute the step of determining the current sampling point and the target sampling point in each sampling point until the model loss value meets the preset training termination condition; and obtain the state prediction model.

[0134] In some embodiments, the training sample set of the process system under test further includes the optimization target parameters of the process system under test. The training module is further configured to: for any current sampling point among the sampling points, obtain the prediction result evaluation data of the current sampling point based on the optimization target parameters, the process control parameters of the process system under test corresponding to the current sampling point, and the prediction state data; wherein, the prediction result evaluation data of the current sampling point characterizes the accuracy of the prediction state data of the current sampling point.

[0135] In some embodiments, the prediction unit 402 is further configured to: determine each process control parameter corresponding to each detection point in the process control sequence as each process control parameter to be optimized corresponding to each detection point; perform at least one iterative optimization on each process control parameter to be optimized corresponding to each detection point using a given parameter optimization algorithm to obtain each optimized process control parameter corresponding to each detection point; and predict each state of the process system under test corresponding to each detection point based on each optimized process control parameter corresponding to each detection point to obtain each predicted state data corresponding to each detection point.

[0136] In some embodiments, the given parameter optimization algorithm includes the cross-entropy algorithm.

[0137] In some embodiments, the prediction unit 402 is further configured to: initialize the parameter values ​​of the parameter optimization algorithm; perform parameter optimization on each process control parameter to be optimized corresponding to each detection point based on the parameter values ​​of the parameter optimization algorithm to obtain each intermediate process control parameter corresponding to each detection point; predict each state of the process system under test corresponding to each detection point according to each intermediate process control parameter corresponding to each detection point to obtain each intermediate predicted state data and each prediction result evaluation data corresponding to each detection point; update the parameter values ​​of the parameter optimization algorithm according to each prediction result evaluation data corresponding to each detection point, update each intermediate process control parameter corresponding to each detection point to each process control parameter to be optimized corresponding to each detection point, and execute the step of performing parameter optimization on each process control parameter to be optimized corresponding to each detection point based on the parameter values ​​of the parameter optimization algorithm to obtain each intermediate process control parameter corresponding to each detection point, until the parameter optimization meets the preset parameter optimization termination condition; and obtain each optimized process control parameter for each detection point.

[0138] In some embodiments, when the number of times the parameter optimization is performed on each process control parameter to be optimized corresponding to each detection point meets a preset iterative optimization threshold, a judgment result is obtained that the parameter optimization meets the preset parameter optimization termination condition.

[0139] In some embodiments, the at least one target detection point includes the first detection point among the detection points; the at least one detection point to be adjusted includes at least one detection point following the target detection point among the detection points.

[0140] electronic devices

[0141] Figure 5 This is a schematic diagram of an electronic device provided in Embodiment 4 of this application. The specific embodiments of this application do not limit the specific implementation of the electronic device. See also... Figure 5 The electronic device 500 provided in this application embodiment includes: a processor 502, a communications interface 504, a memory 506, and a bus 508. Wherein:

[0142] The processor 502, communication interface 504, and memory 506 communicate with each other via bus 508.

[0143] Communication interface 504 is used to communicate with other electronic devices or servers.

[0144] The processor 502 is used to execute program 510, specifically the relevant steps in the above-described process control parameter optimization method embodiment.

[0145] Specifically, program 510 may include program code that includes computer operation instructions.

[0146] The processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The smart device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0147] Memory 506 is used to store program 510. Memory 506 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0148] Specifically, program 510 can be used to cause processor 502 to execute the process control parameter optimization method in any of the foregoing embodiments.

[0149] The specific implementation of each step in program 510 can be found in the corresponding steps and units described in the above-mentioned process control parameter optimization method embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the equipment and modules described above can be referred to the corresponding process descriptions in the aforementioned method embodiments, and will not be repeated here.

[0150] Computer-readable storage media

[0151] This application also provides a computer-readable storage medium storing instructions for causing a machine to perform the process control parameter optimization method as described herein. Specifically, a system or apparatus equipped with a storage medium storing software program code that implements the functions of any of the embodiments described above, and enabling the computer (or CPU or MPU) of the system or apparatus to read and execute the program code stored in the storage medium.

[0152] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of this application.

[0153] Examples of storage media used to provide program code include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), magnetic tapes, non-volatile memory cards, and ROMs. Alternatively, program code can be downloaded from a server computer via a communication network.

[0154] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0155] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0156] In this patent application, nouns and pronouns relating to people are not limited to specific genders.

[0157] In the above embodiments, the hardware modules can be implemented mechanically or electrically. For example, a hardware module may include permanent dedicated circuitry or logic (such as a dedicated processor, FPGA, or ASIC) to perform the corresponding operations. The hardware module may also include programmable logic or circuitry (such as a general-purpose processor or other programmable processor), which can be temporarily configured by software to perform the corresponding operations. The specific implementation method (mechanical, dedicated permanent circuitry, or temporarily configured circuitry) can be determined based on cost and time considerations.

[0158] The present application has been shown and described in detail above with reference to the accompanying drawings and preferred embodiments. However, the present application is not limited to these disclosed embodiments. Based on the above multiple embodiments, those skilled in the art will know that more embodiments of the present application can be obtained by combining the code review methods in the different embodiments above. These embodiments are also within the protection scope of the present application.

Claims

1. A method for optimizing process control parameters, comprising: Based on the process control parameters corresponding to each detection point in the process control sequence, predict the state of the process system under test corresponding to each detection point, and obtain the predicted state data corresponding to each detection point. Based on at least one target detection point and at least one detection point to be adjusted determined in each detection point, at least one target state data corresponding to the at least one target detection point is determined in each predicted state data, and the state of the process system under test corresponding to the at least one target detection point is detected to obtain at least one actual state data of the at least one target detection point. Analyze the at least one target state data and the at least one actual state data to obtain state analysis data for the at least one target detection point; Based on the status analysis data of the at least one target detection point, optimize at least one process control parameter of the at least one detection point to be adjusted.

2. The method of claim 1, wherein, The method includes: Using a state prediction model, the steps of predicting the state of the process system under test corresponding to each detection point based on the process control parameters corresponding to each detection point in the process control sequence are executed, and the predicted state data corresponding to each detection point are obtained. Furthermore, the state prediction model is trained in the following manner: Obtain a training sample set for the process system under test, wherein the training sample set includes at least the process control parameters, environmental influencing factor parameters, and actual state data of the process system under test corresponding to each sampling point; In each sampling point, a current sampling point and a target sampling point are determined, wherein the target sampling point is a sampling point that follows the current sampling point in each sampling point; Using a state prediction model to be trained, the state of the process system under test corresponding to the target sampling point is predicted based on the process control parameters and environmental influencing factor parameters of the process system under test corresponding to the current sampling point, thereby obtaining the preset state data of the target sampling point. Based on the predicted state data and actual state data of the target sampling points, the state prediction model to be trained is trained to obtain the state prediction model.

3. The method of claim 2, wherein, The state prediction model includes a sparse Gaussian process regression model.

4. The method according to claim 2, wherein, For any current sampling point among all sampling points The process control parameters of the current sampling point are determined based on the given control parameters and adjustable parameters of the current sampling point. The environmental influencing factor parameters of the current sampling point are obtained by at least one sensing device of the process system under test at the detection time corresponding to the current sampling point. The environmental impact factor parameters of the current sampling point include at least one of the production environment detection parameters, production raw material detection parameters, and equipment parameters of the current sampling point.

5. The method of claim 2, wherein, The step of training the state prediction model to be trained based on the predicted state data and actual state data of the target sampling point, and obtaining the state prediction model, includes: By comparing the predicted state data and the actual state data of the target sampling points, the model loss value of the state prediction model to be trained is obtained. Based on the model loss value, update the model parameters of the state prediction model to be trained; Using the updated state prediction model to be trained, the step of determining the current sampling point and the target sampling point in each sampling point is performed until the model loss value meets the preset training termination condition. The state prediction model is obtained.

6. The method of claim 2, wherein, The training sample set of the process system under test also includes the optimization target parameters of the process system under test. Furthermore, the training of the state prediction model also includes: For any current sampling point among all sampling points Based on the optimized target parameters, the process control parameters of the process system under test corresponding to the current sampling point, and the predicted state data, the predicted result evaluation data of the current sampling point is obtained. The prediction result evaluation data of the current sampling point represents the accuracy of the prediction state data of the current sampling point.

7. The method according to claim 1 or 6, wherein, The step of predicting the state of the process system under test corresponding to each detection point based on the process control parameters corresponding to each detection point in the process control sequence, and obtaining the predicted state data corresponding to each detection point, includes: Each process control parameter corresponding to each detection point in the process control sequence is determined as the process control parameter to be optimized for each detection point. Using the given parameter optimization algorithm, perform at least one iterative optimization on each process control parameter to be optimized corresponding to each detection point to obtain each optimized process control parameter corresponding to each detection point. Based on the optimized process control parameters corresponding to each detection point, the states of the process system under test corresponding to each detection point are predicted, and the predicted state data corresponding to each detection point are obtained.

8. The method of claim 7, wherein, The given parameter optimization algorithm includes the cross-entropy algorithm.

9. The method of claim 7, wherein, The step of using a given parameter optimization algorithm to perform at least one iterative optimization on each process control parameter to be optimized corresponding to each detection point, to obtain each optimized process control parameter corresponding to each detection point, includes: Initialize the parameter values ​​of the parameter optimization algorithm; Based on the parameter values ​​of the parameter optimization algorithm, parameter optimization is performed on each process control parameter to be optimized corresponding to each detection point to obtain each intermediate process control parameter corresponding to each detection point. Based on the intermediate process control parameters corresponding to each detection point, predict the state of the process system under test corresponding to each detection point, and obtain the intermediate predicted state data and prediction result evaluation data corresponding to each detection point. The parameter values ​​of the parameter optimization algorithm are updated based on the evaluation data of the prediction results corresponding to each detection point. The intermediate process control parameters corresponding to each detection point are updated to the process control parameters to be optimized corresponding to each detection point. The parameter optimization algorithm is then executed to optimize the process control parameters to be optimized corresponding to each detection point, and the intermediate process control parameters corresponding to each detection point are obtained. This process continues until the parameter optimization meets the preset parameter optimization termination condition. Obtain the optimized process control parameters for each detection point.

10. The method according to claim 9, wherein, The determination result that the parameter optimization meets the preset parameter optimization termination condition is obtained through the following method: When the number of times the parameter optimization process is performed on each process control parameter to be optimized corresponding to each detection point meets the preset iterative optimization threshold, a judgment result is obtained that the parameter optimization meets the preset parameter optimization termination condition.

11. The method according to claim 1, wherein, The at least one target detection point includes the first detection point among all detection points; The at least one detection point to be adjusted includes at least one detection point that follows the target detection point among all detection points.

12. A process control parameter optimization device, comprising: The prediction unit is used to predict the state of the process system under test corresponding to each detection point based on the process control parameters corresponding to each detection point in the process control sequence, and to obtain the predicted state data corresponding to each detection point. The determining unit is configured to determine at least one target state data corresponding to the at least one target detection point in each predicted state data based on at least one target detection point and at least one detection point to be adjusted in each detection point, and detect the state of the process system under test corresponding to the at least one target detection point, thereby obtaining at least one actual state data of the at least one target detection point. Analysis unit, used to analyze the at least one target state data and the at least one actual state data to obtain state analysis data of the at least one target detection point; An optimization unit is used to optimize at least one process control parameter of the at least one target detection point to be adjusted based on the status analysis data of the at least one target detection point.

13. An electronic device, comprising: The processor, the communication interface, the memory, and the bus are connected, and the processor, the communication interface, and the memory communicate with each other via the bus. The memory is used to store at least one executable instruction that causes the processor to perform an operation corresponding to any one of the methods described in claims 1 to 11.

14. A computer-readable storage medium storing computer instructions that, when executed by a processor, cause the processor to perform the method of any one of claims 1 to 11.

Images