A real-time operation monitoring method and system for super-large leaching stirring tank

Abstract

The application relates to the technical field of industrial equipment monitoring, and discloses a real-time operation monitoring method and system for a super-large leaching stirring tank. The system comprises an operation environment construction module, a real-time monitoring execution module, an intelligent evaluation feedback module, a case matching engine module, a self-adaptive optimization module and a comprehensive report generation module. Mechanical vibration spectrum, dynamic torque fluctuation, ore pulp flow state distribution and electrochemical parameter dynamic response data are collected through a multi-source sensing array, and original operation state data sets are integrated and constructed to cover multi-dimensional operation states of the equipment. Based on wavelet packet energy entropy, torque mutation density, turbulence intensity and leaching reaction activity characteristics, operation state characteristic vector sets are generated to ensure the comprehensiveness of monitoring data and the accuracy of feature representation and reduce the risk of monitoring blind areas.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial equipment monitoring, in particular to a real-time operation monitoring method and system for a super-large leaching agitator. BACKGROUND

[0002] The agitator is a special equipment for leaching and cleaning liquid in the metallurgical industry, and its structure design is similar to that of the flotation machine. According to the stirring mode, it can be divided into two types: air agitator and mechanical agitator. The agitator is suitable for leaching of fine grinding high-grade ore and needs to be combined with high-temperature and high-acid conditions, but it is mostly used in special scenes such as sulfide ore due to the high operation and maintenance cost.

[0003] At present, in the operation monitoring process of the super-large leaching agitator, due to the complex structure of the equipment, the harsh operation environment and the variable working conditions, the traditional method mainly relies on single and limited types of sensors for data acquisition, which leads to incomplete acquisition of multi-dimensional operation state information such as mechanical vibration state, ore pulp flow state distribution, dynamic torque fluctuation and leaching reaction electrochemical parameters, and cannot real-time construct a complete data set reflecting the overall operation state of the equipment.

[0004] Therefore, the present application provides a real-time operation monitoring method and system for a super-large leaching agitator to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a real-time operation monitoring method and system for a super-large leaching agitator to solve the problem of incomplete acquisition of multi-dimensional operation state information and inability to real-time construct a complete data set reflecting the overall operation state of the equipment.

[0006] To achieve the above purpose, the present application provides the following technical solution: a real-time operation monitoring method for a super-large leaching agitator, the method comprising the following steps:

[0007] S1, collecting the operation state data of the leaching agitator through a multi-source sensor array to generate an original operation state data set;

[0008] S2, performing multi-modal feature extraction processing on the original operation state data set to generate an operation state feature vector set;

[0009] S3, performing agitator operation abnormal risk quantification evaluation processing based on the operation state feature vector set to generate real-time risk level evaluation data;

[0010] S4, performing agitator operation state dynamic classification processing according to the real-time risk level evaluation data to generate operation state classification identification data;

[0011] S5, when the running state hierarchical identification data is an abnormal state, triggering stirring tank running fault tracing analysis processing, generating multi-dimensional fault tracing positioning data;

[0012] S6, based on the multi-dimensional fault tracing positioning data and historical fault case library, fault disposal strategy matching processing is performed to generate target fault disposal strategy feature data;

[0013] S7, constructing real-time monitoring summary data to perform stirring tank running state visualization reconstruction processing, generating three-dimensional dynamic running monitoring graph;

[0014] S8, according to the three-dimensional dynamic running monitoring graph and the preset running safety threshold, stirring tank running safety early warning decision processing is performed to generate hierarchical safety early warning instruction set;

[0015] S9, executing the device control operation corresponding to the hierarchical safety early warning instruction set through the industrial control bus.

[0016] Preferably, the running state data of the leaching stirring tank in S1 includes the following steps:

[0017] S11, collecting tank mechanical vibration frequency spectrum data through a piezoelectric vibration sensor array distributedly arranged on the leaching stirring tank body;

[0018] S12, collecting main shaft dynamic torque fluctuation data through a torque-speed composite sensor embeddedly arranged on the leaching stirring tank transmission shaft;

[0019] S13, collecting ore pulp solid-liquid two-phase flow state distribution data through a multi-depth tomographic imaging probe of ore pulp medium;

[0020] S14, collecting ore pulp electrochemical parameter dynamic response data through a tank body ring electrode array;

[0021] S15, integrating the mechanical vibration frequency spectrum data, dynamic torque fluctuation data, flow state distribution data and electrochemical parameter dynamic response data to construct an original running state data set.

[0022] Preferably, the multi-modal feature extraction processing in S2 includes the following steps:

[0023] S21, wavelet packet energy entropy feature extraction processing is performed on the mechanical vibration frequency spectrum data to generate a vibration energy entropy feature vector;

[0024] S22, torque mutation point statistical feature extraction processing is performed on the dynamic torque fluctuation data to generate a torque mutation density feature vector;

[0025] S23, ore pulp flow state vortex scale analysis processing is performed on the flow state distribution data to generate a flow state turbulent intensity feature vector;

[0026] S24, reaction activity index calculation processing is performed on the electrochemical parameter dynamic response data to generate leaching reaction activity feature vectors;

[0027] S25, the vibration energy entropy feature vectors, the torque mutation density feature vectors, the turbulent intensity feature vectors and the leaching reaction activity feature vectors are fused to construct a set of operating state feature vectors.

[0028] Preferably, the stirring tank operating abnormality risk quantitative evaluation processing in S3 comprises the following steps:

[0029] S31, an abnormality risk quantitative evaluation model based on a deep belief network is constructed, and an input layer of the model receives the set of operating state feature vectors;

[0030] S32, a multi-feature nonlinear coupling relationship analysis processing is performed through a hidden layer to generate state feature coupling response data;

[0031] S33, a risk probability mapping processing is performed in an output layer to generate real-time risk level evaluation data, and the risk level includes a safe level, a warning level and a dangerous level.

[0032] Preferably, the stirring tank operating state dynamic classification processing in S4 comprises the following steps:

[0033] S41, the real-time risk level evaluation data is received, and a three-dimensional classification coordinate system including a mechanical state dimension, a fluid state dimension and a reaction state dimension is established;

[0034] S42, in the mechanical state dimension, three mechanical operating states are divided based on the vibration energy entropy feature vectors: a stable level, a fluctuation level and an instability level;

[0035] S43, in the fluid state dimension, three ore slurry flow states are divided according to the flow turbulence intensity feature vectors: a laminar flow level, a transition level and a turbulent flow level;

[0036] S44, in the reaction state dimension, three reaction states are divided according to the leaching reaction activity feature vectors: a normal activity level, a fluctuation activity level and an abnormal activity level;

[0037] S45, a dynamic classification decision strategy is established, and corresponding identification is generated when the three-dimensional classification results meet the following conditions:

[0038] Safe identification: mechanical stable level + laminar flow level + normal activity level;

[0039] Warning identification: any dimension reaches the fluctuation level or the transition level;

[0040] Dangerous identification: any dimension reaches the instability level, the turbulent flow level or the abnormal activity level;

[0041] S46, mapping the running state grading identification data to the corresponding area of the three-dimensional dynamic running monitoring graph in real time through a grading result feedback mechanism.

[0042] Preferably, the stirring tank running fault tracing analysis processing in S5 comprises the following steps:

[0043] S51, when the running state grading identification data is an abnormal state, activating a fault tracing analysis engine, locating the physical space coordinates where the abnormal state occurs, and determining the initial time node where the abnormal state occurs;

[0044] S52, identifying a dominant fault influence factor set through multi-parameter coupling analysis;

[0045] S53, integrating the physical space coordinates, the initial time node, and the dominant fault influence factor set to construct multi-dimensional fault tracing positioning data.

[0046] Preferably, the fault disposal strategy matching processing in S6 comprises the following steps:

[0047] S61, establishing a historical fault case knowledge graph, the nodes of which include fault type, occurrence location, disposal measure, and recovery effect;

[0048] S62, performing graph structure similarity matching of the multi-dimensional fault tracing positioning data and the historical fault case knowledge graph;

[0049] S63, when the matching similarity is greater than or equal to 0.85, directly outputting the corresponding fault disposal strategy as the target fault disposal strategy feature data;

[0050] S64, when the matching similarity is less than 0.85, starting a disposal strategy optimization engine based on reinforcement learning to generate a new disposal strategy.

[0051] Preferably, the stirring tank running state visual reconstruction processing in S7 comprises the following steps:

[0052] S71, constructing a physical field dynamic model inside the tank by fusing real-time mechanical vibration distribution data, ore pulp flow state motion data, and temperature field gradient data;

[0053] S72, mapping risk level data to the model surface using color coding technology;

[0054] S73, representing the ore pulp flow trajectory through dynamic particle flow visualization technology;

[0055] S74, integrating to generate a three-dimensional dynamic running monitoring graph containing physical field state, risk distribution, and flow trajectory.

[0056] Preferably, the stirring tank running safety early warning decision processing in S8 comprises the following steps:

[0057] S81, set double threshold early warning trigger mechanism:

[0058] Primary threshold triggers yellow early warning instruction: equipment speed reduction;

[0059] High threshold triggers red early warning instruction: emergency shutdown;

[0060] S82, trigger red early warning instruction when the coverage rate of the dangerous area in the three-dimensional dynamic operation monitoring graph is greater than or equal to 15%;

[0061] S83, trigger yellow early warning instruction when the leaching reaction activity characteristic value deviates from the reference value by ± 30% and lasts for 120 seconds;

[0062] S84, generate a hierarchical safety early warning instruction set containing early warning levels, affected areas and disposal suggestions.

[0063] A real-time operation monitoring method and system for a super-large leaching stirring tank, the system comprising:

[0064] An operation environment construction module generates a stirring tank operation virtual environment using a three-dimensional scene reconstruction unit, sets safety specifications through an operation rule configuration unit, and outputs environment data through an abnormal working condition simulation unit;

[0065] A real-time monitoring execution module receives the environment data, collects device operation states through a multi-modal sensing unit, performs state deduction using a physical process simulation unit, and outputs monitoring data through a fault propagation calculation unit;

[0066] An intelligent evaluation feedback module receives the monitoring data, constructs a dynamic evaluation matrix through a data acquisition and processing unit, outputs evaluation results through a risk evaluation generation unit, and provides safety early warning feedback through a real-time early warning unit;

[0067] A case matching engine module receives the evaluation results, generates disposal instructions by calling a case library through a feature extraction unit and a similarity matching unit, and feeds back to the operation environment construction module through a decision deduction unit to adjust monitoring parameters;

[0068] An adaptive optimization module receives the evaluation results output by the intelligent evaluation feedback module, analyzes operation characteristics through a device state portrait construction unit, identifies system vulnerable items through a weak link identification unit, and generates optimization instructions through a monitoring strategy control unit;

[0069] A comprehensive report generation module integrates the evaluation results and operation data, and generates a visual monitoring report through a multi-index fusion unit and a weight calculation unit.

[0070] Beneficial effects: compared with the prior art, the present application provides a real-time operation monitoring method and system for super-large leaching stirring tank, which has the following beneficial effects:

[0071] 1. In the present application, the mechanical vibration spectrum, dynamic torque fluctuation, ore slurry flow pattern distribution and electrochemical parameter dynamic response data are collected by a multi-source sensing array, and the original operation state data set is integrated and constructed, covering the multi-dimensional operation state of the equipment. Based on wavelet packet energy entropy, torque mutation density, turbulence intensity and leaching reaction activity characteristics extraction processing, the operation state feature vector set is generated, ensuring the comprehensiveness of the monitoring data and the accuracy of the feature representation, and reducing the risk of monitoring blind area.

[0072] 2. In the present application, the state feature coupling response data is generated through multi-feature nonlinear coupling relationship analysis, and is mapped into real-time risk level evaluation data. According to the three-dimensional grading coordinate system, the mechanical state, fluid state and reaction state dimension level are dynamically divided, the grading decision strategy is established to generate operation state grading identification data, the abnormal state is identified in real time, and the dynamic nature of risk grading and timely warning ability are ensured.

[0073] 3. In the present application, the mechanical vibration distribution, ore slurry flow movement and temperature field gradient data are fused to construct the internal physical field dynamic model of the tank, generate the three-dimensional dynamic operation monitoring atlas, execute the double-threshold early warning trigger mechanism according to the preset operation safety threshold, generate the grading safety early warning instruction set and control the equipment through the industrial control bus, intuitively display the risk distribution and implement grading control, ensure the safety of equipment operation and the visualization of operation. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 The flowchart of the real-time operation monitoring method for the super-large leaching stirring tank of the present application;

[0075] Figure 2 The framework diagram of the real-time operation monitoring system for the super-large leaching stirring tank of the present application. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0077] Specific embodiment: a real-time operation monitoring method for a super-large leaching stirring tank, the method comprising the following steps:

[0078] S1, collecting operation state data of the leaching stirring tank by a multi-source sensing array to generate an original operation state data set;

[0079] S2, performing multi-modal feature extraction processing on the original operation state data set to generate an operation state feature vector set;

[0080] S3, performing stirring tank operation abnormal risk quantitative evaluation processing based on the operation state feature vector set to generate real-time risk level evaluation data;

[0081] S4, performing stirring tank operation state dynamic classification processing according to the real-time risk level evaluation data to generate operation state classification identification data;

[0082] S5, when the operation state classification identification data is an abnormal state, triggering stirring tank operation fault source analysis processing to generate multi-dimensional fault source positioning data;

[0083] S6, performing fault disposal strategy matching processing based on the multi-dimensional fault source positioning data and the historical fault case library to generate target fault disposal strategy feature data;

[0084] S7, constructing real-time monitoring summary data to perform stirring tank operation state visual reconstruction processing to generate a three-dimensional dynamic operation monitoring graph;

[0085] S8, performing stirring tank operation safety early warning decision processing according to the three-dimensional dynamic operation monitoring graph and the preset operation safety threshold to generate a set of classified safety early warning instructions;

[0086] S9, executing device control operations corresponding to the set of classified safety early warning instructions through an industrial control bus.

[0087] In S1, the operation state data of the leaching stirring tank is collected, including the following steps:

[0088] S11, collecting tank body mechanical vibration frequency spectrum data by a piezoelectric vibration sensor array distributedly arranged on the leaching stirring tank body;

[0089] S12, collecting main shaft dynamic torque fluctuation data by a torque-speed composite sensor embeddedly arranged on the leaching stirring tank drive shaft system;

[0090] S13, collecting ore slurry solid-liquid two-phase flow state distribution data by a multi-depth tomographic imaging probe of the ore slurry medium;

[0091] S14, collecting ore slurry electrochemical parameter dynamic response data by a tank body ring electrode array;

[0092] S15, integrating the mechanical vibration frequency spectrum data, the dynamic torque fluctuation data, the flow state distribution data, and the electrochemical parameter dynamic response data to construct the original operation state data set.

[0093] The multimodal feature extraction processing in S2 includes the following steps:

[0094] S21, wavelet packet energy entropy feature extraction processing is performed on the mechanical vibration spectrum data to generate a vibration energy entropy feature vector;

[0095] S22, torque mutation point statistical feature extraction processing is performed on the dynamic torque fluctuation data to generate a torque mutation density feature vector;

[0096] S23, mine slurry flow state vortex scale analysis processing is performed on the flow state distribution data to generate a flow state turbulence intensity feature vector;

[0097] S24, reaction activity index calculation processing is performed on the electrochemical parameter dynamic response data to generate a leaching reaction activity feature vector;

[0098] Reaction activity index calculation: ; wherein, represents the reaction activity index, pH, redox potential and current density are weight coefficients, is the normalized value of the corresponding parameter; S25, fusion of vibration energy entropy feature vector, torque mutation density feature vector, turbulence intensity feature vector and leaching reaction activity feature vector, to construct a running state feature vector set.

[0099] The agitator tank running abnormal risk quantitative evaluation processing in S3 includes the following steps:

[0100] S31, an abnormal risk quantitative evaluation model based on a deep belief network is constructed, and the input layer receives the running state feature vector set;

[0101] The specific implementation steps of the abnormal risk quantitative evaluation model are:

[0102] First, initialize the model architecture, including an input layer, two hidden layers and an output layer; second, use the historical running data set for pre-training to optimize the weight parameters and ensure that the model can learn the nonlinear relationship of multiple source features; finally, in the model deployment stage, apply the Dropout technology to prevent overfitting; the output layer is designed as a three-node structure, corresponding to the probability distribution of the safe level, the warning level and the dangerous level, respectively. After the model is constructed, it is integrated into the real-time monitoring system to realize end-to-end risk quantitative evaluation.

[0103] S32, multi-feature nonlinear coupling relationship analysis processing is performed through the hidden layer to generate state feature coupling response data;

[0104] S33, risk probability mapping processing is performed in the output layer to generate real-time risk level evaluation data, and the risk level includes: safe level, warning level and dangerous level: ; ; wherein: is the probability of the kth risk level, k = 1 corresponds to the safe level, k = 2 corresponds to the warning level, and k = 3 corresponds to the danger level; is the linear output of the kth node of the output layer, is the weight vector, is the feature coupling response data of the hidden layer output, is the output layer bias, is an index variable.

[0105] The dynamic classification processing of the stirring tank running state in S4 includes the following steps:

[0106] S41, receiving real-time risk level evaluation data, and establishing a three-dimensional classification coordinate system including mechanical state dimension, fluid state dimension, and reaction state dimension;

[0107] S42, in the mechanical state dimension, dividing three mechanical running states based on the vibration energy entropy feature vector: stable level, fluctuation level, and instability level;

[0108] S43, in the fluid state dimension, dividing three ore pulp flow states according to the flow state turbulence intensity feature vector: laminar flow level, transition level, and turbulent flow level;

[0109] S44, in the reaction state dimension, dividing three reaction states according to the leaching reaction activity feature vector: normal activity level, activity fluctuation level, and abnormal activity level;

[0110] S45, establishing a dynamic classification decision strategy, and generating corresponding identification when the three-dimensional classification results meet the following conditions:

[0111] Safe identification: mechanical stable level + laminar flow level + normal activity level;

[0112] Warning identification: any dimension reaches the fluctuation level, transition level;

[0113] Dangerous identification: any dimension reaches the instability level, turbulent flow level, or abnormal activity level;

[0114] The specific implementation steps are as follows: first, establish the mechanical state dimension, fluid state dimension, and reaction state dimension; second, define the decision rule matrix: when the three-dimensional state is all safe conditions, output the safe identification; when any dimension reaches the warning condition, output the warning identification; when any dimension reaches the danger condition, output the danger identification; finally, through the classification result feedback mechanism, real-time mapping of the identification to the three-dimensional dynamic running monitoring map ensures the dynamicity and real-time of the decision matrix

[0115] S46, through the classification result feedback mechanism, real-time mapping of the running state classification identification data to the corresponding area of the three-dimensional dynamic running monitoring map.

[0116] The troubleshooting and analysis of operational faults in the S5 mixing tank includes the following steps:

[0117] S51. When the operation status classification identification data is in an abnormal state, activate the fault tracing analysis engine, locate the physical space coordinates of the abnormal state, and determine the initial time node of the abnormal state.

[0118] S52. Identify the set of dominant fault influencing factors through multi-parameter coupling analysis;

[0119] S53. Integrate physical space coordinates, initial time nodes, and the set of dominant fault influencing factors to construct multi-dimensional fault source tracing and location data.

[0120] The fault handling strategy matching process in S6 includes the following steps:

[0121] S61. Establish a knowledge graph of historical failure cases, whose nodes include failure type, location of occurrence, handling measures and recovery effect;

[0122] S62. Perform graph structure similarity matching between multi-dimensional fault tracing and localization data and historical fault case knowledge graph: ;in Indicates matching similarity. This is the source graph for the current fault, with a set of nodes. Represents the failure factor, edge set Indicates a cause-and-effect relationship. For historical case knowledge graphs; S63, when the matching similarity is ≥0.85, directly output the corresponding fault handling strategy as the feature data of the target fault handling strategy;

[0123] S64. When the matching similarity is less than 0.85, start the reinforcement learning-based disposal strategy optimization engine to generate a new disposal strategy.

[0124] The visualization and reconstruction of the stirring tank's operating status in S7 includes the following steps:

[0125] S71. By integrating real-time mechanical vibration distribution data, slurry flow motion data and temperature field gradient data, a dynamic model of the physical field inside the tank is constructed. ;in To represent three-dimensional spatial coordinates The overall physical field strength at that location These are the weighting coefficients for vibration, flow regime, and temperature. To standardize vibration and turbulence intensity, To standardize the temperature gradient;

[0126] S73. Use color coding technology to map risk level data onto the model surface;

[0127] S74, characterizing the flow trajectory of the ore pulp by dynamic particle flow visualization technology;

[0128] S75, integrating a three-dimensional dynamic operation monitoring atlas containing physical field state, risk distribution and flow trajectory.

[0129] The stirring tank operation safety early warning decision processing in S8 includes the following steps:

[0130] S81, setting a double-threshold early warning trigger mechanism:

[0131] The primary threshold triggers a yellow early warning instruction: the device runs at a reduced speed;

[0132] The high-level threshold triggers a red early warning instruction: emergency shutdown;

[0133] S82, triggering a red early warning instruction when the coverage rate of the dangerous area in the three-dimensional dynamic operation monitoring atlas is greater than or equal to 15%;

[0134] S83, triggering a yellow early warning instruction when the leaching reaction activity characteristic value deviates from the reference value by ± 30% and lasts for 120 seconds;

[0135] S84, generating a hierarchical safety early warning instruction set containing early warning levels, affected areas and disposal suggestions.

[0136] A real-time operation monitoring method and system for a super-large leaching stirring tank, the system comprising:

[0137] An operation environment construction module generates a stirring tank operation virtual environment using a three-dimensional scene reconstruction unit, sets safety specifications through an operation rule configuration unit, and outputs environment data through an abnormal working condition simulation unit;

[0138] A real-time monitoring execution module receives environment data, collects device operation states through a multi-modal sensing unit, performs state deduction using a physical process simulation unit, and outputs monitoring data through a fault propagation calculation unit;

[0139] An intelligent evaluation feedback module receives monitoring data, constructs a dynamic evaluation matrix through a data acquisition and processing unit, outputs evaluation results through a risk evaluation generation unit, and provides safety early warning feedback through a real-time early warning unit;

[0140] A case matching engine module receives evaluation results, generates disposal instructions by calling a case library through a feature extraction unit and a similarity matching unit, and feeds back to the operation environment construction module through a decision deduction unit to adjust monitoring parameters;

[0141] An adaptive optimization module receives the evaluation results output by the intelligent evaluation feedback module, analyzes the operation characteristics through the device state portrait construction unit, identifies the system vulnerable items through the weak link identification unit, and generates optimization instructions through the monitoring strategy regulation unit;

[0142] A comprehensive report generation module integrates the evaluation results and operation data to generate a visual monitoring report through a multi-index fusion unit and a weight calculation unit.

[0143] The operation steps of the method and system are as follows:

[0144] Step one: Collecting operation state data through a multi-source sensor array to construct an original data set

[0145] In this step, the system deploys multiple sensor arrays to comprehensively collect the operation state data of the leaching stirring tank; specifically, piezoelectric vibration sensors are distributedly installed on the tank body to capture real-time mechanical vibration frequency spectrum data; sensors are used to monitor the dynamic torque fluctuation data of the transmission shaft system; probes are used to obtain solid-liquid two-phase flow state distribution data; and electrochemical parameter dynamic response data is collected. These multi-source data are integrated into a comprehensive original operation state data set, covering multiple dimensions of machinery, fluid, and reaction, ensuring the comprehensiveness of data collection and laying a foundation for subsequent processing.

[0146] Step two: Performing multi-modal feature extraction processing to generate a feature vector set

[0147] This step performs high-level feature extraction on the original data set to accurately represent the device state; the system applies wavelet packet energy entropy analysis to the mechanical vibration frequency spectrum data to generate a feature vector reflecting the vibration energy distribution; performs mutation point statistical analysis on the dynamic torque fluctuation data to extract a torque mutation density feature vector; processes the flow state distribution data through ore slurry flow state vortex scale analysis to output a flow state turbulence intensity feature vector; and calculates the reaction activity index of the electrochemical parameters to generate a leaching reaction activity feature vector; finally, these feature vectors are fused into a unified operation state feature vector set, improving the representation accuracy of the features and reducing the monitoring blind area.

[0148] Step three: Quantitative evaluation of abnormal risks based on the feature vector set to generate risk level data

[0149] The system evaluates the operation abnormal risks, the model input layer receives the feature vector set, the hidden layer analyzes the nonlinear coupling relationship between multiple features, generates state feature coupling response data; the output layer maps these responses to risk probabilities, and outputs real-time risk level evaluation data; this process optimizes the weight parameters through historical data pre-training, ensuring that the model can dynamically identify potential faults and provide quantitative risk evaluation results.

[0150] Step four: generate classification identification according to risk level data

[0151] This step establishes a three-dimensional classification coordinate system and dynamically divides the operating state level. In the mechanical state dimension, it is divided into stable level, fluctuation level and instability level based on vibration energy entropy characteristics; in the fluid state dimension, it is divided into laminar level, transition level and turbulent level according to flow state turbulence intensity characteristics; in the reaction state dimension, it is divided into normal activity level, fluctuation level and abnormal level according to leaching reaction activity characteristics. The system outputs operating state classification identification data through decision strategy and maps the results to the visualization graph in real time to ensure timely and dynamic classification.

[0152] Step five: trigger fault tracing analysis when the classification identification is abnormal to generate positioning data

[0153] When an abnormal state is detected, the system activates the fault tracing engine, which locates the physical space coordinates and initial time node of the abnormality, identifies the dominant fault influencing factor through multi-parameter coupling analysis, and integrates these information into multi-dimensional fault tracing positioning data to provide accurate basis for subsequent disposal and achieve rapid positioning and causal analysis of faults.

[0154] Step six: match disposal strategies based on fault positioning data to generate target strategy feature data

[0155] The system uses historical fault case knowledge graph for strategy matching, and the knowledge graph nodes include fault type, location, measures and effects; the system matches the fault tracing positioning data with the graph structure similarity to finally generate target fault disposal strategy feature data, which ensures the reliability and adaptability of the disposal strategy.

[0156] Step seven: build summary data and perform visualization reconstruction to generate a three-dimensional dynamic graph

[0157] This step integrates real-time data to build a dynamic model of the internal physical field of the tank, applies color coding technology to map risk level data to the model surface to represent risk distribution, and uses dynamic particle flow visualization technology to display the flow trajectory of the ore pulp. Finally, a three-dimensional dynamic operation monitoring graph is generated to intuitively present the physical field state, risk area and flow path, enhancing the operator's understanding of the equipment state.

[0158] Step eight: generate instruction set based on three-dimensional graph and safety threshold for early warning decision

[0159] The system implements a double-threshold early warning mechanism, sets a primary threshold to trigger a yellow early warning instruction, causing the equipment to run at a reduced speed, and a high-level threshold to trigger a red early warning instruction, requiring emergency shutdown. The system generates a hierarchical safety warning instruction set containing warning levels, affected areas and disposal suggestions to ensure accurate and hierarchical response to warnings.

[0160] Step nine: execute the early warning instruction through the industrial control bus to regulate the operation of the equipment

[0161] The final step converts the hierarchical safety early warning instruction set into actual operation through the industrial control bus. The system automatically executes the corresponding equipment regulation instruction, realizes closed-loop control, ensures the timely execution of the early warning decision, and maintains the safety of equipment operation.

[0162] It should be noted that, in this article, relationship terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of other identical elements in the process, method, article or equipment that includes the element.

[0163] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for real-time operational monitoring of a very large leaching stirred tank, characterized by: The method comprises the following steps: S1, collecting operation state data of the leaching stirring tank through a multi-source sensing array to generate an original operation state data set; S2, performing multi-modal feature extraction processing on the original operation state data set to generate an operation state feature vector set; S3, performing stirring tank operation abnormal risk quantification evaluation processing based on the operation state feature vector set to generate real-time risk level evaluation data; S4, performing stirring tank operation state dynamic classification processing according to the real-time risk level evaluation data to generate operation state classification identification data; S5, when the operation state classification identification data is an abnormal state, triggering stirring tank operation fault traceability analysis processing to generate multi-dimensional fault traceability positioning data; S6, performing fault disposal strategy matching processing based on the multi-dimensional fault traceability positioning data and a historical fault case library to generate target fault disposal strategy feature data; S7, constructing real-time monitoring summary data to perform stirring tank operation state visual reconstruction processing to generate a three-dimensional dynamic operation monitoring graph; S8, performing stirring tank operation safety early warning decision processing according to the three-dimensional dynamic operation monitoring graph and a preset operation safety threshold to generate a classification safety early warning instruction set; S9, performing device control operations corresponding to the classification safety early warning instruction set through an industrial control bus.

2. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The operation state data of the leaching stirring tank in S1 comprises the following steps: S11, collecting tank body mechanical vibration frequency spectrum data through a piezoelectric vibration sensor array distributedly arranged on the tank body of the leaching stirring tank; S12, collecting main shaft dynamic torque fluctuation data through a torque-speed composite sensor embeddedly arranged on the transmission shaft system of the leaching stirring tank; S13, collecting ore pulp solid-liquid two-phase flow state distribution data through a multi-depth tomographic imaging probe; S14, collecting ore pulp electrochemical parameter dynamic response data through a tank body ring electrode array; S15, integrating the mechanical vibration frequency spectrum data, the dynamic torque fluctuation data, the flow state distribution data and the electrochemical parameter dynamic response data to construct the original operation state data set.

3. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The multi-modal feature extraction processing in S2 comprises the following steps: S21, performing wavelet packet energy entropy feature extraction processing on the mechanical vibration frequency spectrum data to generate a vibration energy entropy feature vector; S22, performing torque mutation point statistical feature extraction processing on the dynamic torque fluctuation data to generate a torque mutation density feature vector; S23, performing ore pulp flow state vortex scale analysis processing on the flow state distribution data to generate a flow state turbulence intensity feature vector; S24, performing reaction activity index calculation processing on the electrochemical parameter dynamic response data to generate a leaching reaction activity feature vector; S25, fusing the vibration energy entropy feature vector, the torque mutation density feature vector, the turbulence intensity feature vector and the leaching reaction activity feature vector to construct the operation state feature vector set.

4. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The stirring tank operation abnormal risk quantification evaluation processing in S3 comprises the following steps: S31, constructing an abnormal risk quantification evaluation model based on a deep belief network, wherein an input layer of the model receives the operation state feature vector set; S32, a multi-feature nonlinear coupling relationship analysis process is performed through the hidden layer to generate state feature coupling response data; S33, a risk probability mapping process is performed in the output layer to generate real-time risk level evaluation data, and the risk level includes a safety level, a warning level, and a danger level.

5. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The S4 dynamic grading process of the stirring tank operating state includes the following steps: S41, the real-time risk level evaluation data is received, and a three-dimensional grading coordinate system including a mechanical state dimension, a fluid state dimension, and a reaction state dimension is established; S42, in the mechanical state dimension, three levels of mechanical operating states are divided based on the vibration energy entropy feature vector: a stable level, a fluctuation level, and an instability level; S43, in the fluid state dimension, three levels of ore slurry flow states are divided according to the flow state turbulence intensity feature vector: a laminar flow level, a transition level, and a turbulent flow level; S44, in the reaction state dimension, three levels of reaction states are divided according to the leaching reaction activity feature vector: a normal activity level, a fluctuation activity level, and an abnormal activity level; S45, a dynamic grading decision strategy is established, and corresponding identifiers are generated when the three-dimensional grading results meet the following conditions: Safety identifier: mechanical stable level + laminar flow level + normal activity level; Warning identifier: any dimension reaches the fluctuation level or the transition level; Danger identifier: any dimension reaches the instability level, the turbulent flow level, or the abnormal activity level; S46, through a grading result feedback mechanism, the operating state grading identifier data is mapped to the corresponding area of the three-dimensional dynamic operation monitoring graph in real time.

6. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The S5 stirring tank operating fault tracing analysis process includes the following steps: S51, when the operating state grading identifier data is an abnormal state, the fault tracing analysis engine is activated, the physical space coordinates of the abnormal state are located, and the initial time node of the abnormal state is determined; S52, the dominant fault influence factor set is identified through multi-parameter coupling analysis; S53, the physical space coordinates, the initial time node, and the dominant fault influence factor set are integrated to construct multi-dimensional fault tracing positioning data.

7. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1 wherein: The S6 fault handling strategy matching process includes the following steps: S61, a historical fault case knowledge graph is established, and the nodes thereof include fault types, occurrence positions, handling measures, and recovery effects; S62, the multi-dimensional fault tracing positioning data is matched with the historical fault case knowledge graph in terms of graph structure similarity; S63, when the matching similarity is greater than or equal to 0.85, the corresponding fault handling strategy is directly output as the target fault handling strategy feature data; S64, when the matching similarity is less than 0.85, a handling strategy optimization engine based on reinforcement learning is started to generate a new handling strategy.

8. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The S7 stirring tank operating state visualization reconstruction process includes the following steps: S71, real-time mechanical vibration distribution data, ore slurry flow state motion data, and temperature field gradient data are fused to construct a physical field dynamic model inside the tank body; S72, color coding technology is used to map the risk level data to the model surface; S73, the ore slurry flow trajectory is represented through dynamic particle flow visualization technology; S74, a three-dimensional dynamic operation monitoring graph including physical field states, risk distributions, and flow trajectories is integrated and generated.

9. A method for real time operational monitoring of ultra large leaching stirred tanks as claimed in claim 1, wherein: The safety early warning decision processing of the stirring tank in S8 comprises the following steps: S81, setting a double-threshold early warning trigger mechanism: The primary threshold triggers a yellow early warning instruction: equipment running at a reduced speed; The advanced threshold triggers a red early warning instruction: emergency shutdown; S82, triggering a red early warning instruction when the coverage rate of the dangerous area in the three-dimensional dynamic running monitoring graph is greater than or equal to 15%; S83, triggering a yellow early warning instruction when the leaching reaction activity characteristic value deviates from the reference value by ±30% and lasts for 120 seconds; S84, generating a hierarchical safety early warning instruction set containing the early warning level, the affected area, and the disposal suggestion.

10. A real-time operation monitoring system for a super large leaching stirred tank for implementing the real-time operation monitoring method for a super large leaching stirred tank according to any one of claims 1 to 9, characterized in that: The system comprises: An operating environment construction module that generates a stirring tank running virtual environment using a three-dimensional scene reconstruction unit, sets safety specifications through an operating rule configuration unit, and outputs environment data through an abnormal working condition simulation unit; A real-time monitoring execution module that receives the environment data, collects equipment running states through a multi-modal sensing unit, performs state deduction using a physical process simulation unit, and outputs monitoring data through a fault propagation calculation unit; An intelligent evaluation feedback module that receives the monitoring data, constructs a dynamic evaluation matrix through a data acquisition processing unit, outputs evaluation results through a risk evaluation generation unit, and provides safety early warning feedback through a real-time early warning unit; A case matching engine module that receives the evaluation results, generates disposal instructions by calling a case library through a feature extraction unit and a similarity matching unit, and feeds back to the operating environment construction module through a decision deduction unit to adjust the monitoring parameters; An adaptive optimization module that receives the evaluation results output by the intelligent evaluation feedback module, analyzes running characteristics through a device state portrait construction unit, identifies system vulnerable items through a weak link identification unit, and generates optimization instructions through a monitoring strategy regulation unit; A comprehensive report generation module that integrates the evaluation results and operating data, generates a visual monitoring report through a multi-index fusion unit and a weight calculation unit.

Images