Material dynamic regulation and control system and control method

Abstract

The invention relates to the technical field of industrial intelligent control, and provides a material dynamic regulation and control system and a control method, which are different from a traditional material processing mode depending on artificial experience or single control by constructing an intelligent regulation and control system from data perception, strategy generation to precise control. The regulation and control system realizes multi-dimensional real-time monitoring and dynamic optimization of the material state, can quickly respond to the material state change and the process demand difference, remarkably improves the material processing accuracy, stability and production efficiency, forms a self-learning ability by continuously accumulating historical regulation and control data and optimizing a regulation and control strategy, and improves the material processing efficiency. The manual intervention cost and the production risk are effectively reduced, and an innovative solution is provided for automatic and intelligent regulation and control of materials in industrial production.

Description

Technical Field

[0001] This application relates to the field of industrial intelligent control technology, and in particular to material dynamic control systems and methods. Background Technology

[0002] In modern industrial production, material control technology is widely used in many fields such as food processing, chemical industry and pharmaceutical industry. Its control effect directly affects product quality, production efficiency and energy consumption level. With the development of industrial automation and intelligence, traditional material control technology is also constantly evolving.

[0003] Early material control relied mainly on manual experience. Operators would manually adjust equipment parameters based on subjective judgments such as equipment operating status and material appearance. This method was not only inefficient, but also highly dependent on the operator's experience and condition, making it difficult to guarantee consistency and accuracy in control. The drawbacks were even more pronounced when dealing with complex materials or large-scale production.

[0004] With technological advancements, automated control systems based on single control models have emerged, such as PID control systems. These systems use fixed parameters to perform closed-loop control of material parameters like temperature and flow rate, improving stability and efficiency to some extent. However, single control models are often based on idealized assumptions and have poor adaptability to changes in material properties and complex operating conditions. When the density, viscosity, and structure of materials fluctuate, the fixed-parameter control model cannot adjust in time, easily leading to lagging or over-regulation, affecting production quality and efficiency. Most of these systems fail to address how to achieve intelligent material control by sensing multi-dimensional data under complex operating conditions and linking it to the material's state. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides a material dynamic control system and control method.

[0006] In a first aspect, this application provides a material dynamic control system, which includes: acquiring modal data of materials in a target device, extracting material features from the modal data as material density distribution and material structure features, fusing the material density distribution and material structure features to generate a feature tensor, and determining the material state at different spatial locations in the target device through spatiotemporal coding;

[0007] The weight matrix of the process parameter set and feature tensor is determined based on the attention mechanism. The process parameter set and feature tensor are multiplied by the weight matrix to generate the association tensor. The association tensor is then fused and processed with the process knowledge graph to output the association features. At the same time, the control strategy is generated by combining the material state at different spatial locations.

[0008] The actuator controls the material in the target equipment according to the control strategy, monitors the control effect, matches the actuator impedance based on the material structure characteristics, and adjusts the actuator's execution parameters. At the same time, the control effect is compared with the target effect to obtain the effect deviation. Based on the effect deviation, it is determined whether to update the control strategy and the material status at different locations in the target equipment.

[0009] As an optional implementation, the logic for generating the feature tensor includes:

[0010] Modal data of materials in the target equipment are acquired, and the modal data are preprocessed using an adaptive noise reduction algorithm;

[0011] The preprocessed modal data is mapped to the density feature space and the structural feature space by a variational autoencoder. The density changes of the modal data are extracted by a convolutional neural network to obtain the material density distribution. The topological relationship between the particles of the material is analyzed by a graph neural network to obtain the material structural features.

[0012] Feature weights for material density distribution and material structure characteristics are generated through an attention mechanism, and feature tensors are generated through weighted summation.

[0013] As an optional implementation, the logic for determining the material state includes:

[0014] Assign spatiotemporal coordinate information to the feature tensor, mark the spatial location and time series of the material in the target equipment, and perform feature interaction on the feature tensor through a multi-head attention mechanism to extract the change features of the material in spatial location and time series;

[0015] The material state of the target device at different spatial locations is predicted based on the changing characteristics using a Bayesian neural network, and the prediction uncertainty is output.

[0016] Local and global features are extracted from the feature tensor. Fusion weights are generated based on the predicted material state and prediction uncertainty. The local and global features are then fused according to the fusion weights to obtain fused features. The fused features are then mapped to determine the material state of the target device at different spatial locations.

[0017] As an optional implementation, the output logic of the associated feature includes:

[0018] Receive the feature tensor and obtain the process parameter set. Calculate the correlation weights between the process parameter set and the feature tensor in the spatial dimension, temporal dimension, and feature dimension through the attention mechanism. Multiply the correlation weights by the weight matrix of the process parameter set and the feature tensor.

[0019] The causal relationship between the process parameter set and material characteristics is identified by the causal reasoning network, and the weight matrix of the process parameter set and feature tensor is corrected according to the causal relationship.

[0020] Tensor product operation is performed on the process parameter set and feature tensor based on the corrected weight matrix to generate the association tensor. The association tensor is then fused with the process knowledge graph. The fused process knowledge graph is then inferred through a graph neural network to output the association features.

[0021] As an optional implementation, the generation logic of the control strategy includes:

[0022] Reinforcement learning is constructed, where the state space is the global material state formed by splicing the material states and associated features at different spatial locations in the target device, and the action space includes the parameter configuration set of the actuator, and the control efficiency, energy consumption and stability are used as reward functions.

[0023] Based on the global material state and related characteristics, a global control strategy is output. At the same time, with the global control strategy as a constraint, a local control strategy is output for the spatial location where the material state is abnormal. In addition, a time-series control strategy is output according to the changes in the material state to generate a control strategy.

[0024] The prediction uncertainty increment is added to the state space, and the regulation strategy is screened and optimized by a genetic algorithm to obtain the optimized regulation strategy. The feasibility of the optimized regulation strategy is then verified to correct the regulation strategy.

[0025] As an optional implementation, the logic for adjusting the execution parameters includes:

[0026] The execution parameters of the actuator in the control strategy are extracted, and the execution parameters are decomposed into basic parameters and modulation parameters. The adjustment range of the execution parameters is determined by combining the prediction uncertainty, and the confidence interval of the adjusted execution parameters is calculated by Monte Carlo algorithm.

[0027] Based on process requirements, historical control data, and material status, the basic parameters and modulation parameters are prioritized, and the execution parameters are adjusted according to the priority order. The feedback on the adjustment effect of the execution parameters is monitored, and the execution parameters are readjusted based on the confidence interval of the adjusted execution parameters.

[0028] After each adjustment of the execution parameters, the coordination between the basic parameters and the modulation parameters is checked to trigger a reverse adjustment based on priority order, and the effect of each adjustment is updated and recorded in the historical control data.

[0029] As an optional implementation, the actuator impedance matching sub-logic includes:

[0030] The material structure features are received and decomposed into component features through wavelet transform. The component features include micro-features, macro-features and meso-features.

[0031] Actuator impedance includes mass, damping, and stiffness. A mapping relationship between component characteristics and actuator impedance is established based on historical control data to determine the adjustment direction and adjustment range of actuator impedance.

[0032] During material handling, the changes in material state and the operation feedback of the actuator are monitored in real time, so as to dynamically adjust the mapping relationship between the actuator impedance and component characteristics and the actuator impedance through a dual time scale control mechanism.

[0033] As an optional implementation, the update logic of the control strategy includes:

[0034] The deviation between the control effect and the target effect is calculated by Mahalanobis distance, and the correlation between the deviations is measured. The deviations include state deviation, energy consumption deviation and stability deviation.

[0035] A decision tree is constructed based on historical control data. The direction of adjustment of the control strategy is generated based on the combination of effect deviations. The decision tree is then optimized periodically based on the success rate of updating the control strategy.

[0036] The constraints for adjusting the execution parameters are determined based on the physical limitations of the actuator and the properties of the material, and the execution parameters are adjusted iteratively through adaptive step size.

[0037] Real-time monitoring of the control effect of adjusted execution parameters on materials in the target equipment, determination of effect deviation, and closed-loop optimization of control strategies.

[0038] As an optional implementation, the material status update mechanism includes:

[0039] Configure a deviation threshold. When any effect deviation exceeds the deviation threshold, calculate the causal probability that the effect deviation is caused by the change in the material state based on the causal relationship between the process parameter set and the material characteristics.

[0040] The explanatory power of the parameters that verify the effect deviation is verified by adjusting the execution parameters, and the update status of the material status is judged by combining the causal probability and the explanatory power of the parameters.

[0041] When the material status needs to be updated, the update step size of the material status is dynamically adjusted through fuzzy logic control. After the material status is updated, the effect deviation is recalculated to iteratively optimize the material status.

[0042] Secondly, this application provides a control method for a material dynamic control system, the method comprising: acquiring modal data of materials in a target device, and extracting material characteristics from the modal data as material density distribution and material structural characteristics;

[0043] Material density distribution and material structure features are fused to generate a feature tensor, and the material state at different spatial locations in the target equipment is determined by spatiotemporal coding.

[0044] The weight matrix of the process parameter set and feature tensor is determined based on the attention mechanism. The process parameter set and feature tensor are then multiplied by tensor based on the weight matrix to generate the associated tensor.

[0045] The correlation tensor is fused and processed with the process knowledge graph to output correlation features, and control strategies are generated by combining the material state at different spatial locations.

[0046] The actuator controls the material in the target equipment according to the control strategy, monitors the control effect, matches the actuator impedance based on the material structure characteristics, and adjusts the actuator's execution parameters.

[0047] By comparing the control effect with the target effect, the effect deviation is obtained. Based on the effect deviation, it is determined whether to update the control strategy and the material state at different spatial locations in the target equipment.

[0048] Compared with existing technologies, the beneficial effects of this application are as follows: By constructing an intelligent control system that integrates data perception, strategy generation, and precise control, this system differs from traditional material handling methods that rely on manual experience or single control. The control system achieves multi-dimensional real-time monitoring and dynamic optimization of material states, enabling rapid response to changes in material states and differences in process requirements. This significantly improves the accuracy, stability, and production efficiency of material handling. At the same time, by continuously accumulating historical control data and optimizing control strategies, the control system develops self-learning capabilities, effectively reducing the cost of manual intervention and production risks. This provides an innovative solution for the automated and intelligent control of materials in industrial production.

[0049] By acquiring modal data of materials and extracting material density distribution and material structural characteristics, the system provides accurate basic information on material state for the control system. The generated feature tensor integrates multi-dimensional information and locates the state changes of materials in the target equipment through spatiotemporal coding. This enables the control system to grasp the characteristics of materials in different spatial locations and time series in real time and accurately, providing reliable data support for subsequent process decisions and control, and avoiding control deviations caused by inaccurate information.

[0050] Based on attention mechanisms and causal reasoning networks, we can deeply analyze the relationship between process parameters and material characteristics, uncover potential causal logic, and combine data-driven analysis with industry experience knowledge by integrating with process knowledge graphs to generate associated features with rich semantic information. On this basis, we can generate control strategies by combining material status, which effectively solves the problems of low matching degree between parameters and material characteristics and single strategy in traditional control, improves the pertinence and optimization effect of control strategies, and helps to achieve refined control of the production process.

[0051] Guided by the control strategy, the system achieves precise control of materials through actuators. During the control process, the actuator impedance is matched based on the material's structural characteristics, and the actuator parameters can be dynamically adjusted according to the material's properties. This avoids equipment damage or poor material handling results caused by impedance mismatch. At the same time, by comparing the control effect with the target effect, the system determines in real time whether to update the control strategy and material status, forming a closed-loop optimization mechanism. This enables the control system to quickly adapt to changes in the production process, continuously optimize the control strategy, improve material handling quality and production efficiency, reduce energy consumption, and enhance the stability of the control system. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0053] Figure 1 This is a system flowchart of the material dynamic control system provided in the embodiments of this application;

[0054] Figure 2 This is a logic diagram for determining the material state of the material dynamic control system provided in the embodiments of this application;

[0055] Figure 3 The matching sub-logic diagram of the actuator impedance of the material dynamic control system provided in the embodiments of this application;

[0056] Figure 4 This is a flowchart illustrating the control method of the material dynamic control system provided in this application embodiment. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of this application more apparent and understandable, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0058] Example 1

[0059] like Figure 1 The diagram shown is a system flowchart of a material dynamic control system provided in this application embodiment. The system includes a modal sensing module, a process association module, and a dynamic control module.

[0060] The actual application scenario here takes automatic sugar and water addition as an example. The target equipment is a water tank, a sugar tank, and a mixing tank, and the materials are water and sugar.

[0061] The modal perception module is used to acquire modal data of materials in the target equipment, extract material features from the modal data as material density distribution and material structure features, fuse material density distribution and material structure features to generate feature tensors, and determine the material state at different spatial locations in the target equipment through spatiotemporal coding.

[0062] Specifically, the logic for generating feature tensors includes:

[0063] Modal data of materials in the target equipment are acquired, and the modal data are preprocessed using an adaptive noise reduction algorithm;

[0064] The preprocessed modal data is mapped to the density feature space and the structural feature space by a variational autoencoder. The density changes of the modal data are extracted by a convolutional neural network to obtain the material density distribution. The topological relationship between the particles of the material is analyzed by a graph neural network to obtain the material structural features.

[0065] Feature weights for material density distribution and material structure characteristics are generated through an attention mechanism, and feature tensors are generated through weighted summation.

[0066] The raw modal data of the materials in the target equipment are easily affected by environmental noise. Directly using them for feature extraction will lead to inaccurate results. Therefore, the modal data needs to be preprocessed to improve data quality. The physical properties of sugar and water are very different, so customized sensor configuration and noise reduction strategies are required. Specifically, the sugar tank needs to monitor agglomeration and bridging, while the water tank needs to focus on liquid level fluctuations and turbulence. Vibration sensors are deployed on the outside of the sugar tank to detect particle flow, a laser particle size analyzer is installed on the top to monitor sugar particle distribution in real time, and a pressure sensor at the bottom detects the risk of blockage at the discharge port. For the water tank, an ultrasonic level gauge is used to measure the water level, acoustic sensors are installed on the tank wall to capture liquid flow noise, and water quality sensors monitor indicators such as dissolved oxygen.

[0067] The acquired modal data is processed using an adaptive noise reduction algorithm. For sugar particle flow signals, it is necessary to enhance the filtering of high-frequency vibration noise, while for water flow signals, Kalman filtering is used to eliminate noise caused by liquid level fluctuations. The sensor configuration is optimized according to material characteristics, which improves the ability to perceive key states such as sugar agglomeration and water turbulence. Adaptive noise reduction processing ensures the reliability of modal data. Accurate sugar particle distribution data and water level information provide high-quality input for subsequent density and structural feature extraction.

[0068] Raw modal data cannot be directly used for material state analysis. It needs to be transformed into density distribution and structural features that reflect the essential properties of the material in order to generate feature tensors. The particle characteristics of sugar and the fluid characteristics of water need to be extracted using different algorithms. The particle characteristics of sugar include particle size distribution and bulk density, while the fluid characteristics of water include flow velocity and liquid level. The preprocessed modal data is mapped to the density feature space and structural feature space respectively by a variational autoencoder. In terms of density feature extraction, the bulk density changes at different heights in the sugar tank are identified by a convolutional neural network using data from a laser particle size analyzer and the pressure distribution of the sugar tank. When an abnormal increase in local pressure is detected, it is determined that there is a clumping phenomenon. The convolutional neural network automatically extracts density change features at different scales in the data through multiple convolutional layers and pooling layers.

[0069] In terms of structural feature extraction, based on the water flow noise captured by acoustic sensors, graph neural networks are used to analyze the turbulence level and vortex structure of the water flow, including identifying the sound wave patterns of specific frequencies to determine whether there is abnormal turbulence caused by pipe blockage; the topological relationship between sugar particles is constructed using data from vibration sensors, and the aggregation state of sugar particles is analyzed by graph neural networks. When vibration signal attenuation is detected, bridging phenomenon is determined; feature extraction algorithms are customized for the characteristics of sugar and water, which can accurately capture the risk of agglomeration of particulate materials and the abnormal flow of liquid materials. The extracted material density distribution and material structure features provide key inputs for subsequent feature tensor generation, ensuring that the feature tensors can fully reflect the material state.

[0070] To more comprehensively and efficiently represent material characteristics, it is necessary to fuse material density distribution and structural features to generate a unified feature tensor, facilitating subsequent spatiotemporal coding and material state determination. This means integrating the particle characteristics of sugar with the fluid characteristics of water to form a unified representation for subsequent spatiotemporal coding analysis. An attention mechanism is introduced to analyze material density distribution and structural features, generating feature weights for both. This mechanism automatically assigns weights based on the importance of different features in reflecting the material state. When there is a risk of clumping in the sugar tank, the weight of the sugar structural features is automatically increased; when the water tank level approaches the warning line, the weight of the water structural features is increased, and then... By using a weighted summation method, the density distribution and structural features of the material are fused to generate a feature tensor. This feature tensor integrates the density and structural information of the material and is stored in the form of a multidimensional array, which can clearly reflect the combination of features of the material in different dimensions. The feature weights are dynamically adjusted so that the feature tensor can highlight key information according to the working conditions, thereby improving the ability to represent the state of the material. The fused feature tensor contains the spatiotemporal variation information of sugar and water, realizing the efficient fusion and unified representation of material features, and can more comprehensively describe the state of the material. Compared with using a single feature, it greatly improves the ability to represent the state of the material and provides a rich data foundation for subsequent spatiotemporal coding.

[0071] Specifically, such as Figure 2 As shown, the logic for determining the state of materials includes:

[0072] Assign spatiotemporal coordinate information to the feature tensor, mark the spatial location and time series of the material in the target equipment, and perform feature interaction on the feature tensor through a multi-head attention mechanism to extract the change features of the material in spatial location and time series;

[0073] The material state of the target device at different spatial locations is predicted based on the changing characteristics using a Bayesian neural network, and the prediction uncertainty is output.

[0074] Local and global features are extracted from the feature tensor. Fusion weights are generated based on the predicted material state and prediction uncertainty. The local and global features are then fused according to the fusion weights to obtain fused features. The fused features are then mapped to determine the material state of the target device at different spatial locations.

[0075] The state of materials in the target equipment is not only related to their own characteristics, but also affected by spatial location and time series. In order to accurately determine the state of materials, it is necessary to perform spatiotemporal encoding on the feature tensor and extract its change features in the spatiotemporal dimension. First, the feature tensor is given spatiotemporal coordinate information to clarify the spatial location and time series of the materials in the target equipment. That is, three-dimensional coordinate systems are established for the sugar tank and the water tank respectively, and the spatial locations of sugar and water at different time points are marked, including the outlet position of the sugar tank and the inlet position of the water tank.

[0076] Then, a multi-head attention mechanism is used to perform feature interaction on the feature tensor. The multi-head attention mechanism analyzes the feature tensor from multiple perspectives, capturing the correlation and changes of material features between different spatial locations and time points. In the sugar-water mixing stage, the relationship between sugar dissolution rate and water flow rate is focused on to capture the spatiotemporal features of the uneven mixing area, thereby extracting the changes of material in spatial location and time series. Spatiotemporal encoding and feature interaction enable the full mining and integration of material feature information in the spatiotemporal dimension, which can more accurately describe the dynamic change process of material in the target equipment, providing richer spatiotemporal information for material state prediction. The extracted change features provide key input for Bayesian neural network to predict material state.

[0077] After acquiring the spatiotemporal variation information of material characteristics, it is necessary to predict the material state at different locations in the target equipment using a Bayesian neural network to provide a basis for subsequent material control decisions. In practice, this means predicting states such as the degree of sugar dissolution and water level changes to provide decision support for automatic sugar and water addition. The Bayesian neural network processes the extracted variation features to predict the material state at different spatial locations in the target equipment. The Bayesian neural network can not only output the predicted value of the material state but also the prediction uncertainty. By learning from a large amount of historical data, the Bayesian neural network establishes a probabilistic relationship between material characteristics and material state. During the prediction process, based on the current input variation features and the learned probabilistic relationship, it calculates the probability of different material states and selects the state with the highest probability as the predicted value. At the same time, it outputs the uncertainty of the prediction result to reflect the reliability of the prediction. That is, inputting features such as sugar particle distribution and water flow velocity, the output is the predicted value and uncertainty of the sugar dissolution state, i.e., predicting the percentage of sugar dissolved in water and providing a confidence interval. When there is uncertainty in the dissolution process, the dependence on sugar concentration sensor data is increased, and the prediction weight of the Bayesian neural network is reduced.

[0078] The application of Bayesian neural networks enables material state prediction to not only provide the prediction result, i.e., the sugar dissolution state, but also to assess the uncertainty of the prediction, providing more comprehensive information for decision-making. This helps users to rationally formulate control strategies based on the reliability of the prediction, reduce decision-making risks, and specifically adopt a conservative strategy when the uncertainty is high, i.e., extend the stirring time. The predicted material state value and the prediction uncertainty provide important references for subsequent local and global feature fusion, which are used to generate fusion weights and guide the feature fusion process to more accurately determine the final state of the material.

[0079] To comprehensively consider both the local details and global trends of materials and accurately determine their final state at different locations within the target equipment, it is necessary to fuse the local and global features of the feature tensor and make decisions based on the predicted material state and uncertainties. In practice, this means comprehensively considering both the local details and global trends of sugar dissolution to accurately determine the mixing state. First, the local and global features of the feature tensor are extracted. Local features are obtained by analyzing local regions of the feature tensor, focusing on subtle changes and local characteristics of the material, i.e., the arrangement of material particles within a small area. Global features are obtained through comprehensive analysis of the entire feature tensor, reflecting the overall distribution and macroscopic trends of the material, i.e., the overall density distribution of the material within the equipment.

[0080] Then, based on the predicted material state and prediction uncertainty, fusion weights are generated. If the prediction uncertainty is high, the fusion weight of local features is appropriately increased to focus more on the local details of the material and ensure accurate judgment of the material state; conversely, the fusion weight of global features is increased. Next, the local and global features are weighted and fused according to the fusion weights to obtain fused features. Finally, the fused features are input into the mapping rules to determine the final state of the material at different locations in the target equipment. In practical applications, local features focus on the dissolution state of sugar particles near the tank outlet to determine whether there are undissolved sugar lumps, while global features analyze the sugar concentration distribution throughout the mixing tank to assess the mixing uniformity. When the prediction uncertainty is high, the fusion weight of local features is increased to focus on checking for lumps. When the scheduling system is stable, the fusion weight of global features is increased to focus on the overall mixing effect.

[0081] The mapping rules include classifying material states into three levels of judgment results: normal, warning, and abnormal. Relevant rules are formulated for different material types and equipment areas. For example, in the bottom area of ​​the sugar tank, if the fusion feature shows that the local density is greater than the density threshold and the global density gradient is abnormal, it is judged as a clumping warning. However, if the vibration sensor signal is attenuated at the same time, it is upgraded to a clumping warning. In the mixing area of ​​the water tank, if the fusion feature shows that the local turbulence intensity is less than the standard intensity and the global flow velocity is uneven, it is judged as a mixing abnormality.

[0082] By fusing local and global features and combining them with predicted material states for decision-making, the material state can be determined comprehensively and accurately. This fully considers the characteristics of materials at different scales and with different levels of reliability, and can accurately judge the sugar dissolution state and mixing uniformity under different operating conditions. Compared with single feature analysis, it significantly improves the accuracy and reliability of material state determination, avoids misjudgments caused by local anomalies, and provides accurate basis for the process association module to generate control strategies. This allows the control strategies to be more targeted in controlling the materials, and also provides clear objectives for the dynamic control module to perform control operations, thereby improving the control effect and operating efficiency of the entire material dynamic control system.

[0083] The process association module is used to determine the weight matrix of the process parameter set and feature tensor based on the attention mechanism. According to the weight matrix, the process parameter set and feature tensor are multiplied by tensor to generate the association tensor. This is then fused and processed with the process knowledge graph to output the association features. At the same time, the module generates control strategies by combining the material state at different spatial locations.

[0084] Specifically, the output logic for associated features includes:

[0085] Receive the feature tensor and obtain the process parameter set. Calculate the correlation weights between the process parameter set and the feature tensor in the spatial dimension, temporal dimension, and feature dimension through the attention mechanism. Multiply the correlation weights by the weight matrix of the process parameter set and the feature tensor.

[0086] The causal relationship between the process parameter set and material characteristics is identified by the causal reasoning network, and the weight matrix of the process parameter set and feature tensor is corrected according to the causal relationship.

[0087] Tensor product operation is performed on the process parameter set and feature tensor based on the corrected weight matrix to generate the association tensor. The association tensor is then fused with the process knowledge graph. The fused process knowledge graph is then inferred through a graph neural network to output the association features.

[0088] During the mixing process of sugar and water, the distribution of sugar particles and the turbulent state of water are significantly affected by the spatiotemporal influence of process parameters. It is necessary to analyze the coupling relationship between parameters and features through multi-dimensional weighting to avoid control lag or over-adjustment. Here, an attention mechanism is used to calculate the correlation weights of the process parameter set and feature tensor in the spatial, temporal, and feature dimensions. The spatial correlation weights are calculated by deploying sensor nodes at the sugar tank outlet, water tank mixing zone, and pipe bends, including pressure, vibration, and level sensors. For each process parameter, including sugar addition rate, stirring speed, and sugar tank valve opening, a two-dimensional convolutional kernel scans the spatial grid to calculate the impact of the process parameter on material characteristics at different spatial locations. For example, when the valve opening at the bottom of the sugar tank increases, the convolutional kernel outputs the weight of this action on the increase in sugar density at the bottom of the mixing tank; that is, the darker the color, the higher the weight, visually displaying the spatial influence range.

[0089] The temporal correlation weights are calculated by processing historical control data through a bidirectional gated loop unit, with a time window set to 2 hours to capture the temporal dependency between process parameter adjustments and material state changes. When analysis reveals that the lag time for an increase in stirring speed leading to increased water turbulence intensity is approximately 30 seconds, the bidirectional gated loop unit automatically assigns a higher weight to the impact of this process parameter on turbulence characteristics within the next 30 seconds. The feature-level correlation weights require constructing a semantic correlation matrix between process parameters and material characteristics. The matrix rows represent process parameters, and the matrix lists represent material characteristics, including sugar agglomeration rate and water turbulence intensity. The correlation weights are calculated using a self-attention mechanism. The semantic similarity between the elements of the semantic association matrix, namely the stirring speed and the water turbulence intensity, reaches 0.8, while the semantic similarity between the stirring speed and the sugar agglomeration rate is only 0.3, thus determining the association weights of the feature dimensions. The three-dimensional weight matrix realizes the spatiotemporal fine analysis of the influence of process parameters on material characteristics, which can accurately locate the problem of insufficient valve opening on the left side of the sugar tank leading to low sugar concentration in the lower right of the mixing tank, avoiding the global average adjustment mode of traditional PID control. The generated weight matrix provides spatiotemporally labeled association data for causal inference, where the spatial dimension weight can help determine whether there is a real causal relationship between the sugar addition rate and local agglomeration in the mixing tank.

[0090] During the mixing of sugar and water, there may be statistically correlated but causally unrelated parameters. Causal inference is needed to eliminate spurious associations and ensure that the weight matrix reflects the true action path. A causal inference network is constructed by inputting process parameters and material characteristics. An intervention experiment simulates the impact of changes in process parameters on the material state. Preferably, the water temperature is fixed at 25℃, and the stirring speed is forcibly increased from 500 rpm to 800 rpm. The change in sugar dissolution time is observed. If the dissolution time shortens, the causal effect of stirring speed on the dissolution rate is confirmed; if it remains unchanged, it is considered a spurious association, and its weight matrix value is reduced. Simultaneously, the strength of the causal effect is determined. If the influence of stirring speed on the dissolution rate reaches 70%, the weight matrix is ​​adjusted. If a weak causal relationship is found between stirring speed and sugar agglomeration rate, its weight is reduced to avoid excessive adjustment of the stirring speed. Causal inference effectively eliminates spurious associations in the data, making the weight matrix more physically interpretable and improving the reliability of subsequent feature fusion. The corrected weight matrix provides a more accurate weighting basis for tensor product operations, ensuring that the correlation tensor truly reflects the interaction between process parameters and feature tensors.

[0091] A single process parameter cannot fully reflect the complexity of the mixing process. The sugar addition rate needs to work in conjunction with the stirring speed and water temperature. Therefore, a high-dimensional correlation feature needs to be constructed through tensor product operations, and the semantic interpretability of the feature tensor needs to be improved by combining process knowledge. Based on the corrected weight matrix, tensor product operations are performed on the process parameters and feature tensors to generate a correlation tensor containing spatiotemporal and feature dimensions. Specifically, the sugar flow rate parameter is multiplied with the sugar density features at different heights of the mixing tank, highlighting the impact of flow rate on local density. The process knowledge graph is invoked, which stores rules such as excessively rapid sugar addition leading to clumping and the need to extend stirring time due to low water temperature. Each rule serves as a knowledge node, and the matching degree between the association tensor and the knowledge node is inferred through a graph neural network. When the association tensor indicates that the sugar addition rate is too fast and the water temperature is too low, the system automatically matches the clumping risk node in the process knowledge graph, enhancing the ability to identify abnormal states. Tensor product operation realizes the multidimensional fusion of process parameters and feature tensors, while the process knowledge graph introduces the experience of domain experts, so that the association features can reflect real-time data and conform to common sense in the process. The output association features integrate data-driven quantitative analysis and knowledge-driven qualitative judgment, providing a comprehensive decision-making basis for the generation of control strategies.

[0092] Specifically, the generation logic of the regulation strategy includes:

[0093] Reinforcement learning is constructed, where the state space is the global material state formed by splicing the material states and associated features at different spatial locations in the target device, and the action space includes the parameter configuration set of the actuator, and the control efficiency, energy consumption and stability are used as reward functions.

[0094] Based on the global material state and related characteristics, a global control strategy is output. At the same time, with the global control strategy as a constraint, a local control strategy is output for the spatial location where the material state is abnormal. In addition, a time-series control strategy is output according to the changes in the material state to generate a control strategy.

[0095] The prediction uncertainty increment is added to the state space, and the regulation strategy is screened and optimized by a genetic algorithm to obtain the optimized regulation strategy. The feasibility of the optimized regulation strategy is then verified to correct the regulation strategy.

[0096] The mixing of sugar and water requires optimized control of efficiency, energy consumption, and stability. Traditional control methods struggle to achieve dynamic equilibrium, necessitating reinforcement learning to search for the optimal strategy combination in the state space. The state space is composed of the material state and associated features at various locations within the mixing tank. When the sugar concentration at the bottom of the mixing tank is below the target value and the associated features indicate a high weight for the sugar addition rate, the state space highlights the abnormal state in that area. The action space includes discretized combinations of execution parameters such as the opening degree of the sugar tank valve and the stirring speed. The stirring speed is divided into three levels (low, medium, and high), which, combined with the sugar flow rate (fast, medium, and slow), form nine action options. The reward function is designed as a weighted sum of control efficiency, energy consumption, and stability. Reinforcement learning can automatically explore the optimal parameter combination, adapting to the dynamic changes in execution parameters such as the opening degree of the sugar tank valve and the stirring speed during the sugar and water mixing process. The constructed reinforcement learning provides an optimization objective for the generation of hierarchical control strategies, ensuring the synergy of global, local, and temporal control strategies.

[0097] During the mixing process, global uniformity, local anomaly handling, and timing lag require the coordinated action of strategies at different levels. The global strategy ensures overall concentration meets the target, the local strategy eliminates bottom clumping, and the timing strategy compensates for the delay effect of parameter adjustments. Based on the overall sugar concentration distribution and correlation characteristics, a global control strategy is generated. When the global average sugar concentration is less than 20% of the target value, a global command to increase the sugar flow rate to 120% of the rated value is output to quickly increase the overall concentration. For the bottom region where clumping is detected in the mixing tank, a local control strategy is generated, constrained by the global control strategy, to increase the sugar flow rate globally. Simultaneously, pulse vibrations are added to the bottom region to eliminate clumping. Considering the lag effect of parameter adjustment, it takes 5 minutes for the sugar to completely dissolve after addition. A strategy sequence with a time window is generated, that is, sugar is added quickly at a high flow rate first, and high-speed stirring is started after 5 minutes to ensure dissolution, ensuring that the timing of actions and effects is aligned. The hierarchical control strategy realizes three-dimensional control from global optimization to local repair and then to temporal smoothing, avoiding global imbalance or untimely local processing caused by a single strategy. The generated control strategy provides a rich set of candidate solutions for the optimization of the genetic algorithm, improving the diversity and robustness of the control strategy.

[0098] Factors such as sugar particle size fluctuations and sensor noise can lead to uncertainties in the state space. Sugar particle size fluctuations include differences in the dissolution rates of coarse and fine sugars, requiring quantification of risks during the optimization of the control strategy to avoid aggressive decisions in high-uncertainty scenarios. The predicted uncertainty output from the modal perception module is added to the state space as input for reinforcement learning. When missing sugar particle size detection data leads to high uncertainty in the dissolution time prediction, the control system automatically switches to a conservative strategy, i.e., extending the stirring time by 10%. The control strategy generated by reinforcement learning is further optimized using a genetic algorithm, prioritizing control strategies with strong stability in high-uncertainty scenarios. Specifically, in conditions with large water temperature fluctuations, a control strategy of first stirring at a low speed and then gradually increasing the speed is selected. The feasibility of the optimized control strategy is verified by checking whether the motor power exceeds the limit and whether the pipeline is blocked. The consideration of uncertainty allows the control system to dynamically adjust the aggressiveness of the control strategy based on data reliability, avoiding blind optimization in high-risk scenarios. After feasibility verification, the optimized control strategy is output to the dynamic control module for execution, ensuring the safety and operability of the control strategy.

[0099] The dynamic control module is used to control the materials in the target equipment according to the control strategy through the actuator, monitor the control effect, match the actuator impedance based on the material structure characteristics, and adjust the actuator's execution parameters. At the same time, it compares the control effect with the target effect to obtain the effect deviation, and determines whether to update the control strategy and the material status at different locations in the target equipment based on the effect deviation.

[0100] Furthermore, such as Figure 3 As shown, the matching sub-logic for actuator impedance includes:

[0101] The material structure features are received and decomposed into component features through wavelet transform. The component features include micro-features, macro-features and meso-features.

[0102] Actuator impedance includes mass, damping, and stiffness. A mapping relationship between component characteristics and actuator impedance is established based on historical control data to determine the adjustment direction and adjustment range of actuator impedance.

[0103] During material handling, the changes in material state and the operation feedback of the actuator are monitored in real time, so as to dynamically adjust the mapping relationship between the actuator impedance and component characteristics and the actuator impedance through a dual time scale control mechanism.

[0104] Different materials exhibit significant differences in structural characteristics, making it difficult to adapt to a single control method. Decomposing the material's structural features into microscopic, mesoscopic, and macroscopic dimensions allows for more precise matching of actuator impedance. After acquiring the material's structural features through a modal sensing module, wavelet transform technology is used for multi-scale decomposition to obtain component features. These component features include microscopic, macroscopic, and mesoscopic characteristics. For granular materials like sugar, the microscopic level analyzes particle surface roughness and particle size distribution, the mesoscopic level studies particle agglomerate size and morphology, and the macroscopic level focuses on material bulk density and flowability. For water, feature decomposition is performed from molecular structure (microscopic), water flow vortex morphology (mesoscopic), and overall liquid level height (macroscopic). This achieves refined analysis of the material's structural features, providing multi-dimensional basis for subsequent actuator impedance adjustment and avoiding a single control method. The decomposed component features provide a foundation for establishing a mapping relationship with actuator impedance, making impedance adjustment more targeted.

[0105] The mass, damping, and stiffness parameters of the actuator significantly affect the material handling effect. A corresponding relationship needs to be established based on the material's structural characteristics to ensure that the actuator impedance matches these characteristics. A mapping relationship between component characteristics and actuator impedance is established based on historical control data. When sugar agglomeration (mesoscopic anomaly) is detected, the actuator stiffness needs to be increased to break up the agglomerates. When the water flow is unstable (macroscopic anomaly), the damping parameters need to be adjusted to stabilize the water flow. The control system uses these mapping relationships to determine the direction and approximate magnitude of actuator impedance adjustment. This allows for the rapid identification of the actuator impedance that best matches the current material state, improving control efficiency and reducing equipment wear or poor material handling results caused by actuator impedance mismatch. A clear mapping relationship provides a reference standard for real-time dynamic adjustment of actuator impedance, making the adjustment process more efficient and accurate.

[0106] The material state changes in real time during production, and the rate of change varies, requiring a dual-timescale control mechanism to balance rapid response and long-term optimization. On the rapid timescale, the control system monitors material state changes and actuator operation feedback in real time, including motor current and vibration frequency. If an anomaly is detected, preferably a blockage at the sugar tank outlet causing a sudden increase in motor current, a small-scale impedance parameter adjustment is immediately performed to instantly increase stiffness and quickly resolve the immediate problem. On the slow timescale, based on long-term operating data, a deep reinforcement learning algorithm optimizes the mapping relationship. If adjusting the impedance according to the original mapping is found to be ineffective under certain operating conditions, the control system automatically learns and corrects the corresponding relationship, making the actuator impedance matching more adaptable to complex operating conditions. The rapid timescale ensures the control system's timely response to sudden situations, while the slow timescale optimizes long-term operating conditions, improving the stability and adaptability of the control system. The dynamically adjusted actuator impedance provides a stable basis for adjusting execution parameters, and the data accumulated during the adjustment process also provides a reference for updating the control strategy.

[0107] Specifically, the logic for adjusting the execution parameters includes:

[0108] The execution parameters of the actuator in the control strategy are extracted, and the execution parameters are decomposed into basic parameters and modulation parameters. The adjustment range of the execution parameters is determined by combining the prediction uncertainty, and the confidence interval of the adjusted execution parameters is calculated by Monte Carlo algorithm.

[0109] Based on process requirements, historical control data, and material status, the basic parameters and modulation parameters are prioritized, and the execution parameters are adjusted according to the priority order. The feedback on the adjustment effect of the execution parameters is monitored, and the execution parameters are readjusted based on the confidence interval of the adjusted execution parameters.

[0110] After each adjustment of the execution parameters, the coordination between the basic parameters and the modulation parameters is checked to trigger a reverse adjustment based on priority order, and the effect of each adjustment is updated and recorded in the historical control data.

[0111] The execution parameters are diverse in type and have different effects. Decomposing them allows for precise adjustments based on their characteristics, while considering predictive uncertainty helps avoid over-adjustment due to inaccurate data. The execution parameters in the control strategy are broken down into basic parameters and modulation parameters. Basic parameters directly affect material conveying and processing, such as valve opening, while modulation parameters fine-tune the control effect, such as the fluctuation frequency of the stirring motor speed. Combining the predictive uncertainty output by the modal sensing module, adjustment sensitivities are set for different execution parameters. In areas of high predictive uncertainty, the adjustment range of basic parameters is smaller, while modulation parameters are cautiously fine-tuned. In areas of low uncertainty, the adjustment range can be appropriately increased. The confidence interval for adjusting the execution parameters is calculated using a Monte Carlo algorithm to quantify the reliability of parameter adjustments. This enables refined adjustment of execution parameters by category, reducing the control risk caused by data uncertainty and improving control accuracy. The determined adjustment range and confidence interval provide a quantitative basis for subsequent parameter prioritization and adjustment.

[0112] During different stages and operating conditions in the production process, the impact of each execution parameter on the control effect varies. Prioritizing the adjustment of key parameters can improve control efficiency. Based on process requirements, historical control data, and real-time material status, basic and modulation parameters are prioritized. When material blockage occurs, the opening of the sugar tank valve and the vibration frequency of the vibrator are given higher priority. When it is necessary to improve the mixing uniformity, the stirring motor speed is given higher priority. Execution parameters are adjusted according to priority, and the feedback of the adjusted effect is monitored in real time. If the effect does not meet expectations, a second adjustment is made based on the confidence interval of the adjusted parameter. This ensures that key issues are addressed first under complex operating conditions, avoids blindly adjusting parameters which could lead to instability in the control system, and improves control efficiency and effectiveness. The feedback of the adjusted execution parameters is used to check the coordination between parameters and to accumulate data for updating the control strategy.

[0113] The execution parameters are interconnected; adjusting a single parameter can affect the effects of other parameters. Checking their coordination can prevent conflicts between parameters and ensure the stable operation of the control system. After each adjustment of the execution parameters, the control system automatically checks the coordination between the basic parameters and the modulation parameters. When increasing the opening of the sugar tank valve to increase the sugar addition rate, it checks whether the stirring motor speed matches the speed. If the speed is too low, it will cause material accumulation. At the same time, it checks whether the vibration frequency of the vibrator needs to be adjusted accordingly to prevent blockage. If incoordination between execution parameters is found, a reverse adjustment based on priority is triggered. If the stirring motor load is too high, the valve opening is appropriately reduced or the speed is increased to ensure the stable operation of the control system. After the adjustment is completed, the effect of this adjustment is updated and recorded in the historical control data to provide a reference for subsequent adjustments. This allows for the timely detection and resolution of conflicts between execution parameters, ensuring the stable operation of the control system. At the same time, by recording feedback data, the control strategy is continuously optimized. The coordinated execution parameters ensure the effective execution of the current control strategy, and the accumulated data provides a basis for updating the control strategy and the material status.

[0114] Specifically, the update logic of the regulatory strategy includes:

[0115] The deviation between the control effect and the target effect is calculated by Mahalanobis distance, and the correlation between the deviations is measured. The deviations include state deviation, energy consumption deviation and stability deviation.

[0116] A decision tree is constructed based on historical control data. The direction of adjustment of the control strategy is generated based on the combination of effect deviations. The decision tree is then optimized periodically based on the success rate of updating the control strategy.

[0117] The constraints for adjusting the execution parameters are determined based on the physical limitations of the actuator and the properties of the material, and the execution parameters are adjusted iteratively through adaptive step size.

[0118] Real-time monitoring of the control effect of adjusted execution parameters on materials in the target equipment, determination of effect deviation, and closed-loop optimization of control strategies.

[0119] To determine the effectiveness of the current control strategy, it is necessary to compare the actual control effect with the target effect, calculate the deviation, and analyze its correlation to provide direction for adjusting the control strategy. The effect deviation between the control effect and the target effect is calculated using Mahalanobis distance, encompassing state deviation, energy consumption deviation, and stability deviation. State deviation includes the difference between the actual and target material density; energy consumption deviation includes the comparison between actual and expected energy consumption; and stability deviation includes the vibration amplitude and parameter fluctuations of the equipment. Simultaneously, the correlation between each effect deviation is measured. If a large material density deviation is found to be accompanied by high energy consumption, and the correlation between the two is strong, it indicates a dual problem caused by an unreasonable control strategy. This comprehensively quantifies the effectiveness of the control strategy, identifies the root cause of the problem, and avoids misjudgments caused by analysis of a single indicator. The calculated effect deviation and correlation analysis results provide data support for constructing a decision tree and determining the direction for adjusting the control strategy.

[0120] Historical control data contains a wealth of experience, which can be transformed into rules by constructing a decision tree to quickly determine the direction of control strategy adjustments. The decision tree, built based on historical control data, uses combinations of effect deviations as decision nodes. If the state deviation exceeds the state deviation threshold and energy consumption deviation is also high, the decision tree directs the adjustment of basic parameters to optimize material handling efficiency. If stability deviation is significant, it directs the inspection of equipment operating parameters and adjustment of modulation parameters. Based on the currently calculated effect deviation, the decision tree automatically generates the direction of control strategy adjustments. Simultaneously, the control system periodically optimizes the decision tree based on the success rate of control strategy updates, deleting invalid branches and adding new effective rules to better align with actual production needs. Transforming historical experience into actionable rules improves the targeting and efficiency of control strategy adjustments, reduces trial-and-error costs, and provides clear guidance for determining the constraints of execution parameter adjustments and iteratively optimizing the strategy.

[0121] Based on the adjustment direction of the control strategy, and under the constraints of actuator physical limitations and material properties, the execution parameters are adjusted, and closed-loop monitoring ensures continuous optimization of the control strategy. The constraints for adjusting the execution parameters are determined based on the actuator's physical limitations and material properties. The actuator's physical limitations include the maximum motor power and the maximum valve opening, while the material properties include the maximum pressure and temperature range the material can withstand. An adaptive step-size iterative adjustment of the execution parameters is adopted. Initially, a larger step size is used to quickly approach the target; subsequently, the step size is gradually reduced for fine-tuning based on feedback. The control effect of the adjusted execution parameters on the material in the target equipment is monitored in real time, and the effect deviation is recalculated. If the effect deviation still does not reach the target range, the execution parameters are adjusted again, forming a closed-loop optimization until a satisfactory control effect is achieved. This ensures continuous optimization of the control strategy while maintaining equipment safety and material quality, improving the stability and production efficiency of the control system. The optimized control strategy provides a guarantee for the continuous operation of the dynamic control module, and the data during the adjustment process also provides a reference for updating the material status.

[0122] Specifically, the material status update mechanism includes:

[0123] Configure a deviation threshold. When any effect deviation exceeds the deviation threshold, calculate the causal probability that the effect deviation is caused by the change in material state based on the causal relationship between the process parameter set and material characteristics.

[0124] The explanatory power of the parameters that verify the effect deviation is verified by adjusting the execution parameters, and the update status of the material status is judged by combining the causal probability and the explanatory power of the parameters.

[0125] When the material status needs to be updated, the update step size of the material status is dynamically adjusted through fuzzy logic control. After the material status is updated, the effect deviation is recalculated to iteratively optimize the material status.

[0126] Not all deviations require updating the material state. By setting thresholds to filter out significant deviations and analyzing their causal relationship with changes in material state, unnecessary update operations can be avoided. Deviation thresholds are pre-configured. When any deviation exceeds the corresponding threshold, the subsequent process is triggered. Based on the causal relationship between the process parameter set and material characteristics established by the process association module, the causal probability of the deviation being caused by a change in material state is calculated. If a material density deviation is detected to exceed the standard, the control system queries the causal relationship to analyze whether it is caused by process parameter adjustments or changes in the material's own state, and calculates the corresponding causal probability. This allows for the accurate identification of deviations requiring attention, reducing invalid calculations and erroneous operations, improving the operating efficiency of the control system, and providing an important basis for determining whether to update the material state.

[0127] Some performance deviations can be resolved by adjusting execution parameters. However, it is necessary to verify the explanatory power of the parameters and comprehensively determine whether a material state update is truly necessary. By fine-tuning the execution parameters and observing the changes in performance deviations, the explanatory power of the parameters can be verified. If the material density deviation decreases significantly after appropriately reducing the sugar addition rate, it indicates that the deviation can be resolved by parameter adjustment, and the material state does not need to be updated. If the deviation does not change significantly, the causal probability is used to determine whether a material state update is necessary. When the causal probability is high and the parameter explanatory power is low, it is determined that a material state update is necessary; otherwise, the current state is maintained. This avoids erroneous material state updates due to misjudgment, ensures the stable operation of the control system, and reduces unnecessary control operations. A confirmed material state update decision initiates subsequent state update operations or maintains the existing control strategy.

[0128] During the material status update process, the update step size is difficult to fix under different operating conditions. Fuzzy logic control can dynamically adjust it according to the actual situation to ensure a smooth and effective update process. When it is determined that the material status needs to be updated, the update step size is dynamically adjusted by the fuzzy logic controller. The fuzzy logic controller uses the magnitude of the effect deviation and the causal probability as input variables. If the effect deviation is large and the causal probability is high, a larger update step size is output to speed up the material status adjustment. If the effect deviation is small, a smaller step size is used for fine-tuning. After updating the material status, the effect deviation is recalculated. If it still does not reach the target range, the update step size is adjusted again by fuzzy logic control for iterative optimization until the material status meets the production requirements. This achieves adaptive adjustment of the material status update step size, avoids system fluctuations caused by improper step size, and improves the accuracy and stability of material status adjustment. The updated material status is fed back to the modal perception module and the process association module, providing the latest data for subsequent control and serving as a reference for further optimization of control strategies.

[0129] Example 2

[0130] like Figure 4 The diagram shown illustrates a method flowchart for controlling a material dynamic control system, as provided in this application embodiment. The method includes:

[0131] Obtain modal data of materials in the target equipment, and extract material characteristics from the modal data as material density distribution and material structure characteristics;

[0132] Material density distribution and material structure features are fused to generate a feature tensor, and the material state at different spatial locations in the target equipment is determined by spatiotemporal coding.

[0133] The weight matrix of the process parameter set and feature tensor is determined based on the attention mechanism. The process parameter set and feature tensor are then multiplied by tensor based on the weight matrix to generate the associated tensor.

[0134] The correlation tensor is fused and processed with the process knowledge graph to output correlation features, and control strategies are generated by combining the material state at different spatial locations.

[0135] The actuator controls the material in the target equipment according to the control strategy, monitors the control effect, matches the actuator impedance based on the material structure characteristics, and adjusts the actuator's execution parameters.

[0136] By comparing the control effect with the target effect, the effect deviation is obtained. Based on the effect deviation, it is determined whether to update the control strategy and the material state at different spatial locations in the target equipment.

[0137] Since the principle of the method in this application embodiment is similar to that of the system described in this application embodiment, the implementation of the method is the same as that of the system, and the repeated parts will not be described again.

Claims

1. A material dynamic control system, characterized in that, include: Modal data of materials in the target equipment are acquired, and the material features of the modal data are extracted as material density distribution and material structure features. The material density distribution and material structure features are fused to generate a feature tensor, and the material state at different spatial locations in the target equipment is determined by spatiotemporal coding. The weight matrix of the process parameter set and feature tensor is determined based on the attention mechanism. The process parameter set and feature tensor are multiplied by the weight matrix to generate the association tensor. The association tensor is then fused and processed with the process knowledge graph to output the association features. At the same time, the control strategy is generated by combining the material state at different spatial locations. The actuator controls the material in the target equipment according to the control strategy, monitors the control effect, matches the actuator impedance based on the material structure characteristics, and adjusts the actuator's execution parameters. At the same time, the control effect is compared with the target effect to obtain the effect deviation. Based on the effect deviation, it is determined whether to update the control strategy and the material status at different locations in the target equipment.

2. The material dynamic control system as described in claim 1, characterized in that, The logic for generating the feature tensor includes: Modal data of materials in the target equipment are acquired, and the modal data are preprocessed using an adaptive noise reduction algorithm; The preprocessed modal data is mapped to the density feature space and the structural feature space by a variational autoencoder. The density changes of the modal data are extracted by a convolutional neural network to obtain the material density distribution. The topological relationship between the particles of the material is analyzed by a graph neural network to obtain the material structural features. Feature weights for material density distribution and material structure characteristics are generated through an attention mechanism, and feature tensors are generated through weighted summation.

3. The material dynamic control system as described in claim 2, characterized in that, The logic for determining the state of the material includes: Assign spatiotemporal coordinate information to the feature tensor, mark the spatial location and time series of the material in the target equipment, and perform feature interaction on the feature tensor through a multi-head attention mechanism to extract the change features of the material in spatial location and time series; The material state of the target device at different spatial locations is predicted based on the changing characteristics using a Bayesian neural network, and the prediction uncertainty is output. Local and global features are extracted from the feature tensor. Fusion weights are generated based on the predicted material state and prediction uncertainty. The local and global features are then fused according to the fusion weights to obtain fused features. The fused features are then mapped to determine the material state of the target device at different spatial locations.

4. The material dynamic control system as described in claim 3, characterized in that, The output logic for the associated features includes: Receive the feature tensor and obtain the process parameter set. Calculate the correlation weights between the process parameter set and the feature tensor in the spatial dimension, temporal dimension, and feature dimension through the attention mechanism. Multiply the correlation weights by the weight matrix of the process parameter set and the feature tensor. The causal relationship between the process parameter set and material characteristics is identified by the causal reasoning network, and the weight matrix of the process parameter set and feature tensor is corrected according to the causal relationship. Tensor product operation is performed on the process parameter set and feature tensor based on the corrected weight matrix to generate the association tensor. The association tensor is then fused with the process knowledge graph. The fused process knowledge graph is then inferred through a graph neural network to output the association features.

5. The material dynamic control system as described in claim 4, characterized in that, The generation logic of the regulation strategy includes: Reinforcement learning is constructed, where the state space is the global material state formed by splicing the material states and associated features at different spatial locations in the target device, and the action space includes the parameter configuration set of the actuator, and the control efficiency, energy consumption and stability are used as reward functions. Based on the global material state and related characteristics, a global control strategy is output. At the same time, with the global control strategy as a constraint, a local control strategy is output for the spatial location where the material state is abnormal. In addition, a time-series control strategy is output according to the changes in the material state to generate a control strategy. The prediction uncertainty increment is added to the state space, and the regulation strategy is screened and optimized by a genetic algorithm to obtain the optimized regulation strategy. The feasibility of the optimized regulation strategy is then verified to correct the regulation strategy.

6. The material dynamic control system as described in claim 5, characterized in that, The logic for adjusting the execution parameters includes: The execution parameters of the actuator in the control strategy are extracted, and the execution parameters are decomposed into basic parameters and modulation parameters. The adjustment range of the execution parameters is determined by combining the prediction uncertainty, and the confidence interval of the adjusted execution parameters is calculated by Monte Carlo algorithm. Based on process requirements, historical control data, and material status, the basic parameters and modulation parameters are prioritized, and the execution parameters are adjusted according to the priority order. The feedback on the adjustment effect of the execution parameters is monitored, and the execution parameters are readjusted based on the confidence interval of the adjusted execution parameters. After each adjustment of the execution parameters, the coordination between the basic parameters and the modulation parameters is checked to trigger a reverse adjustment based on priority order, and the effect of each adjustment is updated and recorded in the historical control data.

7. The material dynamic control system as described in claim 6, characterized in that, The matching sub-logic for the actuator impedance includes: The material structure features are received and decomposed into component features through wavelet transform. The component features include micro-features, macro-features and meso-features. Actuator impedance includes mass, damping, and stiffness. A mapping relationship between component characteristics and actuator impedance is established based on historical control data to determine the adjustment direction and adjustment range of actuator impedance. During material handling, the changes in material state and the operation feedback of the actuator are monitored in real time, so as to dynamically adjust the mapping relationship between the actuator impedance and component characteristics and the actuator impedance through a dual time scale control mechanism.

8. The material dynamic control system as described in claim 7, characterized in that, The update logic of the control strategy includes: The deviation between the control effect and the target effect is calculated by Mahalanobis distance, and the correlation between the deviations is measured. The deviations include state deviation, energy consumption deviation and stability deviation. A decision tree is constructed based on historical control data. The direction of adjustment of the control strategy is generated based on the combination of effect deviations. The decision tree is then optimized periodically based on the success rate of updating the control strategy. The constraints for adjusting the execution parameters are determined based on the physical limitations of the actuator and the properties of the material, and the execution parameters are adjusted iteratively through adaptive step size. Real-time monitoring of the control effect of adjusted execution parameters on materials in the target equipment, determination of effect deviation, and closed-loop optimization of control strategies.

9. The material dynamic control system as described in claim 8, characterized in that, The material status update mechanism includes: Configure a deviation threshold. When any effect deviation exceeds the deviation threshold, calculate the causal probability that the effect deviation is caused by the change in the material state based on the causal relationship between the process parameter set and the material characteristics. The explanatory power of the parameters that verify the effect deviation is verified by adjusting the execution parameters, and the update status of the material status is judged by combining the causal probability and the explanatory power of the parameters. When the material status needs to be updated, the update step size of the material status is dynamically adjusted through fuzzy logic control. After the material status is updated, the effect deviation is recalculated to iteratively optimize the material status.

10. A control method for a material dynamic control system, characterized in that, include: Obtain modal data of materials in the target equipment, and extract material characteristics from the modal data as material density distribution and material structure characteristics; Material density distribution and material structure features are fused to generate a feature tensor, and the material state at different spatial locations in the target equipment is determined by spatiotemporal coding. The weight matrix of the process parameter set and feature tensor is determined based on the attention mechanism. The process parameter set and feature tensor are then multiplied by tensor based on the weight matrix to generate the associated tensor. The correlation tensor is fused and processed with the process knowledge graph to output correlation features, and control strategies are generated by combining the material state at different spatial locations. The actuator controls the material in the target equipment according to the control strategy, monitors the control effect, matches the actuator impedance based on the material structure characteristics, and adjusts the actuator's execution parameters. By comparing the control effect with the target effect, the effect deviation is obtained. Based on the effect deviation, it is determined whether to update the control strategy and the material state at different spatial locations in the target equipment.

Images