A remote intelligent control system and device for crusher operation

Abstract

This invention belongs to the field of remote control technology for industrial equipment. It discloses a remote intelligent control system and device for crusher operation, including a dynamic sensing field module. This module employs a fusion deployment of distributed fiber optic vibration sensors and a 3D-printed flexible strain sensor array to form a dense monitoring network at key locations in the crushing chamber. It collects equipment operation data and reconstructs the field using a field reconstruction algorithm based on a graph neural network to generate a vibration energy topology map. A cognitive twin evolution module constructs an industrial knowledge graph based on the Neo4j graph database, integrates historical equipment fault databases and equipment operation data, and combines a Transformer-XL model to perform temporal pattern mining, obtaining the equipment's temporal behavior embedding vector. This improves the system's responsiveness and adaptive optimization capabilities under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of remote control technology for industrial equipment, and more specifically, to a remote intelligent control system and device for crusher operation. Background Technology

[0002] Patent publication number CN108325728A discloses a crusher control system, including a main controller, a temperature detection module, a speed detection module, amplifier A, amplifier B, an I / O interface module, an alarm, a main power supply, a backup power supply, a power supply detection and control module, an input module, and a crushing motor. This invention's crusher control system uses a microcontroller as its core component, enabling effective real-time monitoring of the crushing motor's operating status during crusher operation. When the motor temperature is too high or the motor speed becomes abnormal, it can issue an alarm signal and stop the crusher's operation to prevent continuous damage. Furthermore, this design employs dual power supplies, controlling both power sources through power monitoring technology to ensure the normal operation of the equipment. Using this system can effectively extend the crusher's service life.

[0003] Existing remote intelligent control systems and devices for crusher operation mainly have the following problems:

[0004] In existing technologies, controller parameters are fixed, making it difficult to dynamically optimize control accuracy based on operational deviations and fault states. Existing methods rely solely on fault prediction reports or vibration characteristics for risk assessment, lacking a multi-source fusion evaluation mechanism. Fixed risk fusion strategies cannot balance the complementarity between prediction uncertainty and vibration response, leading to the risk of misjudgment or omission. Existing methods cannot automatically adjust confidence levels based on information source confidence, resulting in control strategy failure or instability. When predicting crusher operational risk scores, if the fusion coefficient in the operational risk function is fixed too large, the system will overly rely on the topological propagation intensity of vibration energy, potentially becoming insensitive even if the fault prediction report indicates a high risk. If the fusion coefficient is set too small, it will not respond to changes in the vibration spectrum, potentially missing early warning opportunities for local vibration anomalies and easily leading to misjudgment or omission.

[0005] Existing methods often prioritize control commands based on static indicators, lacking comprehensive modeling of multiple factors such as operational risks, disturbance responses, and task pressure. This results in fixed control strength adjustments that cannot be dynamically adjusted in response to environmental disturbances. Traditional control logic struggles to simulate nonlinear activation / inhibition mechanisms similar to those in biological systems, leading to insufficient response or inappropriate suppression in high-risk or high-pressure task scenarios. Existing technologies typically weight multiple indicators at fixed ratios, failing to introduce adaptive weighting mechanisms based on changes in task background and operational status.

[0006] In view of this, the present invention proposes a remote intelligent control system and device for crusher operation to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: a remote intelligent control system for crusher operation, comprising:

[0008] The dynamic sensing field module employs a fusion deployment of distributed fiber optic vibration sensors and a 3D-printed flexible strain sensor array to form a dense monitoring network at key locations in the crushing chamber. It collects equipment operation data and uses a field reconstruction algorithm based on graph neural networks to reconstruct the field of the equipment operation data, generating a vibration energy topology map.

[0009] The cognitive twin evolution module constructs an industrial knowledge graph based on the Neo4j graph database, integrates the equipment historical fault database and equipment operation data, and performs temporal pattern mining using the Transformer-XL model to obtain the temporal behavior embedding vector of the equipment. Based on the temporal behavior embedding vector, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory to generate a fault prediction report with confidence assessment.

[0010] The adaptive robust execution module is based on a fuzzy PID controller to build the execution core. It combines fault prediction reports and vibration energy topology maps to dynamically sense equipment operation risks and adaptively adjust control parameters to generate control commands. It also introduces a metabolic heuristic algorithm to adjust the control intensity and dynamically adjust the priority of control commands.

[0011] The multi-scale verification module uses a real-time simulator to build a hardware-in-the-loop test platform, integrates FPGA programmable fault injection tools, and generates a deviation analysis report on virtual debugging and control command execution.

[0012] The dynamic meta-optimization module constructs a strategy optimization mainline based on the MLOps architecture, integrates vibration energy topology diagrams, fault prediction reports, control commands, and deviation analysis reports, and uses a meta-model selector to quickly determine the current operating conditions and execute the corresponding control commands.

[0013] Preferably, the method for forming the dense monitoring network includes:

[0014] Sensor channels and integrated interfaces are pre-set at key locations in the crushing chamber to form a physical platform for sensor deployment. Key locations in the crushing chamber include the inner wall of the crushing chamber, the liner support structure, the main bearing connection, and the edges of the inlet and outlet. Fiber optic lines are laid around the outer perimeter of the crushing chamber and the main structure to form a continuous and seamless vibration sensing link. A composite material with conductive and elastic properties is used as a flexible substrate. Strain sensors are directly printed onto the flexible substrate in the pre-selected installation area using 3D printing technology to form a flexible strain sensor array that adaptively conforms to the curved surface.

[0015] A micro edge computing module is deployed between a distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array to perform edge preprocessing and multimodal fusion of distributed signals, thereby forming a dense monitoring network for multimodal fusion within the fracture cavity.

[0016] Preferably, the method for generating the vibration energy topology map includes:

[0017] Real-time equipment operation data is collected by a distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array deployed in the crushing chamber. The equipment operation data includes dynamic mechanical status data, electrical energy consumption data, and thermal lubrication status data. Each location in the crushing chamber where a sensor is deployed is defined as a vibration energy topology graph node, and edges between all vibration energy topology graph nodes are constructed based on physical adjacency relationships.

[0018] The device operation data collected by the sensors corresponding to each vibration energy topology graph node are preprocessed to obtain the node feature matrix; the vibration energy topology graph structure is constructed based on the obtained vibration energy topology graph node, edge and node feature matrix; the vibration energy topology graph structure is used as the input of the graph neural network, and the node feature fusion representation of each vibration energy topology graph node is obtained by spatiotemporal reconstruction of the node features of the vibration energy topology graph nodes;

[0019] Energy estimation is performed on the node feature fusion representation of each vibration energy topology graph node to obtain the vibration energy distribution value of each vibration energy topology graph node; the vibration energy distribution value of each vibration energy topology graph node is mapped to the vibration energy topology graph structure to form the vibration energy topology graph.

[0020] Preferably, the method for constructing the industrial knowledge graph includes:

[0021] From equipment operation data, semantically meaningful knowledge elements are extracted, including entity elements, attribute elements, and relation elements. Based on the extracted knowledge elements, a data model for an industrial knowledge graph is designed, which includes nodes, edges, and attributes.

[0022] Neo4j is used as the graph database platform to store the industrial knowledge graph structure. Its native graph engine is used for graph traversal, path analysis and pattern matching. A retrieval mechanism for key entities is established through attribute indexing and full-text indexing. Key entities include fault type, operating condition label, equipment component name and status event.

[0023] A business expert rule base is introduced to define the relation weights and node credibility indicators in the industrial knowledge graph. Combined with rule-based logical rules, the reasoning capabilities of the graph are gradually expanded. Based on Neo4j Bloom, the visualization of the industrial knowledge graph structure is supported to obtain the final industrial knowledge graph.

[0024] Preferably, the method for obtaining the timing behavior embedding vector of the device includes:

[0025] The equipment historical fault database includes fault type, fault occurrence time, fault duration, fault location data, and fault repair records; a Transformer-XL model is built and trained based on the equipment historical fault database and equipment operation data. The Transformer-XL model includes an embedding layer, a multi-head attention layer, a feedforward neural network layer, and an output layer.

[0026] The model's embedding layer is used to transform the device's historical fault database and historical device operation data into vector form, and the model's output layer is used to output the device's temporal behavior embedding vector. The mean absolute error is used as the loss function to measure the error of the model's prediction. The trained Transformer-XL model is used to output the device's temporal behavior embedding vector.

[0027] Preferably, the method for obtaining the fault prediction report with confidence assessment includes:

[0028] Based on temporal behavior embedding vectors, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory; the uncertainty reasoning engine includes a fusion mechanism based on Bayesian network structure and DS evidence theory.

[0029] The temporal behavior embedding vector is mapped to nodes in a Bayesian network. The nodes of the Bayesian network are designed based on the equipment operation data. Each node represents a potential fault type. The dependency structure between the nodes of the Bayesian network is learned using the equipment's historical fault database. The conditional probability table of each node of the Bayesian network is determined by maximum likelihood estimation.

[0030] The conditional probability table is mapped to the basic probability assignment function in the DS evidence theory. The Dempster composition rule is applied to combine the basic probability assignment functions to obtain the comprehensive confidence level for each fault type. Based on the comprehensive confidence level, a fault prediction report with confidence assessment is output.

[0031] Preferably, the method for generating the control commands includes:

[0032] The fault prediction report includes the predicted probability and overall confidence level for different fault types, and is received by the crusher through a fuzzy PID controller. Operational deviation at any time and rate of change of deviation ;

[0033] The fuzzy PID controller uses a fuzzy inference mechanism to... and As input, based on a preset fuzzy rule base, the output is the PID parameter correction amount. , and ; This indicates the correction amount for the proportional gain; This represents the proportional term in a fuzzy PID controller; This represents the correction amount for the integral gain; This represents the integral term in a fuzzy PID controller. This represents the correction amount for the differential gain; This represents the derivative term in a fuzzy PID controller.

[0034] Real-time updates of the proportional, integral, and derivative parameters of the fuzzy PID controller; construction of an operational risk function to predict the crusher's operational risk score; operational risk function. ;in, Indicates that the crusher is in Operational risk score at any given moment; Indicates the first Types of faults; Index representing the fault type; Indicates the first Confidence level of the prediction for each type of failure; Indicates the first Risk weighting coefficients for different fault types; This represents the topological propagation intensity of vibration energy within the equipment structure. This represents the time-dependent fusion coefficient;

[0035] The fusion coefficient is adjusted using the fusion coefficient adjustment formula. Make dynamic adjustments. ;in, This represents the information entropy of the current fault prediction. This represents the preset maximum information entropy;

[0036] A confidence threshold function is introduced to limit the confidence of fault types participating in the regulation of the fuzzy PID controller; the confidence threshold function is: ;in, This indicates the preset minimum confidence level; Indicates the first If the confidence level of a certain fault type prediction is greater than or equal to the preset minimum confidence level, that fault type participates in the fuzzy PID controller regulation. Indicates the first If the confidence level of a certain fault type is less than the preset minimum confidence level, that fault type will not participate in the fuzzy PID controller regulation.

[0037] Preferably, the method for dynamically adjusting the priority of control commands includes:

[0038] By analogy with the dynamic variables of biological metabolic output, a control strength formula is constructed to obtain the basic control strength. The nonlinear activation or inhibition effects in the metabolic pathway are simulated, and a nonlinear mapping function is used to adjust the basic control strength. At the same time, the stability index of the historical execution of control commands is introduced as a weighting factor to dynamically adjust and correct the priority of control commands.

[0039] Preferably, the method for quickly determining the current operating condition includes:

[0040] A multi-source data aggregation mechanism is constructed to uniformly aggregate the acquired vibration energy topology map, fault prediction report, control command and deviation analysis report as input data streams into the MLOps working pipeline to form a dynamic dataset; a strategy optimization process is set up on the MLOps mainline to uniformly encode and format the input dynamic dataset to obtain the characteristics of the current operating condition;

[0041] A meta-model selector is constructed to match control strategies from pre-trained strategy meta-models based on the characteristics of the current operating conditions, generating candidate control strategies. A real-time simulator is introduced to simulate and verify the candidate control strategies, select the optimal control strategy, and execute the corresponding control commands. The control commands are then sent to the corresponding crusher execution components in real time through the crusher execution terminal.

[0042] A remote intelligent control device for crusher operation includes a crusher, wherein the crusher is equipped with a remote intelligent control system for crusher operation that integrates a dynamic sensing field module, a cognitive twin evolution module, an adaptive robust execution module, a multi-scale verification module, and a dynamic meta-optimization module.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] This invention introduces information entropy from Shannon's information theory to measure the uncertainty of the prediction results of the running risk function in real time. By measuring the uncertainty of the function prediction through information entropy, the reliance on topological sensing data is increased when the prediction confidence is low, thereby enhancing the system's fault tolerance under abnormal conditions. In scenarios where the prediction data fluctuates greatly or the acquired signals are abnormal, the fusion mechanism can adaptively switch to a more reliable information source, thereby improving the overall robustness of risk assessment. It effectively realizes the fusion reasoning based on risk function prediction and vibration energy topology, enabling the system to establish a highly reliable adaptive feedback path between the perception layer and the cognition layer.

[0045] The confidence threshold mechanism significantly improves the stability and response accuracy of the control logic by eliminating sources of low reliability predictions. It links participation in control with information quality, realizing a confidence-driven control path pruning mechanism to avoid invalid or erroneous information from disturbing the controller and ensuring the robustness of the control loop. It is suitable for high-complexity operating scenarios where multiple potential faults coexist, preventing weak faults or false detection information from misleading the master control strategy and enhancing system security and engineering practicality.

[0046] By integrating multi-source dynamic indicators such as operational risk, structural disturbance response, and control task priority pressure, a basic control strength model is constructed. Differential regulation is achieved through weighting factors, enabling the strength and execution order of control commands to automatically change with the evolution of the system state, thereby improving the flexibility and adaptability of the control logic. A nonlinear adjustment function under the metabolic pathway activation mechanism is constructed, which can achieve rapid increase or gradual suppression of control strength near critical state points, enhancing the emergency response capability of the control system to extreme tasks, sudden disturbances, or high-risk states, and avoiding hysteresis or overshoot problems.

[0047] By introducing a historical execution stability index of control instructions as a priority weighting factor, the actual priority of instructions can be dynamically adjusted according to their historical execution performance. This reduces the problem of frequent instruction switching and scheduling drift caused by environmental changes, and significantly improves the continuity of system response and the controllability of instruction scheduling. By setting a nonlinear adjustment factor and a stability weighting mechanism, this method prioritizes the control of critical tasks under the condition of limited control resources, mitigates the response of non-critical tasks, reduces the overall resource conflict rate, and improves the orderliness of control behavior. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the structure of a remote intelligent control system for crusher operation according to the present invention;

[0049] Figure 2 This is a schematic diagram of a remote intelligent control method for crusher operation according to the present invention;

[0050] Figure 3This invention provides a roadmap for remote intelligent control technology for crusher operation. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] Example 1

[0053] Please see Figure 1 and Figure 3 As shown, this embodiment further illustrates the remote intelligent control system for crusher operation proposed in this invention, including:

[0054] With the development of industrial intelligence, the demand for remote operation monitoring and intelligent control of large-scale key equipment such as crushers is increasing. Existing remote control systems for crushers mostly rely on traditional control models and single data source-driven logic, and generally suffer from the following technical bottlenecks and challenges:

[0055] In existing control systems, controller parameters are typically set statically, making it difficult to dynamically adjust based on feedback. This leads to decreased control accuracy when equipment operating conditions change, making it unable to adapt to complex and variable operating environments. In particular, when there are operational deviations, fault symptoms, or changes in disturbance response, fixed-parameter controllers struggle to adjust control strategies in a timely manner, affecting the stability and safety of equipment operation.

[0056] Current risk assessment methods are mostly based on single information sources, such as fault prediction reports or vibration signal characteristics, lacking the ability to integrate multiple sources for reasoning. When faced with multiple coexisting faults, conflicting data, or incomplete information, existing methods cannot comprehensively assess the complementarity between prediction confidence and vibration response, easily leading to misjudgments or omissions. Furthermore, fixed-weighting strategies cannot dynamically adjust according to the reliability of information sources, resulting in high-confidence information being underutilized while low-confidence information may interfere with system decision-making.

[0057] In traditional systems, the fusion coefficient in the risk function is usually a fixed value, which leads to the following problems: when the fusion coefficient is too large, the system relies excessively on the topological propagation intensity of structural vibration energy, potentially making it insensitive to high-risk predictions; when the fusion coefficient is set too small, the advantages of the topological structure in early anomaly identification may be missed, affecting the timeliness and accuracy of warnings. Existing methods struggle to achieve adaptive adjustment of the fusion strategy, limiting the system's fault tolerance and robustness.

[0058] The generation and scheduling mechanisms of control commands lack dynamic priority models, with most systems using static weight settings, neglecting dynamic factors such as operating status, risk level, and task pressure. Adjustments to control strength are often linear and static, lacking nonlinear activation / inhibition mechanisms similar to those in biological systems. This leads to delayed responses under sudden risks or extreme conditions, and may even result in over-adjustment or control failure.

[0059] The lack of a correlation mechanism between instruction priority and historical execution stability means that existing methods cannot dynamically adjust the scheduling weights based on the stability or past performance of the information source. This leads to frequent switching of control strategies by the control system in response to environmental changes, causing scheduling drift, resource conflicts, and other problems, which reduce the overall system's execution efficiency and response continuity.

[0060] Therefore, there is an urgent need for a control system that integrates multi-source information, possesses the ability to handle predictive uncertainties, allows for adaptive adjustment of control parameters, and employs a nonlinear adjustment mechanism in its control strategy. To effectively address the aforementioned issues, this invention proposes a remote intelligent control system for crusher operation, comprising:

[0061] The dynamic sensing field module employs a fusion deployment of distributed fiber optic vibration sensors and a 3D-printed flexible strain sensor array to form a dense monitoring network at key locations in the crushing chamber. It collects equipment operation data and uses a field reconstruction algorithm based on graph neural networks to reconstruct the field of the equipment operation data, generating a vibration energy topology map.

[0062] The cognitive twin evolution module constructs an industrial knowledge graph based on the Neo4j graph database, integrates the equipment historical fault database and equipment operation data, and performs temporal pattern mining using the Transformer-XL model to obtain the temporal behavior embedding vector of the equipment. Based on the temporal behavior embedding vector, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory to generate a fault prediction report with confidence assessment.

[0063] The adaptive robust execution module is based on a fuzzy PID controller to build the execution core. It combines fault prediction reports and vibration energy topology maps to dynamically sense equipment operation risks and adaptively adjust control parameters to generate control commands. It also introduces a metabolic heuristic algorithm to adjust the control intensity and dynamically adjust the priority of control commands.

[0064] The multi-scale verification module uses a real-time simulator to build a hardware-in-the-loop test platform, integrates FPGA programmable fault injection tools, and generates a deviation analysis report on virtual debugging and control command execution.

[0065] The dynamic meta-optimization module constructs a strategy optimization mainline based on the MLOps architecture, integrates vibration energy topology diagrams, fault prediction reports, control commands, and deviation analysis reports, and uses a meta-model selector to quickly determine the current operating conditions and execute the corresponding control commands.

[0066] Methods for forming dense surveillance networks include:

[0067] Sensor channels and integrated interfaces are pre-installed at key locations within the crushing chamber, forming a physical platform for sensor deployment. The sensor channels feature a high-temperature isolation layer and a structural stress buffer layer, adapting to extreme conditions such as gravel impact and high-temperature dust. Key locations within the crushing chamber include the inner wall, liner support structure, main bearing connection, and inlet / outlet edges. Fiber optic lines are deployed around the crushing chamber and the main structure, forming a continuous and seamless vibration sensing link. The fiber optics are embedded or attached to the metal surface of the crushing chamber through structural colloid encapsulation, enabling simultaneous acquisition of multi-point vibration signals with centimeter-level spatial resolution. A flexible substrate (such as carbon nanotube / PDMS composite) with conductive and elastic properties is used. Strain sensors are directly printed onto the pre-selected installation area using 3D printing technology, forming an adaptive, curved surface flexible strain sensor array. This flexible strain sensor array can integrate multiple sensing channels to capture micro-deformation of the stressed structure, shear stress transmission paths, and local plastic fatigue characteristics in real time. It is suitable for crushing chamber areas with complex surface curvature and high requirements for local micro-vibration response sensitivity.

[0068] A micro edge computing module is deployed between a distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array to perform edge preprocessing and multimodal fusion of distributed signals, thereby forming a dense monitoring network with multimodal fusion within the crushing chamber. The dense monitoring network has the characteristics of high spatial resolution, fast response, strong adaptability and strong anti-interference ability, and can achieve full coverage, high precision and real-time perception of the structural operating status of the crushing chamber. It provides a solid data foundation for structural health prediction, fault precursor identification and maintenance decision-making of the crushing chamber, and is especially suitable for industrial environments with high-intensity impact, complex vibration and multi-source interference.

[0069] Methods for generating vibration energy topology maps include:

[0070] By employing a field reconstruction algorithm based on graph neural networks, the collected equipment operation data is transformed into a visualized and structured vibration energy topology map, enabling spatial perception reconstruction and energy flow modeling of the vibration state of the internal structure of the crushing chamber. This process mainly includes four stages: graph construction, feature embedding, graph neural network propagation, and energy decoding.

[0071] Real-time equipment operation data is collected using distributed fiber optic vibration sensors and a 3D-printed flexible strain sensor array deployed in the crushing chamber. This data includes dynamic mechanical status data, electrical energy consumption data, and thermal lubrication status data. Dynamic mechanical status data includes vibration signal data, rotational speed, impact intensity, impact acceleration (transient fluctuations caused by ore crushing behavior within the crushing chamber), and equipment structural components. Electrical energy consumption data includes equipment current, voltage, power, and motor load rate. Thermal lubrication status data includes bearing temperature, lubricating oil temperature, lubricating oil pressure, lubrication flow rate, and hydraulic pressure. Each sensor location within the crushing chamber is defined as a node in a vibration energy topology graph, and edges between all vibration energy topology graph nodes are constructed based on physical adjacency relationships.

[0072] The device operation data collected by the sensors corresponding to each vibration energy topology graph node are preprocessed to obtain the node feature matrix; the vibration energy topology graph structure is constructed based on the obtained vibration energy topology graph node, edge and node feature matrix; the vibration energy topology graph structure is used as the input of the graph neural network, and the node feature fusion representation of each vibration energy topology graph node is obtained by spatiotemporal reconstruction of the node features of the vibration energy topology graph nodes;

[0073] Energy estimation is performed on the node feature fusion representation of each vibration energy topology graph node to obtain the vibration energy distribution value of each vibration energy topology graph node; the vibration energy distribution value of each vibration energy topology graph node is mapped to the vibration energy topology graph structure to form the vibration energy topology graph.

[0074] Technical advantages and effects: It can form a high-resolution energy distribution map inside a complex geometric structure (fracture chamber); it realizes intelligent perception reconstruction from sensor data to the functional field of spatial structure, significantly improving the cognitive granularity of equipment operation status; it can support fault prediction and structural diagnosis based on vibration energy topology map, enhancing the accuracy of fault location and energy identification.

[0075] Methods for constructing industrial knowledge graphs include:

[0076] From equipment operation data, semantically meaningful knowledge elements are extracted. These elements include entity elements, attribute elements, and relational elements. Entity elements include equipment structural components (spindle, motor, bearing), operating parameters (speed, vibration frequency, temperature), and fault types (overload, abnormal vibration, wear). Attribute elements include the state parameters and performance indicators corresponding to the entity. Relational elements include causal, spatial, and temporal relationships such as "component – ​​belongs to – system," "fault – affects – component," and "parameter – change leads to – fault." Based on the extracted knowledge elements, a data model for an industrial knowledge graph is designed. The data model includes nodes, edges, and attributes. Nodes represent different types of industrial entities (e.g., components, states, faults, operating conditions). Edges represent semantic relationships between entities (e.g., "connected to," "leads to," "associated with," etc.). Attributes assign timestamps, probability values, parameter values, and other attribute information to nodes and edges.

[0077] For example, (Equipment {Name: "Crusher"}) - [Contains] -> (Component {Name: "Motor"}), (Fault {Type: "Overload"}) - [Occurs] -> (Component {Name: "Spindle"});

[0078] Neo4j is used as the graph database platform to store the industrial knowledge graph structure. Its native graph engine is used for graph traversal, path analysis, and pattern matching. A retrieval mechanism is established for key entities (such as fault types and operating condition tags) through attribute indexing and full-text indexing. Key entities include fault types (such as "bearing wear", "spindle jamming", "abnormal vibration", etc., which are from the historical fault database), operating condition tags (such as "high load", "frequent start-stop", "rapid temperature rise", etc., which are the core of status identification and risk labeling), equipment component names (such as "spindle", "motor", "liner", "bearing", which are important nodes in structural disassembly), and status events (such as "temperature exceeds threshold" and "sudden increase in vibration", which have attributes such as timestamp and threshold).

[0079] A business expert rule base is introduced to define the relation weights and node credibility indicators in the industrial knowledge graph. Combined with rule-based logical rules, the reasoning capabilities of the graph are gradually expanded. Based on Neo4j Bloom, the visualization of the industrial knowledge graph structure is supported to obtain the final industrial knowledge graph.

[0080] Methods for obtaining the device's timing behavior embedding vector include:

[0081] The equipment historical fault database includes fault type, fault occurrence time, fault duration, fault location data, and fault repair records; a Transformer-XL model is built and trained based on the equipment historical fault database and equipment operation data. The Transformer-XL model includes an embedding layer, a multi-head attention layer, a feedforward neural network layer, and an output layer.

[0082] The model's embedding layer is used to transform the device's historical fault database and historical device operation data into vector form, and the model's output layer is used to output the device's temporal behavior embedding vector. The mean absolute error is used as the loss function to measure the error of the model's prediction. The trained Transformer-XL model is used to output the device's temporal behavior embedding vector.

[0083] Methods for obtaining fault prediction reports with confidence assessment include:

[0084] Based on temporal behavior embedding vectors, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory; the uncertainty reasoning engine includes a fusion mechanism based on Bayesian network structure and DS evidence theory.

[0085] The temporal behavior embedding vector is mapped to nodes in a Bayesian network. The nodes of the Bayesian network are designed based on the equipment operation data. Each node represents a potential fault type. The dependency structure between the nodes of the Bayesian network is learned using the equipment's historical fault database. The conditional probability table of each node of the Bayesian network is determined by maximum likelihood estimation.

[0086] The conditional probability table is mapped to the basic probability assignment function in the DS evidence theory. The Dempster composition rule is applied to combine the basic probability assignment functions to obtain the comprehensive confidence level for each fault type. Based on the comprehensive confidence level, a fault prediction report with confidence assessment is output.

[0087] Methods for generating control commands include:

[0088] The fault prediction report includes the predicted probability and overall confidence level for different fault types, and is received by the crusher through a fuzzy PID controller. Operational deviation at any time and rate of change of deviation ;

[0089] The fuzzy PID controller uses a fuzzy inference mechanism to... and As input, based on a preset fuzzy rule base, the output is the PID parameter correction amount. , and ; This indicates the correction amount for the proportional gain; This represents the proportional term in a fuzzy PID controller; This represents the correction amount for the integral gain; This represents the integral term in a fuzzy PID controller. This represents the correction amount for the differential gain; This represents the derivative term in a fuzzy PID controller.

[0090] Real-time updates of the proportional, integral, and derivative parameters of the fuzzy PID controller; construction of an operational risk function to predict the crusher's operational risk score; operational risk function. ;in, Indicates that the crusher is in Operational risk score at any given moment; Indicates the first Types of faults; Index representing the fault type; Indicates the first Confidence level of the prediction for each type of failure; Indicates the first Risk weighting coefficients for different fault types; This represents the topological propagation intensity of vibration energy within the equipment structure. This represents a time-dependent fusion coefficient used to adjust the proportion of the predicted operational risk score and the real-time dynamic propagation intensity in the risk assessment.

[0091] If the fusion coefficient in the operating risk function is fixed too large, the system will rely too much on the topological propagation intensity of vibration energy, and may not be sensitive even if the fault prediction report shows a high risk; if the fusion coefficient is set too small, it will not respond to changes in the vibration spectrum, and may miss the early warning opportunity of local vibration anomalies, which may easily lead to misjudgment or missed judgment.

[0092] To address the above issues, the fusion coefficient is adjusted using a fusion coefficient adjustment formula. Make dynamic adjustments. ;in, This represents the information entropy of the current fault prediction. This represents the preset maximum information entropy;

[0093] It should be noted that information entropy is a core concept in Shannon's information theory, used to measure the uncertainty of an information source; a high information entropy indicates strong prediction uncertainty and suggests a lack of confidence at present. Approaching 1, the compensation judgment will rely more on the vibration topology; if the information entropy is low, it indicates a high confidence level in the operational risk function, so the fusion coefficient decreases, and the prediction of the operational risk function itself is trusted; when the operational risk function is completely uncertain... hour, It will depend entirely on the vibration topology; when the running risk function is 0, Then, the trust operation risk function is used for prediction;

[0094] Compared to existing technologies, the beneficial effects are as follows: Information entropy from Shannon's information theory is introduced to measure the uncertainty of the prediction results of the risk function in real time; by measuring the uncertainty of the function prediction through information entropy, the reliance on topological sensing data is increased when the prediction confidence is low, enhancing the system's fault tolerance under abnormal conditions; in scenarios with large fluctuations in prediction data or abnormal acquisition signals, the fusion mechanism can adaptively switch to a more reliable information source (such as vibration topology sensed by sensors), thereby improving the overall robustness of risk assessment; and it effectively realizes fusion reasoning based on risk function prediction and vibration energy topology, enabling the system to establish a highly reliable adaptive feedback path between the perception layer and the cognition layer.

[0095] Integrating multiple dynamic data sources (fault prediction, topology graph, PID status, etc.) results in complex decision feedback paths, leading to information fusion conflicts and the risk of "response drift." To address these issues, a confidence threshold function is introduced to limit the confidence of fault types participating in the fuzzy PID controller's regulation; the confidence threshold function is... ;in, This indicates the preset minimum confidence level; Indicates the first If the confidence level of a certain fault type prediction is greater than or equal to the preset minimum confidence level, that fault type participates in the fuzzy PID controller regulation. Indicates the first If the confidence level of a certain fault type prediction is less than the preset minimum confidence level, that fault type will not participate in the fuzzy PID controller regulation.

[0096] Compared to existing technologies, the advantages are as follows: the confidence threshold mechanism significantly improves the stability and response accuracy of the regulation logic by eliminating low-reliability prediction sources; it associates participation in control with information quality, realizing a confidence-driven regulation path pruning mechanism to avoid invalid or erroneous information from disturbing the controller and ensuring the robustness of the control loop; it is suitable for high-complexity operating scenarios where multiple potential faults coexist, preventing weak faults or false detection information from misleading the main control strategy, thus enhancing system security and engineering practicality.

[0097] Methods for dynamically adjusting the priority of control commands include:

[0098] By drawing an analogy between control intensity and the dynamic variables of biological metabolic output, a control intensity formula is constructed to obtain the basic control intensity; the control intensity formula is: ;in, Indicates the basic control strength; Indicates the structural disturbance response index; This indicates the pressure index for controlling task priority; Indicates the weighting coefficients for the operational risk response dimension; These represent the weighting coefficients for the structural disturbance adjustment dimension; This represents the weighting coefficients for the task priority adjustment dimension; based on expert experience. , , The value range is between 0 and 1, and ;

[0099] It simulates nonlinear activation or inhibition effects in metabolic pathways and uses a nonlinear mapping function to adjust the baseline control strength. The nonlinear mapping function is as follows: ;in, This indicates the control strength after nonlinear adjustment; This indicates the preset basal metabolic rate threshold; This represents the adjustment sensitivity coefficient, controlling the steepness of the nonlinear activation curve. The larger the value, the more sensitive the adjustment. It should be noted that this nonlinear mapping function form originates from the Sigmoid curve modeling approach for enzyme activity regulation in metabolic kinetics, and is also commonly used in neural activation functions, drug-effect-dose relationship modeling, and other fields. When the control intensity is low, the system is in an inhibitory state with a slow response. When the control intensity is close to the threshold, a threshold activation effect occurs, resulting in a rapid response. When the control intensity is far above the threshold, the response tends to saturate, preventing the system from becoming overly aggressive or oscillating.

[0100] At the same time, the stability index of the historical execution of control instructions is introduced as a weighting factor to dynamically adjust and correct the priority of control instructions. ; Indicates the first The priority of each control command; Indicates control commands The execution stability index (reflected by the magnitude of instruction disturbances during historical execution). The weights represent the impact of stability, based on expert experience. The value ranges from 0 to 1.

[0101] The following problems in existing technologies have been solved: In existing methods, the priority of control commands is often set based on static indicators, which lacks comprehensive modeling of multiple factors such as operational risks, disturbance response and task pressure. The control strength adjustment is fixed and cannot be dynamically adjusted with environmental disturbances. Traditional control logic is difficult to simulate nonlinear activation / inhibition mechanisms similar to biological systems, resulting in insufficient response or inappropriate inhibition in high-risk or high-pressure task scenarios. Existing technologies usually weight multiple indicators with a fixed ratio and fail to introduce an adaptive weighting mechanism based on changes in task background and operating status.

[0102] Compared with existing technologies, the advantages are: integrating multi-source dynamic indicators such as operational risk, structural disturbance response and control task priority pressure to construct a basic control strength model, and achieving differentiated regulation through weight factors, so that the strength and execution order of control commands no longer depend on static settings, but can automatically change with the evolution of system state, thereby improving the flexibility and adaptability of control logic.

[0103] A nonlinear regulation function under the metabolic pathway activation mechanism was constructed, which can achieve rapid increase or gradual inhibition of control strength near the critical state point, thereby more realistically simulating the "kinase regulation" mechanism in complex systems, enhancing the emergency response capability of the control system to extreme tasks, sudden disturbances or high-risk states, and avoiding hysteresis or over-tuning problems.

[0104] By introducing the historical execution stability index of control instructions as a priority weighting factor, the actual priority of instructions can be dynamically adjusted according to their historical execution performance, thereby reducing the problem of frequent instruction switching and scheduling drift caused by environmental changes, and significantly improving the continuity of system response and the controllability of instruction scheduling.

[0105] By setting a nonlinear adjustment factor and a stability weighting mechanism, this method prioritizes the control of critical tasks under conditions of limited control resources, mitigates the response of non-critical tasks, reduces the overall resource conflict rate, and improves the orderliness of control behavior.

[0106] Methods for quickly determining the current operating condition include:

[0107] A multi-source data aggregation mechanism is constructed to unify the acquisition of vibration energy topology maps, fault prediction reports, control commands, and deviation analysis reports as input data streams into the MLOps working pipeline, forming a dynamic dataset with time-series consistency and data traceability. A strategy optimization process is set up on the MLOps mainline to uniformly encode and format the input dynamic dataset to obtain the characteristics of the current operating conditions.

[0108] A meta-model selector is constructed to match control strategies from a pre-trained strategy meta-model based on the characteristics of the current operating conditions, generating candidate control strategies. The strategy meta-model is pre-trained from historical operating conditions and simulation scenarios, covering a variety of typical fault situations and control strategy modes. A real-time simulator is introduced to verify the candidate control strategies, select the optimal control strategy, and execute the corresponding control commands. The control commands are then sent to the corresponding crusher execution components in real time through the crusher execution terminal.

[0109] A remote intelligent control device for crusher operation includes a crusher, wherein the crusher is equipped with a remote intelligent control system for crusher operation that integrates a dynamic sensing field module, a cognitive twin evolution module, an adaptive robust execution module, a multi-scale verification module, and a dynamic meta-optimization module.

[0110] The preset basal metabolic intensity threshold is set by staff. By collecting data on different metabolic intensities, the average value of multiple metabolic intensities is taken as the preset basal metabolic intensity threshold. Similarly, the preset maximum information entropy and preset minimum confidence level are set.

[0111] This embodiment introduces information entropy from Shannon's information theory to measure the uncertainty of the prediction results of the risk function in real time. By measuring the uncertainty of the function prediction through information entropy, the system increases its reliance on topological sensing data when the prediction confidence is low, thereby enhancing its fault tolerance under abnormal conditions. In scenarios where the prediction data fluctuates greatly or the acquired signals are abnormal, the fusion mechanism can adaptively switch to a more reliable information source, thereby improving the overall robustness of risk assessment. It effectively realizes the fusion reasoning based on risk function prediction and vibration energy topology, enabling the system to establish a highly reliable adaptive feedback path between the perception layer and the cognition layer.

[0112] The confidence threshold mechanism significantly improves the stability and response accuracy of the control logic by eliminating sources of low reliability predictions. It links participation in control with information quality, realizing a confidence-driven control path pruning mechanism to avoid invalid or erroneous information from disturbing the controller and ensuring the robustness of the control loop. It is suitable for high-complexity operating scenarios where multiple potential faults coexist, preventing weak faults or false detection information from misleading the master control strategy and enhancing system security and engineering practicality.

[0113] By integrating multi-source dynamic indicators such as operational risk, structural disturbance response, and control task priority pressure, a basic control strength model is constructed. Differential regulation is achieved through weighting factors, enabling the strength and execution order of control commands to automatically change with the evolution of the system state, thereby improving the flexibility and adaptability of the control logic. A nonlinear adjustment function under the metabolic pathway activation mechanism is constructed, which can achieve rapid increase or gradual suppression of control strength near critical state points, enhancing the emergency response capability of the control system to extreme tasks, sudden disturbances, or high-risk states, and avoiding hysteresis or overshoot problems.

[0114] By introducing a historical execution stability index of control instructions as a priority weighting factor, the actual priority of instructions can be dynamically adjusted according to their historical execution performance. This reduces the problem of frequent instruction switching and scheduling drift caused by environmental changes, and significantly improves the continuity of system response and the controllability of instruction scheduling. By setting a nonlinear adjustment factor and a stability weighting mechanism, this method prioritizes the control of critical tasks under the condition of limited control resources, mitigates the response of non-critical tasks, reduces the overall resource conflict rate, and improves the orderliness of control behavior.

[0115] Example 2

[0116] Please see Figure 2As shown, parts not described in detail in this embodiment are described in Embodiment 1. A remote intelligent control method for crusher operation is provided, including:

[0117] S1. A distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array are integrated and deployed to form a dense monitoring network at key locations in the crushing chamber; equipment operation data is collected, and the field reconstruction algorithm based on graph neural network is used to reconstruct the field of the equipment operation data to generate a vibration energy topology map;

[0118] S2. Construct an industrial knowledge graph based on the Neo4j graph database, integrate the equipment historical fault database and equipment operation data, and perform time-series pattern mining using the Transformer-XL model to obtain the equipment's time-series behavior embedding vector; based on the time-series behavior embedding vector, construct an uncertainty reasoning engine using Bayesian networks and DS evidence theory to generate a fault prediction report with confidence assessment.

[0119] S3. An execution core is built based on a fuzzy PID controller. Combined with fault prediction reports and vibration energy topology maps, the system dynamically senses equipment operation risks and adaptively adjusts control parameters to generate control commands. A metabolic heuristic algorithm is also introduced to adjust the control intensity and dynamically adjust the priority of control commands.

[0120] S4. A hardware-in-the-loop test platform is built using a real-time simulator, integrating FPGA programmable fault injection tools to generate a deviation analysis report for virtual debugging and control command execution.

[0121] S5. Construct a strategy optimization mainline based on the MLOps architecture, integrate vibration energy topology diagram, fault prediction report, control command and deviation analysis report, use meta-model selector to quickly determine the current operating condition and execute the corresponding control command.

[0122] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0123] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A remote intelligent control system for crusher operation, characterized in that, include: The dynamic sensing field module adopts a fusion deployment of distributed fiber optic vibration sensors and 3D-printed flexible strain sensor arrays to form a dense monitoring network at key locations in the crushing chamber. Collect equipment operation data, and use a field reconstruction algorithm based on graph neural network to reconstruct the field of the equipment operation data to generate a vibration energy topology map; Key locations within the crushing chamber include the inner wall of the crushing chamber, the liner support structure, the main bearing connection area, and the edges of the inlet and outlet; equipment operating data includes dynamic mechanical status data, electrical energy consumption data, and thermal lubrication status data. The cognitive twin evolution module constructs an industrial knowledge graph based on the Neo4j graph database, integrates the equipment historical fault database and equipment operation data, and performs temporal pattern mining using the Transformer-XL model to obtain the temporal behavior embedding vector of the equipment. Based on the temporal behavior embedding vector, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory to generate a fault prediction report with confidence assessment. The adaptive robust execution module is based on a fuzzy PID controller to build the execution core. It combines fault prediction reports and vibration energy topology maps to dynamically perceive equipment operation risks and adaptively adjust control parameters to generate control commands. Furthermore, a metabolic-inspired algorithm is introduced to adjust the control strength and dynamically adjust the priority of control commands; The multi-scale verification module uses a real-time simulator to build a hardware-in-the-loop test platform, integrates FPGA programmable fault injection tools, and generates a deviation analysis report on virtual debugging and control command execution. The dynamic meta-optimization module constructs a strategy optimization mainline based on the MLOps architecture, integrates vibration energy topology diagrams, fault prediction reports, control commands, and deviation analysis reports, and uses a meta-model selector to quickly determine the current operating conditions and execute the corresponding control commands.

2. The remote intelligent control system for crusher operation according to claim 1, characterized in that, The method for forming the dense monitoring network includes: Sensor channels and integrated interfaces are pre-set at key locations in the crushing chamber to form a physical deployment platform for sensors; fiber optic lines are laid around the periphery of the crushing chamber and the main structure to form a continuous and seamless vibration sensing link; a composite material with conductive and elastic properties is used as a flexible substrate, and strain sensors are directly printed on the flexible substrate in the pre-selected installation area using 3D printing technology to form a flexible strain sensor array that adaptively conforms to the curved surface. A micro edge computing module is deployed between a distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array to perform edge preprocessing and multimodal fusion of distributed signals, thereby forming a dense monitoring network for multimodal fusion within the fracture cavity.

3. The remote intelligent control system for crusher operation according to claim 2, characterized in that, The method for generating the vibration energy topology map includes: Real-time equipment operation data is collected by a distributed fiber optic vibration sensor and a 3D-printed flexible strain sensor array deployed in the crushing chamber; each location in the crushing chamber where a sensor is deployed is defined as a vibration energy topology graph node, and edges between all vibration energy topology graph nodes are constructed based on physical adjacency relationships; The device operation data collected by the sensors corresponding to each vibration energy topology graph node are preprocessed to obtain the node feature matrix; the vibration energy topology graph structure is constructed based on the obtained vibration energy topology graph node, edge and node feature matrix; the vibration energy topology graph structure is used as the input of the graph neural network, and the node feature fusion representation of each vibration energy topology graph node is obtained by spatiotemporal reconstruction of the node features of the vibration energy topology graph nodes; Energy estimation is performed on the node feature fusion representation of each vibration energy topology graph node to obtain the vibration energy distribution value of each vibration energy topology graph node; the vibration energy distribution value of each vibration energy topology graph node is mapped to the vibration energy topology graph structure to form the vibration energy topology graph.

4. The remote intelligent control system for crusher operation according to claim 3, characterized in that, The method for constructing the industrial knowledge graph includes: From equipment operation data, semantically meaningful knowledge elements are extracted, including entity elements, attribute elements, and relation elements. Based on the extracted knowledge elements, a data model for an industrial knowledge graph is designed, which includes nodes, edges, and attributes. Neo4j is used as the graph database platform to store the industrial knowledge graph structure. Its native graph engine is used for graph traversal, path analysis and pattern matching. A retrieval mechanism for key entities is established through attribute indexing and full-text indexing. Key entities include fault type, operating condition label, equipment component name and status event. A business expert rule base is introduced to define the relation weights and node credibility indicators in the industrial knowledge graph. Combined with rule-based logical rules, the reasoning capabilities of the graph are gradually expanded. Based on Neo4j Bloom, the visualization of the industrial knowledge graph structure is supported to obtain the final industrial knowledge graph.

5. The remote intelligent control system for crusher operation according to claim 4, characterized in that, The method for obtaining the temporal behavior embedding vector of the device includes: The equipment historical fault database includes fault type, fault occurrence time, fault duration, fault location data, and fault repair records; a Transformer-XL model is built and trained based on the equipment historical fault database and equipment operation data. The Transformer-XL model includes an embedding layer, a multi-head attention layer, a feedforward neural network layer, and an output layer. The model's embedding layer is used to transform the device's historical fault database and historical device operation data into vector form, and the model's output layer is used to output the device's temporal behavior embedding vector. The mean absolute error is used as the loss function to measure the error of the model's prediction. The trained Transformer-XL model is used to output the device's temporal behavior embedding vector.

6. The remote intelligent control system for crusher operation according to claim 5, characterized in that, The method for obtaining the fault prediction report with confidence assessment includes: Based on temporal behavior embedding vectors, an uncertainty reasoning engine is constructed using Bayesian networks and DS evidence theory; the uncertainty reasoning engine includes a fusion mechanism based on Bayesian network structure and DS evidence theory. The temporal behavior embedding vector is mapped to nodes in a Bayesian network. The nodes of the Bayesian network are designed based on the equipment operation data. Each node represents a potential fault type. The dependency structure between the nodes of the Bayesian network is learned using the equipment's historical fault database. The conditional probability table of each node of the Bayesian network is determined by maximum likelihood estimation. The conditional probability table is mapped to the basic probability assignment function in the DS evidence theory. The Dempster composition rule is applied to combine the basic probability assignment functions to obtain the comprehensive confidence level for each fault type. Based on the comprehensive confidence level, a fault prediction report with confidence assessment is output.

7. The remote intelligent control system for crusher operation according to claim 6, characterized in that, The method for generating the control commands includes: The fault prediction report includes the predicted probability and overall confidence level for different fault types, and is received by the crusher through a fuzzy PID controller. Operational deviation at any time and rate of change of deviation ; The fuzzy PID controller uses a fuzzy inference mechanism to... and As input, based on a preset fuzzy rule base, the output is the PID parameter correction amount. , and ; This indicates the correction amount for the proportional gain; This represents the proportional term in a fuzzy PID controller; This represents the correction amount for the integral gain; This represents the integral term in a fuzzy PID controller. This represents the correction amount for the differential gain; This represents the derivative term in a fuzzy PID controller. Real-time updates of the proportional, integral, and derivative parameters of the fuzzy PID controller; construction of an operational risk function to predict the crusher's operational risk score; operational risk function. ;in, Indicates the crusher is in Operational risk score at any given moment; Indicates the first Types of faults; Index representing the fault type; Indicates the first Confidence level of the prediction for each type of failure; Indicates the first Risk weighting coefficients for different fault types; This represents the topological propagation intensity of vibration energy within the equipment structure. This represents the time-dependent fusion coefficient; The fusion coefficient is adjusted using the fusion coefficient adjustment formula. Make dynamic adjustments. ;in, This represents the information entropy of the current fault prediction. This represents the preset maximum information entropy; A confidence threshold function is introduced to limit the confidence of fault types participating in the regulation of the fuzzy PID controller; the confidence threshold function is: ;in, This indicates the preset minimum confidence level; Indicates the first If the confidence level of a certain fault type prediction is greater than or equal to the preset minimum confidence level, that fault type participates in the fuzzy PID controller regulation. Indicates the first If the confidence level of a certain fault type is less than the preset minimum confidence level, that fault type will not participate in the fuzzy PID controller regulation.

8. The remote intelligent control system for crusher operation according to claim 7, characterized in that, The method for dynamically adjusting the priority of control commands includes: By analogy with the dynamic variables of biological metabolic output, a control strength formula is constructed to obtain the basic control strength. The nonlinear activation or inhibition effects in the metabolic pathway are simulated, and a nonlinear mapping function is used to adjust the basic control strength. At the same time, the stability index of the historical execution of control commands is introduced as a weighting factor to dynamically adjust and correct the priority of control commands.

9. The remote intelligent control system for crusher operation according to claim 8, characterized in that, The method for quickly determining the current operating condition includes: A multi-source data aggregation mechanism is constructed to uniformly aggregate the acquired vibration energy topology map, fault prediction report, control command and deviation analysis report as input data streams into the MLOps working pipeline to form a dynamic dataset; a strategy optimization process is set up on the MLOps mainline to uniformly encode and format the input dynamic dataset to obtain the characteristics of the current operating condition; A meta-model selector is constructed to match control strategies from pre-trained strategy meta-models based on the characteristics of the current operating conditions, generating candidate control strategies. A real-time simulator is introduced to simulate and verify the candidate control strategies, select the optimal control strategy, and execute the corresponding control commands. The control commands are then sent to the corresponding crusher execution components in real time through the crusher execution terminal.

10. A remote intelligent control device for crusher operation, employing a remote intelligent control system for crusher operation as described in any one of claims 1-9, characterized in that: The invention includes a crusher, which is equipped with a remote intelligent control system for crusher operation that integrates a dynamic sensing field module, a cognitive twin evolution module, an adaptive robust execution module, a multi-scale verification module, and a dynamic meta-optimization module.

Images