Self-adaptive pressure regulation vacuum pump closed-loop control system and method

Abstract

The invention relates to a self-adaptive pressure regulation vacuum pump closed-loop control system and method. The system comprises a model building module which is used for acquiring real-time data flow of a vacuum pump and building a nonlinear dynamic model to obtain a pressure-flow-power feature mapping relation; the predictive analysis module calculates a pressure-flow predictive value based on the relationship and compares the pressure-flow predictive value with the real-time power data for analysis. And generating a parameter adjustment instruction once the deviation value exceeds a preset threshold value. And the control scheme generation module corrects the control parameters according to the parameter adjustment instruction and generates an optimized power adjustment scheme. And if the data is abnormal, generating a state update value. And the control scheme updating module fuses the state updating value and the pressure-flow prediction value to obtain an optimized model parameter so as to adjust a weight coefficient of an adaptive control strategy and generate a control scheme. The system effectively improves the accuracy of vacuum pump pressure control, greatly enhances the ability of the system to deal with abnormal conditions and equipment performance changes, and ensures long-term stable operation of the vacuum pump.

Description

Technical Field

[0001] The present invention belongs to the technical field of vacuum pump control, and in particular relates to a vacuum pump closed-loop control system and method with adaptive pressure regulation. Background Art

[0002] With the development of vacuum pump control technology, a closed-loop control technique for vacuum pumps with adaptive pressure regulation has emerged. Vacuum pumps are widely used in industrial production and scientific research. From the high-precision vacuum requirements of semiconductor manufacturing to maintaining reduced pressure environments during chemical reactions, to simulating specific vacuum conditions in scientific research experiments, stable operation of vacuum pumps is essential. However, existing vacuum pump control systems suffer from numerous drawbacks. Firstly, they struggle to effectively cope with complex and changing operating conditions. Under conditions such as frequent load fluctuations and ambient temperature variations, they cannot accurately maintain the required pressure, leading to production process defects and increased errors in experimental results. Secondly, most traditional systems utilize open-loop control, lacking real-time feedback and adjustment capabilities for the actual operating status of the vacuum pump. This makes it impossible to promptly compensate for performance degradation caused by equipment aging and component wear. This not only shortens the vacuum pump's service life, but also significantly increases operating costs and makes maintenance increasingly difficult. Summary of the Invention

[0003] Based on this, it is necessary to provide a vacuum pump closed-loop control system and method with adaptive pressure regulation that can effectively improve the accuracy of vacuum pump pressure control and enhance the ability to cope with abnormal situations and equipment performance changes in order to address the above technical problems.

[0004] In a first aspect, the present application provides a vacuum pump closed-loop control system with adaptive pressure regulation, comprising:

[0005] The model building module is used to obtain the real-time data stream of pressure, flow and power when the vacuum pump is running; feature extraction and key parameter nonlinear dynamic model construction are performed on the real-time data stream to obtain the characteristic mapping relationship of pressure-flow-power.

[0006] The prediction and analysis module is used to obtain training data based on the feature mapping relationship and use the support vector regression algorithm to calculate the pressure-flow prediction value under different working conditions and compare and analyze it with the real-time power data. If the deviation value exceeds the preset threshold, a parameter adjustment instruction is obtained.

[0007] The control scheme generation module is used to modify the control parameters using the adaptive control strategy according to the parameter adjustment instruction and generate an optimized power regulation scheme; if there are abnormal values ​​in the data when the optimized power regulation scheme is executed, the self-check calibration mechanism is triggered to generate a status update value.

[0008] The control scheme update module is used to fuse the state update value with the pressure-flow prediction value, and use the Kalman filter algorithm to iteratively optimize the nonlinear dynamic model to obtain the optimized model parameters; it is also used to adjust the weight coefficient of the adaptive control strategy according to the optimized model parameters to generate an updated control scheme; when the control scheme reaches the preset stability standard, the feature mapping relationship is updated at a fixed period.

[0009] In one embodiment, feature extraction and key parameter nonlinear dynamic model construction are performed on the real-time data stream to obtain a characteristic mapping relationship between pressure, flow and power, including:

[0010] Nonlinear weight factors are extracted from the real-time data stream to construct a nonlinear dynamic model; the nonlinear dynamic model includes pressure fluctuation rate, flow change gradient and power response threshold.

[0011] The parameter coupling coefficient between the pressure fluctuation rate and the flow change gradient is calculated using the nonlinear regression algorithm based on the nonlinear dynamic model.

[0012] The parameter coupling coefficient is subjected to latent variable correlation analysis to obtain the variable mapping relationship between the power response threshold and the parameter coupling coefficient.

[0013] A time-varying covariance matrix is ​​generated according to the variable mapping relationship; the time-varying covariance matrix is ​​used to update the residual sequence of the nonlinear dynamic model.

[0014] The nonlinear weight factor is adjusted using an iterative optimization algorithm based on the time-varying covariance matrix to obtain the characteristic mapping relationship of pressure-flow-power.

[0015] In one embodiment, the parameter coupling coefficient between the pressure fluctuation rate and the flow change gradient is calculated using a nonlinear regression algorithm based on a nonlinear dynamic model, including:

[0016] The pressure fluctuation rate and flow rate change gradient data in the nonlinear dynamic model are preprocessed to obtain normalized data suitable for nonlinear regression analysis.

[0017] The preprocessed normalized data are used to perform parameter estimation using a nonlinear regression algorithm to obtain parameter estimates of the pressure fluctuation rate and flow rate gradient.

[0018] The parameter estimates are calculated using the following formula

[0019]

[0020] Among them, PV represents the pressure fluctuation rate, FG represents the flow gradient, and Z i Indicates other possible influencing factors in the nonlinear regression model, β0,β1,...,βn+2 represents the parameter to be estimated, ∈ represents the random error term, J(β) represents the objective function of the least squares method to estimate the parameter, m represents the number of samples, PV j , FG j and Z ij are the pressure fluctuation rate, flow rate change gradient and other influencing factors of the jth sample, Represents the objective function J(β) with respect to each parameter β k (k=0,1,...,n+2) calculate partial derivatives and solve the normal equations to get parameter estimates

[0021] The obtained parameter estimation value is calculated using the coupling coefficient formula to obtain the parameter coupling coefficient between the pressure fluctuation rate and the flow change gradient.

[0022]

[0023] Where CZ represents the parameter coupling coefficient, Representation parameters The derivative of the flow change gradient FG indicates other possible influencing factors.

[0024] In one embodiment, the predictive analysis module further includes:

[0025] The working condition parameters and sensor signals in the feature mapping relationship are obtained to generate a multi-dimensional feature vector.

[0026] The historical operating condition database is matched according to the multi-dimensional feature vector to extract the pressure-flow sample set under the corresponding operating condition.

[0027] The grid search algorithm is used to optimize the hyperparameters of the support vector regression model and generate the optimized regression kernel function.

[0028] The pressure-flow sample set is input into the optimized regression kernel function to calculate the pressure-flow prediction matrix.

[0029] Based on the mapping relationship between the pressure-flow prediction matrix and the real-time operating parameters, dynamic residual analysis results are generated.

[0030] Update the pressure-flow sample weight coefficients in the historical operating condition database based on the dynamic residual analysis results.

[0031] The distribution space of the multidimensional feature vector is reconstructed according to the updated pressure-flow sample weight coefficients.

[0032] The sliding window mechanism is used to intercept the high-frequency feature intervals in the distribution space and generate incremental training data blocks.

[0033] The incremental training data block is input into the optimized regression kernel function, and the updated pressure-flow prediction value is output.

[0034] In one embodiment, the state update value and the pressure-flow prediction value are data-fused, and the nonlinear dynamic model is iteratively optimized using the Kalman filter algorithm to obtain optimized model parameters, including:

[0035] The covariance matrix is ​​calculated based on the residuals between the state update value and the pressure-flow prediction value.

[0036] The covariance matrix is ​​used to update the Kalman gain matrix, and the Kalman gain matrix is ​​used to correct the dynamic equation coefficients of the nonlinear dynamic model to obtain the correction amount of the dynamic equation coefficients; the correction amount of the dynamic equation coefficients is determined by the product of the gain matrix and the residual.

[0037] The correction amount is judged based on a preset threshold. If the correction amount is less than the preset threshold, the optimized model parameters are generated. Otherwise, the corrected dynamic equation coefficients are used as input to recalculate the pressure-flow prediction value.

[0038] In one embodiment, adjusting the weight coefficients of the adaptive control strategy according to the optimized model parameters to generate an updated control scheme includes:

[0039] Obtain real-time environmental variable data in the optimized model parameters; the real-time environmental variable data includes the equipment operating status and external disturbance characteristics.

[0040] A dynamic feature matrix is ​​extracted according to the real-time environmental variable data, and the dynamic feature data in the dynamic feature matrix is ​​normalized to obtain a normalized feature vector.

[0041] The normalized eigenvector is compared with the preset stability index threshold. If the normalized eigenvector exceeds the stability index threshold, the weight correction mechanism is triggered to generate a weight coefficient update instruction.

[0042] The weight coefficient update instruction is input into the adaptive control strategy execution queue and the response data of the target device is monitored. The mapping rules of the dynamic characteristic matrix are iteratively updated according to the response data to generate an updated control scheme.

[0043] In a second aspect, the present application further provides a closed-loop control method for a vacuum pump with adaptive pressure regulation, the method comprising:

[0044] The real-time data stream of pressure, flow and power is obtained when the vacuum pump is running; the feature extraction and key parameter nonlinear dynamic model of the real-time data stream are carried out to obtain the characteristic mapping relationship of pressure-flow-power.

[0045] Based on the feature mapping relationship, training data is obtained and the support vector regression algorithm is used to calculate the pressure-flow prediction value under different working conditions and compare and analyze it with the real-time power data. If the deviation value exceeds the preset threshold, a parameter adjustment instruction is obtained.

[0046] According to the parameter adjustment instruction, the adaptive control strategy is used to correct the control parameters and generate an optimized power regulation scheme; if there are abnormal values ​​in the data when the optimized power regulation scheme is executed, the self-check calibration mechanism is triggered to generate a status update value.

[0047] The state update value and the pressure-flow prediction value are fused, and the Kalman filter algorithm is used to iteratively optimize the nonlinear dynamic model to obtain the optimized model parameters. The weight coefficients of the adaptive control strategy are adjusted according to the optimized model parameters to generate an updated control scheme. When the control scheme reaches the preset stability standard, the feature mapping relationship is updated at a fixed period.

[0048] In a third aspect, the present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above system when executing the computer program.

[0049] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above system is implemented.

[0050] In the above-mentioned closed-loop control system and method for adaptive pressure regulation of a vacuum pump, the model building module acquires real-time data streams of pressure, flow, and power during vacuum pump operation. By extracting features from the data, it constructs a nonlinear dynamic model of key parameters, thereby obtaining a characteristic mapping relationship between pressure, flow, and power. The prediction and analysis module obtains training data based on this characteristic mapping relationship and uses a support vector regression algorithm to calculate pressure-flow predictions under different operating conditions. These predictions are then compared and analyzed with the real-time power data. If the deviation exceeds a preset threshold, a parameter adjustment instruction is generated. Based on the received parameter adjustment instruction, the control scheme generation module uses an adaptive control strategy to modify control parameters and generate an optimized power regulation plan. During the execution of this plan, if the data contains abnormal values, a self-checking and calibration mechanism is triggered to generate a state update value. The control scheme update module fuses the state update value with the pressure-flow prediction value and iteratively optimizes the nonlinear dynamic model using a Kalman filter algorithm. The optimized model parameters are then used to adjust the weight coefficients of the adaptive control strategy and generate an updated control plan. When the control plan reaches a preset stability standard, the characteristic mapping relationship is updated at a fixed period to ensure that the system remains adaptable to changes in the vacuum pump's operating state. Through real-time data monitoring, precise prediction and adaptive adjustment, the accuracy of vacuum pump pressure control can be effectively improved, the system's ability to respond to abnormal situations and changes in equipment performance can be greatly enhanced, the risk of system failure can be reduced, energy utilization efficiency can be improved, operating costs can be reduced, and the long-term stable operation of the vacuum pump can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A structural block diagram of a vacuum pump closed-loop control system with adaptive pressure regulation provided by an embodiment of the present invention;

[0053] Figure 2 A flowchart of the characteristic mapping relationship between pressure, flow rate and power obtained by extracting features from real-time data streams and constructing a nonlinear dynamic model of key parameters provided by an embodiment of the present invention;

[0054] Figure 3 A flowchart of an embodiment of the present invention for fusing state update values ​​with pressure-flow prediction values ​​and iteratively optimizing a nonlinear dynamic model using a Kalman filter algorithm to obtain optimized model parameters;

[0055] Figure 4A flowchart of a closed-loop control method for a vacuum pump with adaptive pressure regulation provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0057] In one embodiment, Figure 1 As shown, the present application provides a vacuum pump closed-loop control system with adaptive pressure regulation, which may include:

[0058] The model building module 101 is used to obtain the real-time data stream of pressure, flow and power when the vacuum pump is running; perform feature extraction on the real-time data stream and build a nonlinear dynamic model of key parameters to obtain a characteristic mapping relationship between pressure, flow and power.

[0059] Specifically, a high-precision sensor network accurately collects real-time data streams of pressure, flow, and power during vacuum pump operation. This data contains a wealth of information about the vacuum pump's operation. Advanced signal processing and machine learning algorithms are used to perform in-depth feature extraction on the real-time data stream. For example, by calculating the statistical and frequency characteristics of the data, key parameters that reflect the essence of the vacuum pump's operating state are discovered. Based on these key parameters, complex nonlinear modeling techniques such as neural networks and deep learning models are used to construct a nonlinear dynamic model that closely matches the actual operating laws of the vacuum pump. Through continuous iterative training and optimization of model parameters, a characteristic mapping relationship between pressure, flow, and power was ultimately successfully obtained. This relationship accurately depicts the inherent connection between pressure, flow, and power of the vacuum pump under different operating conditions.

[0060] The prediction and analysis module 102 is used to obtain training data based on the feature mapping relationship and use the support vector regression algorithm to calculate the pressure-flow prediction value under different working conditions and compare and analyze it with the real-time power data. If the deviation value exceeds the preset threshold, a parameter adjustment instruction is obtained.

[0061] Specifically, based on the pressure-flow-power characteristic mapping relationship, representative training data is screened out. These training data cover a variety of typical operating conditions of vacuum pumps, providing rich samples for the training of subsequent prediction models. With the help of the powerful nonlinear fitting ability of the support vector regression algorithm, this module accurately predicts the pressure-flow under different working conditions. During the actual operation process, the theoretical power corresponding to the predicted pressure-flow value is carefully compared and analyzed with the power data collected in real time. By setting a reasonable deviation threshold, once the power deviation value exceeds the preset threshold, it is immediately determined that the operating status of the vacuum pump is abnormal. At this time, the prediction and analysis module quickly generates parameter adjustment instructions to adjust the operating parameters of the vacuum pump to ensure its stable and efficient operation.

[0062] The control scheme generating module 103 is used to modify the control parameters using the adaptive control strategy according to the parameter adjustment instruction and generate an optimized power regulation scheme; if there are abnormal values ​​in the data when the optimized power regulation scheme is executed, the self-checking calibration mechanism is triggered to generate a state update value.

[0063] Upon receiving the parameter adjustment instruction, the control scheme generation module quickly initiates the adaptive control strategy. This strategy is based on an understanding of the operating principles and characteristics of the vacuum pump, and combines real-time operating data to intelligently correct control parameters such as the speed of the vacuum pump motor and the valve opening. By precisely adjusting these key control parameters, an optimized power regulation scheme is generated, enabling the vacuum pump to operate in a more optimal state under the new operating conditions. During the execution of the optimized power regulation scheme, the control scheme generation module constantly monitors various types of data. Once abnormal values ​​are found in the data, such as sudden changes in sensor readings or abnormal actuator feedback, the self-test and calibration mechanism is immediately triggered. A comprehensive inspection and calibration of all parts of the equipment is carried out to quickly generate status update values, providing an important basis for the stable operation of the system and subsequent optimization of the control scheme.

[0064] The control scheme update module 104 is used to fuse the state update value with the pressure-flow prediction value, and use the Kalman filter algorithm to iteratively optimize the nonlinear dynamic model to obtain optimized model parameters; it is also used to adjust the weight coefficient of the adaptive control strategy according to the optimized model parameters to generate an updated control scheme; when the control scheme reaches the preset stability standard, the feature mapping relationship is updated at a fixed period.

[0065] Specifically, the state update value generated after the self-check calibration mechanism is triggered is deeply fused with the obtained pressure-flow prediction value. On this basis, the Kalman filter algorithm is used to iteratively optimize the nonlinear dynamic model. The Kalman filter algorithm can accurately estimate the system state based on the system's dynamic characteristics and measurement noise, and continuously correct the model parameters to obtain more accurate optimized model parameters. Based on these optimized model parameters, the weight coefficients of the adaptive control strategy are intelligently adjusted to generate an updated control scheme. At the same time, it continuously monitors the execution effect of the control scheme. When the control scheme reaches the preset stability standard, the characteristic mapping relationship between pressure, flow, and power is updated at a fixed period. This enables the system to closely follow the changes in the operating state of the vacuum pump, continuously optimize the control strategy, and always maintain the optimal operating state, effectively improving the system's adaptability, stability, and reliability.

[0066] In the aforementioned closed-loop control system for a vacuum pump with adaptive pressure regulation, the model building module acquires real-time data streams of pressure, flow, and power during vacuum pump operation. By extracting features from the data, it constructs a nonlinear dynamic model of key parameters, thereby obtaining a characteristic mapping relationship between pressure, flow, and power. The prediction and analysis module obtains training data based on this characteristic mapping relationship and uses a support vector regression algorithm to calculate predicted pressure-flow values ​​under different operating conditions. These values ​​are then compared and analyzed with the real-time power data. If the deviation exceeds a preset threshold, a parameter adjustment instruction is generated. The control scheme generation module, based on the received parameter adjustment instruction, uses an adaptive control strategy to modify control parameters and generate an optimized power regulation plan. During the execution of this plan, if the data contains abnormal values, a self-calibration mechanism is triggered to generate a state update value. The control scheme update module fuses the state update value with the predicted pressure-flow value and iteratively optimizes the nonlinear dynamic model using a Kalman filter algorithm. The optimized model parameters are then used to adjust the weight coefficients of the adaptive control strategy and generate an updated control plan. When the control plan reaches a preset stability standard, the characteristic mapping relationship is updated periodically to ensure that the system remains adaptable to changes in the vacuum pump's operating state. Through real-time data monitoring, precise prediction and adaptive adjustment, the accuracy of vacuum pump pressure control can be effectively improved, the system's ability to respond to abnormal situations and changes in equipment performance can be greatly enhanced, the risk of system failure can be reduced, energy utilization efficiency can be improved, operating costs can be reduced, and the long-term stable operation of the vacuum pump can be guaranteed.

[0067] In one embodiment, Figure 2 As shown, feature extraction of real-time data stream and construction of nonlinear dynamic model of key parameters are performed to obtain the characteristic mapping relationship of pressure-flow-power, which may include the following steps:

[0068] Step S201 , extracting nonlinear weight factors from the real-time data stream to construct a nonlinear dynamic model; the nonlinear dynamic model includes pressure fluctuation rate, flow change gradient and power response threshold.

[0069] Step S202 : calculating the parameter coupling coefficient between the pressure fluctuation rate and the flow rate change gradient using a nonlinear regression algorithm according to the nonlinear dynamic model.

[0070] Step S203 : performing latent variable correlation analysis on the parameter coupling coefficient to obtain a variable mapping relationship between the power response threshold and the parameter coupling coefficient.

[0071] Step S204 , generating a time-varying covariance matrix according to the variable mapping relationship; the time-varying covariance matrix is ​​used to update the residual sequence of the nonlinear dynamic model.

[0072] Step S205 : adjusting the nonlinear weight factor using an iterative optimization algorithm based on the time-varying covariance matrix to obtain a characteristic mapping relationship of pressure-flow-power.

[0073] Specifically, the nonlinear weight factors are first extracted from the real-time data stream generated during vacuum pump operation. Based on this, a nonlinear dynamic model is constructed that includes pressure fluctuation rate, flow rate gradient, and power response threshold. Next, a nonlinear regression algorithm is applied to this model to calculate the parameter coupling coefficient between the pressure fluctuation rate and the flow rate gradient. Subsequently, a latent variable correlation analysis is performed on the obtained parameter coupling coefficient to obtain a variable mapping relationship between the power response threshold and the parameter coupling coefficient. Based on this variable mapping relationship, a time-varying covariance matrix is ​​generated, which is used to update the residual sequence of the nonlinear dynamic model. Finally, based on the time-varying covariance matrix, the nonlinear weight factors are adjusted with the help of an iterative optimization algorithm, ultimately obtaining a characteristic mapping relationship between pressure, flow, and power.

[0074] Through complex algorithm processing and multi-step data mining, this embodiment can deeply analyze the intrinsic relationship between the operating parameters of the vacuum pump, construct a nonlinear dynamic model that is highly consistent with the actual operating conditions, and improve the accuracy of the description of the operating status of the vacuum pump. In terms of prediction and control, the characteristic mapping relationship between pressure, flow, and power provides a strong basis for predicting parameter changes under different working conditions, enabling operators to make decisions in advance, optimize the operation control of the vacuum pump, and effectively avoid equipment failures or production accidents caused by parameter anomalies. At the same time, the update of the residual sequence by the time-varying covariance matrix and the adjustment of the nonlinear weight factor by the iterative optimization algorithm enhance the adaptive ability of the model, enabling it to continuously optimize over time and with changes in working conditions, thereby improving the overall stability and reliability of the system.

[0075] In one embodiment, calculating the parameter coupling coefficient between the pressure fluctuation rate and the flow rate change gradient using a nonlinear regression algorithm according to a nonlinear dynamic model may include the following steps:

[0076] Step S301 : Preprocessing the pressure fluctuation rate and flow rate change gradient data in the nonlinear dynamic model to obtain normalized data suitable for nonlinear regression analysis.

[0077] Step S302 : performing parameter estimation on the pre-processed normalized data using a nonlinear regression algorithm to obtain parameter estimation values ​​of the pressure fluctuation rate and the flow rate change gradient.

[0078] The parameter estimates are calculated using the following formula

[0079]

[0080] Among them, PV represents the pressure fluctuation rate, FG represents the flow gradient, and Z i Indicates other possible influencing factors in the nonlinear regression model, β0,β1,...,β n+2 represents the parameter to be estimated, ∈ represents the random error term, J(β) represents the objective function of the least squares method to estimate the parameter, m represents the number of samples, PV j , FG j and Z ij are the pressure fluctuation rate, flow rate change gradient and other influencing factors of the jth sample, Represents the objective function J(β) with respect to each parameter β k (k=0,1,...,n+2) calculate partial derivatives and solve the normal equations to get parameter estimates

[0081] Step S303 : Calculate the parameter coupling coefficient between the pressure fluctuation rate and the flow rate change gradient using the coupling coefficient formula based on the obtained parameter estimation value.

[0082]

[0083] Where CZ represents the parameter coupling coefficient, Representation parameters The derivative of the flow change gradient FG indicates other possible influencing factors.

[0084] This embodiment utilizes a complex nonlinear regression algorithm and least squares method for model construction and parameter solution, comprehensively considering multiple influencing factors. This method accurately captures the complex relationship between pressure fluctuation rate and flow gradient, resulting in precise parameter estimates and parameter coupling coefficients. This is crucial for gaining a deeper understanding of the inherent connection between pressure and flow during vacuum pump operation, helping to optimize vacuum pump operation control strategies and improve their efficiency and stability. This method has significant application value in industrial production, scientific research, and other fields involving vacuum pump applications, effectively improving the performance and reliability of related systems, reducing resource waste, and lowering production costs.

[0085] In one embodiment, the prediction analysis module may further include:

[0086] Step S401: Acquire the operating parameters and sensor signals in the feature mapping relationship to generate a multi-dimensional feature vector.

[0087] Step S402 : matching the historical operating condition database according to the multi-dimensional feature vector to extract the pressure-flow sample set under the corresponding operating condition.

[0088] Step S403 , using a grid search algorithm to optimize the hyperparameters of the support vector regression model to generate an optimized regression kernel function.

[0089] Step S404: Input the pressure-flow sample set into the optimized regression kernel function to calculate and obtain a pressure-flow prediction matrix.

[0090] Step S405 : generating a dynamic residual analysis result according to the mapping relationship between the pressure-flow prediction matrix and the real-time operating condition parameters.

[0091] Step S406: updating the pressure-flow sample weight coefficient in the historical operating condition database based on the dynamic residual analysis result.

[0092] Step S407: reconstruct the distribution space of the multi-dimensional feature vector according to the updated pressure-flow sample weight coefficient.

[0093] Step S408: Using a sliding window mechanism, intercept high-frequency feature intervals in the distribution space to generate incremental training data blocks.

[0094] Step S409: input the incremental training data block into the optimized regression kernel function, and output the updated pressure-flow prediction value.

[0095] Specifically, the operating parameters and sensor signals are first extracted from the feature mapping relationship and integrated to generate a multidimensional feature vector. This multidimensional feature vector contains rich operating condition information, providing a foundation for subsequent analysis. Next, this multidimensional feature vector is matched against a historical operating condition database to extract a set of pressure-flow samples corresponding to the operating condition. Subsequently, a grid search algorithm is used to optimize the hyperparameters of the support vector regression model, generating an optimized regression kernel function that better fits the data. The extracted pressure-flow sample set is input into the optimized regression kernel function to calculate a pressure-flow prediction matrix. By analyzing the mapping relationship between the pressure-flow prediction matrix and the real-time operating condition parameters, a dynamic residual analysis result is generated, which reflects the difference between the predicted and actual values. Based on this dynamic residual analysis result, the weight coefficients of the pressure-flow samples in the historical operating condition database are updated to reflect the importance of different samples under the current operating conditions. Based on the updated weight coefficients, the distribution space of the multidimensional feature vector is reconstructed to better reflect the actual operating conditions. A sliding window mechanism is used to intercept high-frequency feature intervals in the distribution space and generate incremental training data blocks. These data blocks contain the latest and most representative feature information. Finally, the incremental training data blocks are input into the optimized regression kernel function, which outputs the updated pressure-flow prediction value.

[0096] This embodiment fully exploits the information of historical and real-time data through steps such as multi-dimensional feature vector matching, hyperparameter optimization, and dynamic residual analysis. It can more accurately capture the changing patterns of pressure-flow and improve the accuracy and reliability of predictions. Dynamic residual analysis is used to update sample weight coefficients, reconstruct the distribution space, and generate incremental training data blocks. This enables the model to automatically adjust and optimize as operating conditions change. It has good adaptive and self-learning capabilities and can effectively cope with complex and changing operating conditions. This helps to improve the operational stability and efficiency of related systems, provide more reliable decision-making basis for industrial production, equipment control and other fields, and reduce production costs and risks.

[0097] In one embodiment, Figure 3 As shown, the state update value and the pressure-flow prediction value are data-fused, and the nonlinear dynamic model is iteratively optimized using the Kalman filter algorithm to obtain the optimized model parameters, which can include the following steps:

[0098] Step S501 : obtaining a covariance matrix based on the residual between the state update value and the pressure-flow prediction value.

[0099] Step S502, using the covariance matrix to update the Kalman gain matrix, and using the Kalman gain matrix to correct the dynamic equation coefficients of the nonlinear dynamic model to obtain the correction amount of the dynamic equation coefficients; the dynamic equation coefficient correction amount is determined by the product of the gain matrix and the residual.

[0100] Step S503 , judging the correction amount based on a preset threshold, generating optimized model parameters if the correction amount is less than the preset threshold, otherwise recalculating the pressure-flow prediction value using the corrected dynamic equation coefficients as input.

[0101] First, a covariance matrix is ​​calculated based on the residuals between the updated state values ​​and the pressure-flow prediction values. This covariance matrix reflects the distribution characteristics of the residuals and the correlations between variables. Next, the resulting covariance matrix is ​​used to update the Kalman gain matrix. The Kalman gain matrix plays a key role in the entire system, balancing the weights of the predicted and measured values. The updated Kalman gain matrix is ​​used to correct the dynamic equation coefficients of the nonlinear dynamic model, thereby obtaining a correction to the dynamic equation coefficients. This correction is determined by multiplying the gain matrix with the residuals. The correction is then evaluated based on a pre-set threshold. If the correction is less than the preset threshold, it indicates that the current adjustment of the nonlinear dynamic model has reached a relatively ideal state, and the optimized model parameters are generated. Conversely, if the correction is greater than the preset threshold, the corrected dynamic equation coefficients are used as new inputs to recalculate the pressure-flow prediction values ​​for further model optimization.

[0102] By calculating the covariance matrix through residuals and updating the Kalman gain matrix and correcting the dynamic equation coefficients accordingly, dynamic optimization of the nonlinear dynamic model is achieved, which can more accurately capture the dynamic changes of the system, improve the prediction accuracy of the model and its adaptability to complex working conditions. The judgment mechanism of the preset threshold ensures the rationality and stability of the model adjustment. The optimization parameters are generated only when the correction amount reaches a certain standard, avoiding model instability caused by excessive adjustment and enhancing the operational reliability of the system under different working conditions. In actual application scenarios, such as equipment monitoring and control in industrial production processes, this technical solution can effectively improve the prediction accuracy of key parameters, promptly discover potential problems in equipment operation, and provide strong support for optimizing production processes and ensuring safe and stable operation of equipment, thereby reducing production costs and improving production efficiency and product quality.

[0103] In one embodiment, adjusting the weight coefficients of the adaptive control strategy according to the optimized model parameters to generate an updated control solution may include the following steps:

[0104] Step S601: Acquire real-time environmental variable data in the optimized model parameters; the real-time environmental variable data includes the equipment operation status and external disturbance characteristics.

[0105] Step S602 : extracting a dynamic feature matrix based on the real-time environmental variable data, and performing normalization processing on the dynamic feature data in the dynamic feature matrix to obtain a normalized feature vector.

[0106] In step S603, the normalized eigenvector is compared with a preset stability index threshold. If the normalized eigenvector exceeds the stability index threshold, the weight correction mechanism is triggered to generate a weight coefficient update instruction.

[0107] Step S604: input the weight coefficient update instruction into the adaptive control strategy execution queue and monitor the response data of the target device, iteratively update the mapping rule of the dynamic characteristic matrix according to the response data, and generate an updated control solution.

[0108] Specifically, real-time environmental variable data is first acquired from the optimized model parameters. This data covers key information such as the device's operating status and external disturbance characteristics. Next, a dynamic feature matrix is ​​extracted based on the acquired real-time environmental variable data, providing a data foundation for comprehensively reflecting the system's current state. The dynamic feature data in the dynamic feature matrix are normalized to obtain a normalized feature vector. Subsequently, the normalized feature vector is compared with a pre-set stability indicator threshold. If the normalized feature vector exceeds the stability indicator threshold, indicating that the current system state may be unstable, the weight correction mechanism is triggered, generating a weight coefficient update instruction. This instruction is entered into the adaptive control strategy execution queue, while the response data of the target device is closely monitored. Based on this response data, the mapping rules of the dynamic feature matrix are iteratively updated to generate an updated control scheme that is more suitable for the current system state.

[0109] In terms of system stability, this embodiment, through real-time monitoring of environmental variable data and comparison with stability indicator thresholds, can promptly detect potential system instability factors, trigger the weight correction mechanism, and update the control scheme, effectively ensuring the stable operation of the system in complex and changing environments, and reducing the risk of system failure and performance degradation caused by unstable factors. From an adaptability perspective, iteratively updating the mapping rules of the dynamic feature matrix based on the target device response data enables the control scheme to continuously adapt to changes in the device operating status and external environment, improving the system's adaptability and flexibility, and ensuring that the system maintains good performance under different operating conditions.

[0110] In one embodiment, Figure 4 As shown, the present application also provides a vacuum pump closed-loop control method with adaptive pressure regulation, the method comprising:

[0111] Step S701, obtaining the real-time data stream of pressure, flow and power when the vacuum pump is running; performing feature extraction and key parameter nonlinear dynamic model construction on the real-time data stream to obtain the characteristic mapping relationship of pressure-flow-power.

[0112] Step S702 , based on the feature mapping relationship, training data is obtained and the pressure-flow prediction value under different working conditions is calculated using the support vector regression algorithm and compared with the real-time power data for analysis. If the deviation value exceeds the preset threshold, a parameter adjustment instruction is obtained.

[0113] Step S703: According to the parameter adjustment instruction, the control parameters are modified using the adaptive control strategy to generate an optimized power adjustment scheme; if abnormal values ​​exist in the data when the optimized power adjustment scheme is executed, the self-checking and calibration mechanism is triggered to generate a status update value.

[0114] In step S704, the state update value and the pressure-flow prediction value are data-fused, and the nonlinear dynamic model is iteratively optimized using the Kalman filter algorithm to obtain optimized model parameters; the weight coefficient of the adaptive control strategy is adjusted according to the optimized model parameters to generate an updated control scheme; when the control scheme reaches the preset stability standard, the feature mapping relationship is updated at a fixed period.

[0115] The above-mentioned closed-loop control method for a vacuum pump with adaptive pressure regulation first obtains real-time data streams of pressure, flow, and power during vacuum pump operation. In-depth feature extraction is performed on this data, and a nonlinear dynamic model of key parameters is constructed. This method then generates a feature mapping relationship that reflects the complex relationship between pressure, flow, and power. Based on this mapping relationship, training data is obtained. A support vector regression algorithm is used to calculate pressure-flow predictions under different operating conditions, which are then compared and analyzed with real-time power data. If the power deviation exceeds a preset threshold, a parameter adjustment instruction is immediately generated. Based on this instruction, an adaptive control strategy is used to modify the control parameters and generate an optimized power regulation solution. During the execution of the optimization solution, if an outlier value is detected in the data, a self-checking and calibration mechanism is triggered to generate a state update value. The state update value is then fused with the pressure-flow prediction value, and the nonlinear dynamic model is iteratively optimized using a Kalman filter algorithm to obtain more accurate model parameters. The weight coefficients of the adaptive control strategy are then adjusted based on these optimized model parameters to generate an updated control solution. When the control solution reaches a preset stability standard, the feature mapping relationship is updated periodically, enabling the system to better adapt to dynamic changes in the vacuum pump's operating state. Through real-time data monitoring, precise prediction and adaptive adjustment, the accuracy of vacuum pump pressure control can be effectively improved, the system's ability to respond to abnormal situations and changes in equipment performance can be greatly enhanced, the risk of system failure can be reduced, energy utilization efficiency can be improved, operating costs can be reduced, and the long-term stable operation of the vacuum pump can be guaranteed.

[0116] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0117] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the aforementioned adaptive pressure-regulated vacuum pump closed-loop control system and method are implemented.

[0118] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0119] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0120] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A vacuum pump closed-loop control system with adaptive pressure regulation, characterized in that: The system comprises: A model building module is used to obtain real-time data streams of pressure, flow and power when the vacuum pump is running; perform feature extraction and key parameter nonlinear dynamic model construction on the real-time data stream to obtain a characteristic mapping relationship between pressure, flow and power; A prediction and analysis module is used to obtain training data based on the feature mapping relationship and use a support vector regression algorithm to calculate the pressure-flow prediction value under different working conditions and compare and analyze it with the real-time power data. If the deviation value exceeds a preset threshold, a parameter adjustment instruction is obtained; a control scheme generating module, configured to modify control parameters using an adaptive control strategy according to the parameter adjustment instruction to generate an optimized power regulation scheme; and trigger a self-checking and calibration mechanism to generate a state update value if abnormal values ​​are found in the data when the optimized power regulation scheme is executed; A control scheme update module is used to perform data fusion on the state update value and the pressure-flow prediction value, and iteratively optimize the nonlinear dynamic model using the Kalman filter algorithm to obtain optimized model parameters; it is also used to adjust the weight coefficient of the adaptive control strategy according to the optimized model parameters to generate an updated control scheme; when the control scheme reaches a preset stability standard, the feature mapping relationship is updated at a fixed period.

2. The system according to claim 1, wherein: The feature extraction and key parameter nonlinear dynamic model construction of the real-time data stream to obtain the characteristic mapping relationship of pressure-flow-power include: Extracting nonlinear weight factors from the real-time data stream to construct a nonlinear dynamic model; the nonlinear dynamic model includes pressure fluctuation rate, flow change gradient, and power response threshold; A parameter coupling coefficient between the pressure fluctuation rate and the flow rate change gradient is calculated using a nonlinear regression algorithm according to the nonlinear dynamic model; Performing latent variable correlation analysis on the parameter coupling coefficient to obtain a variable mapping relationship between the power response threshold and the parameter coupling coefficient; Generating a time-varying covariance matrix according to the variable mapping relationship; the time-varying covariance matrix is ​​used to update the residual sequence of the nonlinear dynamic model; The nonlinear weight factor is adjusted using an iterative optimization algorithm based on the time-varying covariance matrix to obtain a characteristic mapping relationship of pressure-flow-power.

3. The system according to claim 2, characterized in that The parameter coupling coefficient between the pressure fluctuation rate and the flow change gradient is calculated using a nonlinear regression algorithm according to the nonlinear dynamic model, including: Performing data preprocessing on the pressure fluctuation rate and flow rate change gradient data in the nonlinear dynamic model to obtain normalized data suitable for nonlinear regression analysis; Performing parameter estimation on the pre-processed normalized data using a nonlinear regression algorithm to obtain parameter estimation values ​​of the pressure fluctuation rate and the flow rate change gradient; The parameter estimates are calculated using the following formula Among them, PV represents the pressure fluctuation rate, FG represents the flow gradient, and Z i Indicates other possible influencing factors in the nonlinear regression model, β0,β1,...,β n+2 represents the parameter to be estimated, ∈ represents the random error term, J(β) represents the objective function of the least squares method to estimate the parameter, m represents the number of samples, PV j , FG j and Z ij are the pressure fluctuation rate, flow rate change gradient and other influencing factors of the jth sample, Represents the objective function J(β) with respect to each parameter β k (k=0,1,...,n+2) calculate partial derivatives and solve the normal equations to get parameter estimates Calculating the obtained parameter estimation value using a coupling coefficient formula to obtain a parameter coupling coefficient between the pressure fluctuation rate and the flow rate change gradient; Where CZ represents the parameter coupling coefficient, Representation parameters The derivative of the flow change gradient FG indicates other possible influencing factors.

4. The system according to claim 1, wherein: The prediction analysis module also includes: Acquire the operating parameters and sensor signals in the feature mapping relationship to generate a multidimensional feature vector; Matching a historical operating condition database according to the multidimensional feature vector to extract a pressure-flow sample set under corresponding operating conditions; Optimizing hyperparameters of the support vector regression model using a grid search algorithm to generate an optimized regression kernel function; Inputting the pressure-flow sample set into the optimized regression kernel function to calculate a pressure-flow prediction matrix; Generate dynamic residual analysis results based on the mapping relationship between the pressure-flow prediction matrix and the real-time operating parameters; updating the pressure-flow sample weight coefficient in the historical operating condition database based on the dynamic residual analysis result; Reconstructing the distribution space of the multidimensional feature vector according to the updated pressure-flow sample weight coefficient; Using a sliding window mechanism to intercept high-frequency feature intervals in the distribution space to generate incremental training data blocks; The incremental training data block is input into the optimized regression kernel function, and an updated pressure-flow prediction value is output.

5. The system according to claim 1, wherein: The step of fusing the state update value with the pressure-flow prediction value and iteratively optimizing the nonlinear dynamic model using a Kalman filter algorithm to obtain optimized model parameters includes: A covariance matrix is ​​obtained based on the residual calculation between the state update value and the pressure-flow prediction value; Using the covariance matrix to update the Kalman gain matrix, and using the Kalman gain matrix to correct the dynamic equation coefficients of the nonlinear dynamic model to obtain corrections to the dynamic equation coefficients; the dynamic equation coefficient corrections are determined by multiplying the gain matrix by the residual; The correction amount is judged based on a preset threshold value. If the correction amount is less than the preset threshold value, the optimized model parameters are generated. Otherwise, the corrected dynamic equation coefficient is used as input to recalculate the pressure-flow prediction value.

6. The system according to claim 1, wherein: The step of adjusting the weight coefficient of the adaptive control strategy according to the optimized model parameters to generate an updated control scheme includes: Acquiring real-time environmental variable data in the optimized model parameters; the real-time environmental variable data includes equipment operating status and external disturbance characteristics; Extracting a dynamic feature matrix according to the real-time environmental variable data, and normalizing the dynamic feature data in the dynamic feature matrix to obtain a normalized feature vector; Comparing the normalized eigenvector with a preset stability index threshold, and if the normalized eigenvector exceeds the stability index threshold, triggering a weight correction mechanism to generate a weight coefficient update instruction; The weight coefficient update instruction is input into the adaptive control strategy execution queue and the response data of the target device is monitored. The mapping rule of the dynamic characteristic matrix is ​​iteratively updated according to the response data to generate an updated control scheme.

7. A closed-loop control method for a vacuum pump with adaptive pressure regulation, characterized in that: The method comprises: Acquire real-time data streams of pressure, flow, and power during the operation of the vacuum pump; perform feature extraction and key parameter nonlinear dynamic model construction on the real-time data streams to obtain a characteristic mapping relationship between pressure, flow, and power; Based on the feature mapping relationship, training data is obtained and the pressure-flow prediction value under different working conditions is calculated using a support vector regression algorithm and compared with the real-time power data for analysis. If the deviation value exceeds a preset threshold, a parameter adjustment instruction is obtained; According to the parameter adjustment instruction, the control parameters are modified using the adaptive control strategy to generate an optimized power regulation scheme; if abnormal values ​​are found in the data when the optimized power regulation scheme is executed, a self-checking and calibration mechanism is triggered to generate a state update value; The state update value and the pressure-flow prediction value are data-fused, and the nonlinear dynamic model is iteratively optimized using the Kalman filter algorithm to obtain optimized model parameters; the weight coefficient of the adaptive control strategy is adjusted according to the optimized model parameters to generate an updated control scheme; when the control scheme reaches a preset stability standard, the feature mapping relationship is updated at a fixed period.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the system according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the system according to any one of claims 1 to 6 are implemented.

Images