A pressure control system with adjustable pressure

Abstract

The application discloses a kind of adjustable pressure pressure control systems, system includes: data acquisition module, data pre-processing module, constructs pressure state detection model module, pressure state detection model parameter optimization module and pressure control strategy generation module.The application belongs to the field of pressure control technology, specifically refers to a kind of adjustable pressure pressure control system, the scheme uses correlation coefficient and multi-head attention mechanism, and utilizes residual connection to add spatial context information, obtains global spatial pressure feature matrix, respectively calculates local space and time pressure feature and splices, obtains local pressure feature, introduces position coding and multi-head self-attention mechanism to calculate global space-time pressure feature;Based on the division probability, the individual is divided into excellent, general and poor individuals, three different updating methods are designed to update the position of the individual, the optimal model parameters are determined, the accuracy of the pressure state detection is improved, and the overall effect and intelligent level of the pressure control system are improved.

Description

Technical Field

[0001] This invention belongs to the field of pressure control technology, specifically referring to an adjustable pressure control system. Background Technology

[0002] Pressure control systems utilize artificial intelligence algorithms to monitor pressure status in real time, enabling precise adjustment and control of pressure, thereby enhancing system stability and safety. However, existing pressure control systems suffer from problems such as complex pressure data, incomplete feature extraction, and an inability to accurately capture complex correlations and dynamic changes within the pressure data, leading to inaccurate pressure status detection. Furthermore, existing pressure control systems suffer from difficulties in accurately optimizing model parameters and low search efficiency, resulting in inaccurate pressure control strategies. Summary of the Invention

[0003] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an adjustable pressure control system. Addressing the problems of complex pressure data, incomplete feature extraction, and inability to accurately capture complex correlations and dynamic changes in pressure data, leading to inaccurate pressure state detection in existing pressure control systems, this solution constructs a pressure map, uses correlation coefficients and a multi-head attention mechanism, and adds spatial context information using residual connections to obtain a global spatial pressure feature matrix. Local spatial and temporal pressure features are calculated separately and concatenated to obtain local pressure features. Position encoding and a multi-head self-attention mechanism are introduced to calculate global spatiotemporal pressure features, completing pressure state detection. This effectively integrates information from both spatial and temporal dimensions, improving the accuracy of pressure state detection. The accuracy and reliability of the system enable it to better adapt to complex and ever-changing pressure environments and demands. Addressing the issues of inaccurate pressure control strategies caused by difficulties in precisely optimizing model parameters and low search efficiency in existing pressure control systems, this solution categorizes individuals into excellent, average, and poor based on probability partitioning. It introduces random angles and dynamic sensitivity to design a position update method for excellent individuals, control factors and dynamic convergence factors to design a position update method for average individuals, and weighted coefficients and Gaussian functions to design a position update method for poor individuals. This determines the optimal model parameters, improving the efficiency and accuracy of model parameter optimization. This contributes to more precise and reliable pressure state detection, thereby enhancing the overall performance and intelligence level of the pressure control system.

[0004] The present invention provides an adjustable pressure control system, comprising a data acquisition module, a data preprocessing module, a pressure state detection model construction module, a pressure state detection model parameter optimization module, and a pressure control strategy generation module;

[0005] The data acquisition module collects historical pressure data;

[0006] The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data.

[0007] The stress state detection model module constructs a stress map, uses correlation coefficients and multi-head attention mechanism, and adds spatial context information using residual connections to obtain a global spatial stress feature matrix. It calculates and concatenates local spatial and temporal stress features to obtain local stress features. It introduces position encoding and multi-head self-attention mechanism to calculate global spatiotemporal stress features to complete stress state detection.

[0008] The pressure state detection model parameter optimization module divides individuals into excellent, average, and poor individuals based on the partition probability. It introduces random angle and dynamic sensitivity to design a position update method for excellent individuals, introduces control factors and dynamic convergence factors to design a position update method for average individuals, and introduces weighting coefficients and Gaussian functions to design a position update method for poor individuals, thereby determining the optimal model parameters.

[0009] The pressure control strategy generation module understands the current pressure status based on the output labels of the pressure status detection model and formulates a pressure control strategy.

[0010] Furthermore, the data acquisition module collects historical pressure data; the historical pressure data includes timestamps, pressure values, environmental data, equipment operating status data, and pressure status; the pressure status is used as a data label.

[0011] Furthermore, the data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data; data cleaning includes handling missing values, outliers, and duplicate values; data transformation converts the data into vector form; data normalization unifies the data range based on the min-max normalization method; and dataset construction builds a stress detection dataset based on the processed data.

[0012] Furthermore, the module for constructing the pressure state detection model specifically includes the following:

[0013] A stress graph is generated by treating each data point in the stress detection dataset as a node and using the data information of each node as its feature vector. A similarity threshold is pre-set. If the cosine similarity between two data points is greater than the similarity threshold, an edge is added between the corresponding nodes; otherwise, no edge is added. The final result is a node feature matrix G and a stress graph O = (F, δ, D), where F is the set of nodes, δ is the set of edges, and D is the adjacency matrix. Elements in D represent the connectivity between nodes; if there is an edge between two nodes, the element value is 1, otherwise the element value is 0.

[0014] Extract global spatial pressure features, including the following:

[0015] The correlation coefficient between nodes is calculated using the following formula:

[0016] ;

[0017] ;

[0018] In the formula, W is the correlation coefficient between the i-th and j-th nodes, where i and j are node indices. LeakyReLU() is the modified linear unit activation function with leakage, || is the concatenation operation, and W... xb It is the linear transformation weight matrix, g i and g j These are the feature vectors of the i-th and j-th nodes, respectively. and They are g i and g j After linear transformation, the eigenvectors are β, which is the weight vector, and T is the vector transpose operation.

[0019] The attention weights between nodes are calculated using the following formula:

[0020] ;

[0021] In the formula, It is the attention weight between the i-th and j-th nodes;

[0022] The global spatial pressure feature matrix is ​​calculated, attention is embedded in a multi-head manner, and spatial context information is added using residual connections. Through multi-head and residual connections, spatial features are captured more comprehensively. The formulas used are as follows:

[0023] ;

[0024] In the formula, G is the global spatial pressure feature matrix, G is the node feature matrix, Head is the number of attention heads, head is the index of the attention head, and ReLU(·) is the ReLU activation function.

[0025] Local pressure features are extracted by calculating the local spatial pressure feature matrix and the local temporal pressure feature matrix, respectively, and then concatenating the two matrices to obtain the local pressure feature matrix. This includes the following:

[0026] The local spatial pressure feature matrix is ​​calculated to better extract the pressure features in the local space. The formula used is as follows:

[0027] ;

[0028] ;

[0029] In the formula, It is the local spatial pressure feature matrix, and Sigmoid(•) is the Sigmoid activation function, E lap It is the Laplace matrix, W gc It is a learnable matrix, I is the identity matrix, and A is the degree matrix of D+I;

[0030] The local time pressure feature matrix is ​​calculated to accurately capture the pressure change characteristics over time. The formula used is as follows:

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] In the formula, u t m t and n t These are the reset gate, update gate, and candidate hidden state at time step t, respectively, where t is the time step index, and b is the time step index. u b m and b n They are u t m t and n t The bias, B t-1 Let be the hidden state at time step (t-1), GC(·) is the graph convolution operation, and tanh(·) is the hyperbolic tangent activation function. and These are the local time pressure feature matrices extracted at the t-th and t-1-th time steps, respectively; It is the global spatial pressure feature matrix extracted at the t-th time step;

[0036] Extract global spatiotemporal pressure features, based on the local pressure feature matrix Calculate the global spatiotemporal pressure feature matrix, including the following:

[0037] Location encoding, in the local pressure feature matrix Positional encoding information is embedded to enhance the model's ability to perceive temporal and spatial locations. The formula used is as follows:

[0038] ;

[0039] ;

[0040] In the formula, It is the local stress feature matrix after embedding positional encoding information, where x is the feature dimension index and w is the local stress feature matrix. The number of feature dimensions, It is location-coded information;

[0041] Calculate the global spatiotemporal pressure feature matrix and use a multi-head self-attention mechanism from The global spatiotemporal pressure characteristics were calculated; considering the characteristics of different spatiotemporal locations, the formula used is as follows:

[0042] ;

[0043] ;

[0044] ;

[0045] In the formula, Q head K head and V head These are the query vector, key vector, and value vector corresponding to the head-th attention head, respectively. , and These are the query weight matrix, key weight matrix, and value weight matrix corresponding to the head-th attention head. U is the self-attention score of the head-th attention head. 1 It is the global spatiotemporal pressure feature matrix, and Softmax(·) is the normalization exponential function;

[0046] Stress state detection involves processing the global spatiotemporal stress feature matrix into a normalization layer, a feedforward neural layer, and an output layer to obtain the stress state. The loss function of the model is then derived based on the mean squared error, using the following formula:

[0047] ;

[0048] ;

[0049] ;

[0050] In the formula, C and These are the outputs of the normalization layer and the feedforward neural layer, respectively. This is the detection result output by the stress state detection model. LayerNorm is the normalization layer. W1 and W2 are the weight matrices of the first and second linear transformations of the feedforward neural layer, respectively. b1 and b2 are the biases of the first and second linear transformations of the feedforward neural layer, respectively. y and b y These are the weight matrix and bias of the output layer, respectively.

[0051] Furthermore, the pressure state detection model parameter optimization module specifically includes the following:

[0052] Initialize individual positions based on the linear transformation weight matrix W in the pressure state detection model. xb Weight vector β, learnable matrix W gc Query weight matrix Key weight matrix Value weight matrix The weight matrix W of the output layer y The search space is established by the weight matrix W1 of the first linear transformation and the weight matrix W2 of the second linear transformation of the feedforward neural layer. The individual position is used as the representative of the stress state detection model parameters. The loss function of the stress state detection model established based on the model parameters is used as the fitness value of the individual. N individual positions are randomly initialized in the search space.

[0053] Before each position update, the probability of classifying an individual (P) is calculated. If the probability P is greater than 0.7, the individual is classified as an excellent individual, and the dynamic sensitivity position update method is used to update the individual's position. If the probability P is greater than 0.7, the individual is classified as an excellent individual. If the probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position; if the classification probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position. If the individual is classified as a poor performer, a distribution estimation update method is used to update the individual's position. This detailed classification allows for different update strategies to be applied to different types of individuals, improving optimization efficiency and effectiveness. The formula used to calculate the classification probability of an individual is as follows:

[0054] ;

[0055] In the formula, and These represent the partition probability and position of the v-th individual at the h-th iteration, where h is the current iteration number, H is the maximum iteration number, and v is the individual index. It is the fitness value of the v-th individual in the h-th iteration. It is the relative ranking of the v-th individual among N individuals after the fitness value of the v-th individual is increased by 1 in the h-th iteration, and rank() is the ascending ranking function;

[0056] Individual location updates include the following:

[0057] A dynamic sensitivity position update method is proposed, which introduces a random angle θ and a dynamic sensitivity design to update the position of top performers, enabling them to move more flexibly in the search space. The formulas used are as follows:

[0058] ;

[0059] ;

[0060] In the formula, yes The position updated using the dynamic sensitivity position update method. It is the position of the a1-th excellent individual in the h-th iteration, and s(h) is the sensitivity in the h-th iteration. max That is the maximum sensitivity. It is the position of the globally optimal individual at the h-th iteration. The globally optimal individual is the individual with the minimum fitness value. rand() is a random number generation function. The random angle θ ranges from 0 to 2π.

[0061] A dynamic convergence position update method is proposed, which introduces a control factor and a dynamic convergence factor to design the position update method for general individuals, making the update of general individuals more targeted. The formula used is as follows:

[0062] ;

[0063] ;

[0064] In the formula, yes The position is updated using the dynamic convergence position update method. It is the position of the a2th general individual in the h-th iteration, d min and d max These are the minimum and maximum values ​​of the convergence factor, respectively. d(h) is the convergence factor at the h-th iteration. r1 and r2 are the first and second random numbers in the range (0, 1), respectively. k is the control factor.

[0065] The distribution estimation position update method introduces weighted coefficients and a Gaussian function to design a position update mechanism for poorly performing individuals; it guides poorly performing individuals to move in the direction where the best individual will emerge; the formula used is as follows:

[0066] ;

[0067] ;

[0068] ;

[0069] ;

[0070] In the formula, yes The updated position using the distribution estimation position update method. It represents the position of the a3th worst individual in the h-th iteration, where xmean(h) and cov(h) are the weighted average and weighted covariance in the h-th iteration, respectively, and ω is the position of the a3th worst individual. v is the weighting coefficient of the v-th individual, and Gaussian function (·) is the Gaussian function;

[0071] Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the global best individual are updated. When the fitness value corresponding to the global best individual is lower than the fitness threshold, the model parameters corresponding to the position of the global best individual are the optimal model parameters, and a stress state detection model is constructed based on the optimal model parameters. Otherwise, if the maximum number of iterations is reached, the individual positions are re-initialized; otherwise, the iteration count is incremented by 1 and individuals are re-divided for individual position updates.

[0072] Furthermore, the pressure control strategy generation module collects real-time pressure data; the real-time pressure data includes timestamps, pressure values, environmental data, and equipment operating status data. After preprocessing the real-time pressure data, it is input into a pressure status detection model constructed based on optimal model parameters. Based on the output labels of the model, the current pressure status is understood, and a pressure control strategy is formulated.

[0073] The beneficial effects achieved by adopting the above solution are as follows:

[0074] (1) In view of the problems in existing pressure control systems, such as complex pressure data, incomplete feature extraction, inability to accurately capture complex correlations and dynamic changes in pressure data, resulting in inaccurate pressure state detection, this scheme constructs a pressure map, uses correlation coefficient and multi-head attention mechanism, and adds spatial context information by using residual connection to obtain a global spatial pressure feature matrix. Local spatial and temporal pressure features are calculated and spliced ​​to obtain local pressure features. Position encoding and multi-head self-attention mechanism are introduced to calculate global spatiotemporal pressure features to complete pressure state detection. This effectively integrates information in spatial and temporal dimensions, captures the features and relationships of pressure data more comprehensively and accurately, improves the accuracy and reliability of pressure state detection, provides strong support for the precise control of pressure control systems, and can better adapt to complex and ever-changing pressure environments and needs.

[0075] (2) In view of the problems of inaccurate pressure control strategies caused by the difficulty in accurately optimizing model parameters and low search efficiency in existing pressure control systems, this scheme divides individuals into excellent, average and poor individuals based on the partition probability. It introduces random angle and dynamic sensitivity to design the position update method for excellent individuals, introduces control factor and dynamic convergence factor to design the position update method for average individuals, and introduces weighted coefficient and Gaussian function to design the position update method for poor individuals. The optimal model parameters are determined, the efficiency and accuracy of model parameter optimization are improved, the targeting and adaptability of optimization are enhanced, and it can better adapt to different situations. This helps to achieve more accurate and reliable pressure state detection, thereby improving the overall effect and intelligence level of the pressure control system. Attached Figure Description

[0076] Figure 1 A schematic diagram of an adjustable pressure control system provided by the present invention;

[0077] Figure 2 A schematic diagram of the module for constructing a pressure state detection model;

[0078] Figure 3 This is a schematic diagram of the parameter optimization module for the pressure state detection model.

[0079] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0080] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0081] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0082] Example 1, see Figure 1 The present invention provides an adjustable pressure control system, comprising a data acquisition module, a data preprocessing module, a pressure state detection model construction module, a pressure state detection model parameter optimization module, and a pressure control strategy generation module;

[0083] The data acquisition module collects historical pressure data and sends the data to the data preprocessing module;

[0084] The data preprocessing module receives data sent by the data acquisition module, performs data cleaning, data transformation, data normalization and dataset construction on the acquired data, and sends the data to the stress state detection model construction module.

[0085] The stress state detection model construction module receives data sent by the data preprocessing module, constructs a stress map, uses correlation coefficients and multi-head attention mechanism, and adds spatial context information using residual connections to obtain a global spatial stress feature matrix. It calculates and concatenates local spatial and temporal stress features to obtain local stress features, introduces position encoding and multi-head self-attention mechanism to calculate global spatiotemporal stress features, completes stress state detection, and sends the data to the stress state detection model parameter optimization module.

[0086] The pressure state detection model parameter optimization module receives data sent by the pressure state detection model construction module, classifies individuals into excellent, average, and poor individuals based on the partition probability, introduces random angle and dynamic sensitivity to design a position update method for excellent individuals, introduces control factor and dynamic convergence factor to design a position update method for average individuals, and introduces weighting coefficient and Gaussian function to design a position update method for poor individuals, determines the optimal model parameters, and sends the data to the pressure control strategy generation module.

[0087] The pressure control strategy generation module receives data sent by the pressure state detection model parameter optimization module, understands the current pressure state based on the output label of the pressure state detection model, and formulates a pressure control strategy.

[0088] Example 2, see Figure 1 This embodiment is based on the above embodiment. The data acquisition module collects historical pressure data. The historical pressure data includes timestamps, pressure values, environmental data, equipment operating status data, and pressure status. The pressure status is used as a data label.

[0089] Example 3, see Figure 1 This embodiment is based on the above embodiment. The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data. Data cleaning includes handling missing values, outliers, and duplicate values. Data transformation converts the data into vector form. Data normalization unifies the data range based on the min-max normalization method. Dataset construction builds a stress detection dataset based on the processed data.

[0090] Example 4, see Figure 1 and Figure 2This embodiment, based on the above embodiments, specifically includes the following contents in constructing the pressure state detection model module:

[0091] A stress graph is generated by treating each data point in the stress detection dataset as a node and using the data information of each node as its feature vector. A similarity threshold is pre-set. If the cosine similarity between two data points is greater than the similarity threshold, an edge is added between the corresponding nodes; otherwise, no edge is added. The final result is a node feature matrix G and a stress graph O = (F, δ, D), where F is the set of nodes, δ is the set of edges, and D is the adjacency matrix. Elements in D represent the connectivity between nodes; if there is an edge between two nodes, the element value is 1, otherwise the element value is 0.

[0092] Extract global spatial pressure features, including the following:

[0093] The correlation coefficient between nodes is calculated to effectively capture the degree of association between them. The formula used is as follows:

[0094] ;

[0095] ;

[0096] In the formula, W is the correlation coefficient between the i-th and j-th nodes, where i and j are node indices. LeakyReLU() is the modified linear unit activation function with leakage, || is the concatenation operation, and W... xb It is the linear transformation weight matrix, g i and g j These are the feature vectors of the i-th and j-th nodes, respectively. and They are g i and g j After linear transformation, the eigenvectors are β, which is the weight vector, and T is the vector transpose operation.

[0097] The attention weights between nodes are calculated to highlight important node relationships, using the following formula:

[0098] ;

[0099] In the formula, It is the attention weight between the i-th and j-th nodes;

[0100] The global spatial pressure feature matrix is ​​calculated, attention is embedded in a multi-head manner, and spatial context information is added using residual connections. Through multi-head and residual connections, spatial features are captured more comprehensively, improving the model's expressive power. The formulas used are as follows:

[0101] ;

[0102] In the formula, G is the global spatial pressure feature matrix, G is the node feature matrix, Head is the number of attention heads, head is the index of the attention head, and ReLU(·) is the ReLU activation function.

[0103] Local pressure features are extracted by calculating the local spatial pressure feature matrix and the local temporal pressure feature matrix, respectively, and then concatenating the two matrices to obtain the local pressure feature matrix. This includes the following:

[0104] The local spatial pressure feature matrix is ​​calculated to better extract the pressure features in the local space. The formula used is as follows:

[0105] ;

[0106] ;

[0107] In the formula, It is the local spatial pressure feature matrix, and Sigmoid(•) is the Sigmoid activation function, E lap It is the Laplace matrix, W gc It is a learnable matrix, I is the identity matrix, and A is the degree matrix of D+I;

[0108] The local time pressure feature matrix is ​​calculated to accurately capture the pressure change characteristics over time. The formula used is as follows:

[0109] ;

[0110] ;

[0111] ;

[0112] ;

[0113] In the formula, u t m t and n t These are the reset gate, update gate, and candidate hidden state at time step t, respectively, where t is the time step index, and b is the time step index. u b m and b n They are u t m t and n t The bias, B t-1 Let be the hidden state at time step (t-1), GC(·) is the graph convolution operation, and tanh(·) is the hyperbolic tangent activation function. and These are the local time pressure feature matrices extracted at the t-th and t-1-th time steps, respectively; It is the global spatial pressure feature matrix extracted at the t-th time step;

[0114] Extract global spatiotemporal pressure features, based on the local pressure feature matrix Calculate the global spatiotemporal pressure feature matrix, including the following:

[0115] Location encoding, in the local pressure feature matrix Positional encoding information is embedded to enhance the model's ability to perceive temporal and spatial locations. The formula used is as follows:

[0116] ;

[0117] ;

[0118] In the formula, It is the local stress feature matrix after embedding positional encoding information, where x is the feature dimension index and w is the local stress feature matrix. The number of feature dimensions, It is location-coded information;

[0119] Calculate the global spatiotemporal pressure feature matrix and use a multi-head self-attention mechanism from The global spatiotemporal pressure characteristics are calculated; by comprehensively considering the characteristics of different spatiotemporal locations, the accuracy of the model is improved; the formulas used are as follows:

[0120] ;

[0121] ;

[0122] ;

[0123] In the formula, Q head K head and V head These are the query vector, key vector, and value vector corresponding to the head-th attention head, respectively. , and These are the query weight matrix, key weight matrix, and value weight matrix corresponding to the head-th attention head. U is the self-attention score of the head-th attention head. 1 It is the global spatiotemporal pressure feature matrix, and Softmax(·) is the normalization exponential function;

[0124] Stress state detection involves processing the global spatiotemporal stress feature matrix into a normalization layer, a feedforward neural layer, and an output layer to obtain the stress state. The loss function of the model is then derived based on the mean squared error, using the following formula:

[0125] ;

[0126] ;

[0127] ;

[0128] In the formula, C and These are the outputs of the normalization layer and the feedforward neural layer, respectively. This is the detection result output by the stress state detection model. LayerNorm is the normalization layer. W1 and W2 are the weight matrices of the first and second linear transformations of the feedforward neural layer, respectively. b1 and b2 are the biases of the first and second linear transformations of the feedforward neural layer, respectively. y and b y These are the weight matrix and bias of the output layer, respectively.

[0129] By performing the above operations, this solution addresses the problems in existing pressure control systems, such as complex pressure data, incomplete feature extraction, and inability to accurately capture complex correlations and dynamic changes in pressure data, leading to inaccurate pressure state detection. It constructs a pressure map, uses correlation coefficients and a multi-head attention mechanism, and adds spatial context information using residual connections to obtain a global spatial pressure feature matrix. Local spatial and temporal pressure features are calculated and concatenated to obtain local pressure features. Position encoding and a multi-head self-attention mechanism are introduced to calculate global spatiotemporal pressure features, completing pressure state detection. This effectively integrates information from both spatial and temporal dimensions, more comprehensively and accurately capturing the characteristics and relationships of pressure data, improving the accuracy and reliability of pressure state detection, and providing strong support for the precise control of pressure control systems. This allows for better adaptation to complex and ever-changing pressure environments and demands.

[0130] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the pressure state detection model parameter optimization module specifically includes the following:

[0131] Initialize individual positions based on the linear transformation weight matrix W in the pressure state detection model. xb Weight vector β, learnable matrix W gc Query weight matrix Key weight matrix Value weight matrix The weight matrix W of the output layer yThe search space is established by the weight matrix W1 of the first linear transformation and the weight matrix W2 of the second linear transformation of the feedforward neural layer. The individual position is used as the representative of the stress state detection model parameters. The loss function of the stress state detection model established based on the model parameters is used as the fitness value of the individual. N individual positions are randomly initialized in the search space.

[0132] Before each position update, the probability of classifying an individual (P) is calculated. If the probability P is greater than 0.7, the individual is classified as an excellent individual, and the dynamic sensitivity position update method is used to update the individual's position. If the probability P is greater than 0.7, the individual is classified as an excellent individual. If the probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position; if the classification probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position. If the individual is classified as a poor performer, a distribution estimation update method is used to update the individual's position. This detailed classification allows for different update strategies to be applied to different types of individuals, improving optimization efficiency and effectiveness. The formula used to calculate the classification probability of an individual is as follows:

[0133] ;

[0134] In the formula, and These represent the partition probability and position of the v-th individual at the h-th iteration, where h is the current iteration number, H is the maximum iteration number, and v is the individual index. It is the fitness value of the v-th individual in the h-th iteration. It is the relative ranking of the v-th individual among N individuals after the fitness value of the v-th individual is increased by 1 in the h-th iteration, and rank() is the ascending ranking function;

[0135] Individual location updates include the following:

[0136] The dynamic sensitivity position update method introduces a random angle θ and a dynamic sensitivity design to update the position of superior individuals. This allows superior individuals to move more flexibly in the search space, helping to find better model parameters more quickly. The formulas used are as follows:

[0137] ;

[0138] ;

[0139] In the formula, yes The position updated using the dynamic sensitivity position update method. It is the position of the a1-th excellent individual in the h-th iteration, and s(h) is the sensitivity in the h-th iteration. maxThat is the maximum sensitivity. It is the position of the globally optimal individual at the h-th iteration. The globally optimal individual is the individual with the minimum fitness value. rand() is a random number generation function. The random angle θ ranges from 0 to 2π.

[0140] A dynamic convergence position update method is proposed, which introduces a control factor and a dynamic convergence factor to design the position update method for general individuals. This makes the updates of general individuals more targeted and promotes the overall optimization process. The formula used is as follows:

[0141] ;

[0142] ;

[0143] In the formula, yes The position is updated using the dynamic convergence position update method. It is the position of the a2th general individual in the h-th iteration, d min and d max These are the minimum and maximum values ​​of the convergence factor, respectively. d(h) is the convergence factor at the h-th iteration. r1 and r2 are the first and second random numbers in the range (0, 1), respectively. k is the control factor.

[0144] The distribution estimation position update method introduces weighted coefficients and a Gaussian function to design a position update mechanism for poorly performing individuals. This guides poorly performing individuals towards the direction where the best individual will emerge, improving the overall optimization effect. The formula used is as follows:

[0145] ;

[0146] ;

[0147] ;

[0148] ;

[0149] In the formula, yes The updated position using the distribution estimation position update method. It represents the position of the a3th worst individual in the h-th iteration, where xmean(h) and cov(h) are the weighted average and weighted covariance in the h-th iteration, respectively, and ω is the position of the a3th worst individual. v is the weighting coefficient of the v-th individual, and Gaussian function (·) is the Gaussian function;

[0150] Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the global best individual are updated. When the fitness value corresponding to the global best individual is lower than the fitness threshold, the model parameters corresponding to the position of the global best individual are the optimal model parameters, and a stress state detection model is constructed based on the optimal model parameters. Otherwise, if the maximum number of iterations is reached, the individual positions are re-initialized; otherwise, the iteration count is incremented by 1 and individuals are re-divided for individual position updates.

[0151] By performing the above operations, this solution addresses the problems of inaccurate pressure control strategies caused by the difficulty in accurately optimizing model parameters and low search efficiency in existing pressure control systems. Based on probability partitioning, individuals are categorized into excellent, average, and poor individuals. A position update method for excellent individuals is designed using random angles and dynamic sensitivity; a position update method for average individuals is designed using control factors and dynamic convergence factors; and a position update method for poor individuals is designed using weighted coefficients and Gaussian functions. This determines the optimal model parameters, improves the efficiency and accuracy of model parameter optimization, enhances the targeting and adaptability of the optimization, and better adapts to different situations. This contributes to more accurate and reliable pressure state detection, thereby improving the overall performance and intelligence level of the pressure control system.

[0152] Example 6, see Figure 1 This embodiment is based on the above embodiment. The pressure control strategy generation module collects real-time pressure data. The real-time pressure data includes timestamps, pressure values, environmental data, and equipment operating status data. After preprocessing the real-time pressure data, it is input into the pressure status detection model constructed based on the optimal model parameters. Based on the output labels of the model, the current pressure status is understood, and a pressure control strategy is formulated.

[0153] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0154] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0155] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An adjustable pressure control system, characterized in that: It includes a data acquisition module, a data preprocessing module, a pressure state detection model construction module, a pressure state detection model parameter optimization module, and a pressure control strategy generation module; The data acquisition module collects historical pressure data; The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data. The stress state detection model module constructs a stress map, uses correlation coefficients and multi-head attention mechanism, and adds spatial context information using residual connections to obtain a global spatial stress feature matrix. It calculates and concatenates local spatial and temporal stress features to obtain local stress features. It introduces position encoding and multi-head self-attention mechanism to calculate global spatiotemporal stress features to complete stress state detection. The pressure state detection model parameter optimization module divides individuals into excellent, average, and poor individuals based on the partition probability. It introduces random angle and dynamic sensitivity to design a position update method for excellent individuals, introduces control factors and dynamic convergence factors to design a position update method for average individuals, and introduces weighting coefficients and Gaussian functions to design a position update method for poor individuals, thereby determining the optimal model parameters. The pressure control strategy generation module understands the current pressure status based on the output labels of the pressure status detection model and formulates a pressure control strategy.

2. The adjustable pressure control system according to claim 1, characterized in that: The module for constructing the pressure state detection model specifically includes the following: A stress graph is generated by treating each data point in the stress detection dataset as a node and using the data information of each node as its feature vector. A similarity threshold is pre-set. If the cosine similarity between two data points is greater than the similarity threshold, an edge is added between the corresponding nodes; otherwise, no edge is added. The final result is a node feature matrix G and a stress graph O = (F, δ, D), where F is the set of nodes, δ is the set of edges, and D is the adjacency matrix. Elements in D represent the connectivity between nodes; if there is an edge between two nodes, the element value is 1, otherwise the element value is 0. Extract global spatial pressure features; Local pressure features are extracted by calculating the local spatial pressure feature matrix and the local temporal pressure feature matrix, respectively, and then concatenating the two matrices to obtain the local pressure feature matrix. This includes the following: The formula used to calculate the local spatial pressure characteristic matrix is ​​as follows: ； ； In the formula, It is the local spatial pressure feature matrix, and Sigmoid(•) is the Sigmoid activation function, E lap It is the Laplace matrix, W gc It is a learnable matrix, I is the identity matrix, and A is the degree matrix of D+I; The formula used to calculate the local time pressure characteristic matrix is ​​as follows: ； ； ； ； In the formula, u t m t and n t These are the reset gate, update gate, and candidate hidden state at time step t, respectively, where t is the time step index, and b is the time step index. u b m and b n They are u t m t and n t The bias, B t-1 Let be the hidden state at time step (t-1), GC(·) is the graph convolution operation, and tanh(·) is the hyperbolic tangent activation function. and These are the local time pressure feature matrices extracted at the t-th and t-1-th time steps, respectively; It is the global spatial pressure feature matrix extracted at the t-th time step; Extract global spatiotemporal pressure features, based on the local pressure feature matrix Calculate the global spatiotemporal pressure feature matrix; Stress state detection involves processing the global spatiotemporal stress feature matrix into a normalization layer, a feedforward neural layer, and an output layer to obtain the stress state. The loss function of the model is then derived based on the mean squared error, using the following formula: ； ； ； In the formula, C and These are the outputs of the normalization layer and the feedforward neural layer, respectively. This is the detection result output by the stress state detection model. LayerNorm is the normalization layer. W1 and W2 are the weight matrices of the first and second linear transformations of the feedforward neural layer, respectively. b1 and b2 are the biases of the first and second linear transformations of the feedforward neural layer, respectively. y and b y These are the weight matrix and bias of the output layer, respectively.

3. The adjustable pressure control system according to claim 2, characterized in that: The extraction of global spatial pressure features specifically includes the following: The correlation coefficient between nodes is calculated using the following formula: ； ； In the formula, It is the correlation coefficient between the i-th and j-th nodes, where i and j are node indices. It is a modified linear unit activation function with leakage, || is the concatenation operation, and W xb It is the linear transformation weight matrix, g i and g j These are the feature vectors of the i-th and j-th nodes, respectively. and They are g i and g j After linear transformation, the eigenvectors are β, which is the weight vector, and T is the vector transpose operation. The attention weights between nodes are calculated using the following formula: ； In the formula, It is the attention weight between the i-th and j-th nodes; The global spatial pressure feature matrix is ​​calculated, attention is embedded in a multi-head manner, and spatial context information is added using residual connections; the formulas used are as follows: ； In the formula, G is the global spatial pressure feature matrix, G is the node feature matrix, Head is the number of attention heads, head is the index of the attention head, and ReLU(·) is the ReLU activation function.

4. The adjustable pressure control system according to claim 2, characterized in that: The extraction of global spatiotemporal pressure features specifically Includes the following: Location encoding, in the local pressure feature matrix The positional encoding information is embedded in the middle, and the formula used is as follows: ； ； In the formula, It is the local stress feature matrix after embedding positional encoding information, where x is the feature dimension index and w is the local stress feature matrix. The number of feature dimensions, It is location-coded information; Calculate the global spatiotemporal pressure feature matrix and use a multi-head self-attention mechanism from The global spatiotemporal pressure characteristics were calculated using the following formula: ； ； ； In the formula, Q head K head and V head These are the query vector, key vector, and value vector corresponding to the head-th attention head, respectively. , and These are the query weight matrix, key weight matrix, and value weight matrix corresponding to the head-th attention head. U is the self-attention score of the head-th attention head. 1 It is the global spatiotemporal pressure feature matrix, and Softmax(·) is the normalized exponential function.

5. The adjustable pressure control system according to claim 1, characterized in that: The pressure state detection model parameter optimization module specifically includes the following: Initialize individual positions based on the linear transformation weight matrix W in the pressure state detection model. xb Weight vector β, learnable matrix W gc Query weight matrix Key weight matrix Value weight matrix The weight matrix W of the output layer y The search space is established by the weight matrix W1 of the first linear transformation and the weight matrix W2 of the second linear transformation of the feedforward neural layer. The individual position is used as the representative of the stress state detection model parameters. The loss function of the stress state detection model established based on the model parameters is used as the fitness value of the individual. N individual positions are randomly initialized in the search space. Before each position update, the probability of classifying an individual (P) is calculated. If the probability P is greater than 0.7, the individual is classified as an excellent individual, and the dynamic sensitivity position update method is used to update the individual's position. If the probability P is greater than 0.7, the individual is classified as an excellent individual. If the probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position; if the classification probability P is less than or equal to 0.7, the individual is classified as a general individual, and the dynamic convergent position update method is used to update the individual's position. If the individual is classified as a poor individual, the distribution estimation update method is used to update the individual's position; the formula used to calculate the classification probability of the individual is as follows: ； In the formula, and These represent the partition probability and position of the v-th individual at the h-th iteration, where h is the current iteration number, H is the maximum iteration number, and v is the individual index. It is the fitness value of the v-th individual in the h-th iteration. It is the relative ranking of the v-th individual among N individuals after the fitness value of the v-th individual is increased by 1 in the h-th iteration, and rank() is the ascending ranking function; Individual location update; Once the optimal model parameters are determined, a fitness threshold is pre-set, and the fitness values ​​of individuals and the global best individual are updated. When the fitness value corresponding to the global best individual is lower than the fitness threshold, the model parameters corresponding to the position of the global best individual are the optimal model parameters, and a stress state detection model is constructed based on the optimal model parameters; otherwise, if the maximum number of iterations is reached, the individual positions are re-initialized. Otherwise, increment the iteration count by 1 and re-divide the individuals to update their positions.

6. The adjustable pressure control system according to claim 5, characterized in that: The individual location update specifically includes the following: A dynamic sensitivity position update method is proposed, which introduces a random angle θ and a dynamic sensitivity design method for position update to update the position of high-performing individuals; the formula used is as follows: ； ； In the formula, yes The position updated using the dynamic sensitivity position update method. It is the position of the a1-th excellent individual in the h-th iteration, and s(h) is the sensitivity in the h-th iteration. max That is the maximum sensitivity. It is the position of the globally optimal individual at the h-th iteration. The globally optimal individual is the individual with the minimum fitness value. rand() is a random number generation function. The random angle θ ranges from 0 to 2π. A dynamic convergence position update method is proposed, which introduces a control factor and a dynamic convergence factor to design a position update method for general individuals; the formula used is as follows: ； ； In the formula, yes The position is updated using the dynamic convergence position update method. It is the position of the a2th general individual in the h-th iteration, d min and d max These are the minimum and maximum values ​​of the convergence factor, respectively. d(h) is the convergence factor at the h-th iteration. r1 and r2 are the first and second random numbers in the range (0, 1), respectively. k is the control factor. A distribution estimation location update method is proposed, which introduces weighting coefficients and a Gaussian function to design a location update method for poorly performing individuals; the formula used is as follows: ； ； ； ； In the formula, yes The updated position using the distribution estimation position update method. It represents the position of the a3th worst individual in the h-th iteration, where xmean(h) and cov(h) are the weighted average and weighted covariance in the h-th iteration, respectively, and ω is the position of the a3th worst individual. v is the weighting coefficient of the v-th individual, and Gaussian function (·) is the Gaussian function.

7. The adjustable pressure control system according to claim 1, characterized in that: The pressure control strategy generation module collects real-time pressure data, which includes timestamps, pressure values, environmental data, and equipment operating status data. After preprocessing the real-time pressure data, it is input into a pressure status detection model constructed based on optimal model parameters. Based on the output labels of the model, the current pressure status is understood, and a pressure control strategy is formulated.

8. The adjustable pressure control system according to claim 1, characterized in that: The data acquisition module collects historical pressure data, which includes timestamps, pressure values, environmental data, equipment operating status data, and pressure status. The pressure status is used as a data label.

9. The adjustable pressure control system according to claim 1, characterized in that: The data preprocessing module performs data cleaning, data transformation, data normalization, and dataset construction on the collected data; data cleaning includes handling missing values, outliers, and duplicate values. Data transformation converts data into vector form; data normalization unifies the data range based on the min-max normalization method; and dataset construction builds a stress detection dataset based on the processed data.

Images