Dynamic risk assessment system based on multi-field coupling modeling

Abstract

The invention relates to the technical field of Internet of Things, in particular to a dynamic risk assessment system based on multi-field coupling modeling, which comprises a multi-source data acquisition module, a data fusion processing module, a digital twin construction module and a feature extraction module, an operation mode division module, a risk propagation analysis module, a health assessment module, a life prediction module and a decision optimization module. According to the method, the accurate maintenance decision is finally generated by collecting the multi-source monitoring data, constructing the digital twinborn body for multi-physical field simulation, dynamically identifying the operation mode, analyzing the risk propagation path, evaluating the health state and predicting the residual life, so that the accurate evaluation and prediction of the equipment state are realized, and the operation reliability and the maintenance efficiency of the equipment are improved.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and in particular to a dynamic risk assessment system based on multi-field coupling modeling. Background Technology

[0002] In the current field of power plant equipment condition monitoring, discretely deployed monitoring devices such as vibration sensors, infrared thermal imagers, and oil condition sensors, along with process parameters from DCS / SIS systems, create a multi-source, heterogeneous data environment. Due to the lack of a unified data fusion mechanism, data barriers exist between vibration spectra, temperature field distributions, oil quality parameters, and process parameters such as temperature, pressure, and flow rate, leading to fragmented equipment condition information. Simultaneously, traditional monitoring methods rely on single-physics field analysis, making it difficult to capture the coupling effects of mechanical, thermal, and electrical multi-physics fields, resulting in biased equipment health status assessments. While existing technologies employ digital twin technology for equipment modeling, the simulation models are disconnected from real-time operating data, failing to dynamically reflect actual operating condition evolution. This leads to delayed condition warnings and insufficient accuracy in remaining life prediction, making it difficult to meet the requirements of the "Condition-Based Maintenance Guidelines" for accurate equipment health status assessment.

[0003] Chinese patent CN120725435A discloses an intelligent risk early warning method and system based on multi-source data fusion, belonging to the field of construction risk assessment and early warning technology. The method involves: acquiring multi-source information; obtaining early warning control quantities based on the multi-source information and a preset quantitative processing method; constructing an initial early warning cloud map based on the early warning control quantities and preset graded early warning standards; obtaining a three-dimensional early warning cloud map based on the initial early warning cloud map, a two-dimensional matrix distribution, and a preset two-dimensional normal cloud model; optimizing and correcting the three-dimensional early warning cloud map using the posterior probability support vector method to obtain an improved prediction model; and obtaining risk early warning results based on an improved evidence fusion method and the improved prediction model. This invention provides an intelligent risk early warning method and system based on multi-source data fusion, which realizes the construction of a corrected three-dimensional early warning model based on fused multi-source data combined with historical data, using the posterior probability support vector method and an improved evidence fusion method, significantly improving the accuracy, effectiveness, and applicability of risk early warning. However, this solution still suffers from problems in existing equipment health management systems, such as inaccurate equipment status assessment, delayed fault early warning, and maintenance decisions being out of touch with actual conditions due to isolated data, static models, and independent risk assessment. Summary of the Invention

[0004] To address this, the present invention provides a dynamic risk assessment system based on multi-field coupling modeling, which overcomes the problems of inaccurate equipment status assessment, delayed fault warning, and disconnect between maintenance decisions and actual conditions caused by data isolation, static models, and independent risk assessment in existing equipment health management systems.

[0005] To achieve the above objectives, the present invention provides a dynamic risk assessment system based on multi-field coupling modeling, comprising: The multi-source data acquisition module is used to collect vibration spectrum data, temperature field distribution data, oil quality parameters, temperature monitoring data, pressure monitoring data and flow monitoring data to obtain multi-source heterogeneous data; The data fusion processing module is used to fuse and normalize the multi-source heterogeneous data to obtain multi-source fused data; The digital twin construction module is used to construct dynamic digital twins using digital twin construction methods. The feature extraction module is used to input the multi-source fused data into the dynamic digital twin to obtain the operating mode identifier and physical feature vector output by the dynamic digital twin; The operation mode segmentation module is used to dynamically segment the operation modes of the device according to the operation mode identifier and physical feature vector, and obtain the mode segmentation result; The risk propagation analysis module is used to analyze the coupling relationship and propagation path of multiple risk factors in the device based on the modality division results, and to obtain the risk transmission path; The health assessment module is used to assess the health status of the equipment based on the operating mode and feature vector and the risk transmission path, and to obtain the health level assessment result. The lifespan prediction module is used to predict the remaining effective lifespan of the device based on the health level assessment results, and to obtain a remaining lifespan estimate. The decision optimization module is used to generate maintenance decision schemes and resource allocation instructions based on the health level assessment results and the remaining lifespan estimate.

[0006] Furthermore, the multi-source data acquisition module collects vibration spectrum data by deploying vibration sensors, acquires temperature field distribution data by using an infrared thermal imager, acquires oil quality parameters by using an oil condition sensor, and integrates temperature monitoring data, pressure monitoring data, and flow monitoring data from the DCS system and SIS system through an industrial protocol gateway, which together constitute the multi-source heterogeneous data.

[0007] Furthermore, the data fusion processing module performs fusion and normalization processing on the multi-source heterogeneous data to obtain multi-source fused data, specifically including: Data cleaning is performed on multi-source heterogeneous data to remove missing values, outliers, and inconsistencies, resulting in cleaned multi-source heterogeneous data. The cleaned multi-source heterogeneous data is then standardized to obtain standardized multi-source heterogeneous data. Data integration technology is used to merge the standardized multi-source heterogeneous data to obtain merged multi-source heterogeneous data. Finally, the merged multi-source heterogeneous data is normalized to obtain multi-source fused data.

[0008] Furthermore, the digital twin construction method includes: Step S1: Model the geometric and physical properties of the equipment to obtain a parametric 3D model and a material property library; Step S2: Perform multiphysics coupling analysis based on the parametric three-dimensional model and the material property library to obtain a multiphysics coupling finite element model; Step S3: Perform co-simulation integration of the multiphysics coupled finite element model and the system behavior model to obtain the initial digital twin; Step S4: Use real-time acquired multi-source fusion data to calibrate and drive the initial digital twin to obtain a dynamic digital twin.

[0009] Furthermore, the feature extraction module inputs the multi-source fusion data into the dynamic digital twin to obtain the operating mode identifier and physical feature vector output by the dynamic digital twin, including: Step A1: Input the process parameters in the multi-source fusion data as boundary conditions into the dynamic digital twin to drive it to perform transient simulation calculations and obtain high-dimensional physical field data. Step A2: Perform time-frequency domain joint analysis on the high-dimensional physical field data and real-time sensor data to extract key physical features and form a physical feature vector; Step A3: Based on the physical feature vector, identify the operating status mode of the device through a pre-set fuzzy C-means clustering model to obtain the operating mode identifier.

[0010] Furthermore, the operation mode segmentation module dynamically segments the device's operation modes based on the operation mode identifier and physical feature vector, obtaining the mode segmentation results including: Construct a dynamic feature matrix containing the operating mode identifier and physical feature vector; perform fuzzy clustering analysis on the dynamic feature matrix to calculate the membership degree of each sample point to each cluster center; determine the operating mode boundary based on the membership degree matrix, and generate a mode partitioning result containing the operating mode identifier, mode boundary time, and baseline feature vector.

[0011] Furthermore, the risk propagation analysis module analyzes the coupling relationship and propagation path of multiple risk factors in the device based on the modality segmentation results, and obtains the risk transmission path including: Based on the operating mode identifier in the modality partitioning results, the set of dominant risk factors under the current operating condition is determined, a risk node topology is established, and the coupling strength between each risk node in the set of dominant risk factors is calculated through dynamic causal association analysis. A risk coupling matrix is ​​constructed, and based on the risk coupling matrix, a fault tree and Bayesian network fusion analysis method is used to quantify the risk propagation probability and generate a risk transmission path.

[0012] Furthermore, the health assessment module assesses the equipment's health status based on the operating mode and feature vector and the risk transmission path, obtaining a health level assessment result including: Based on the physical feature vector and operating mode identifier in the operating mode and feature vector, and combined with the risk propagation probability in the risk transmission path, the comprehensive health index of the equipment is calculated; the comprehensive health index and real-time equipment operating data are input into the health status classification model to obtain a preliminary health level; based on the key risk node information in the risk transmission path, the preliminary health level is corrected to obtain the final health level assessment result.

[0013] Furthermore, the lifespan prediction module predicts the remaining effective lifespan of the device based on the health level assessment results, and the remaining lifespan estimate includes: Based on the health index value and status classification identifier in the health level assessment results, a device degradation trajectory sequence is constructed; a Wiener process is used to establish a random degradation model to identify parameters of the degradation trajectory sequence; based on the random degradation model, the probability distribution of remaining lifespan is calculated using the first arrival time concept; combined with the risk warning information in the health level assessment results, the probability distribution of remaining lifespan is corrected to obtain a remaining lifespan estimate.

[0014] Furthermore, the decision optimization module generates maintenance decision schemes and resource allocation instructions based on the health level assessment results and the remaining lifespan estimate, including: Based on the status classification identifier in the health level assessment results and the expected value of remaining life in the remaining life estimation, a maintenance priority assessment matrix is ​​established; according to the maintenance priority assessment matrix and resource constraints, a mixed integer programming optimization model is constructed; the branch and bound algorithm is used to solve the mixed integer programming model to generate a preliminary maintenance decision scheme; based on equipment correlation and risk transmission path, the preliminary maintenance decision scheme is collaboratively optimized to generate a maintenance decision scheme and resource allocation instructions.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: the system collects multi-source monitoring data, constructs a digital twin for multi-physics simulation, dynamically identifies operating modes, analyzes risk propagation paths, assesses health status and predicts remaining lifespan, and ultimately generates precise maintenance decisions, achieving accurate assessment and prediction of equipment status, improving equipment operational reliability and maintenance efficiency. The system comprehensively acquires multi-dimensional monitoring data such as vibration, temperature, and oil levels through a multi-source data acquisition module, providing a complete data foundation for health assessment. Furthermore, the system uses a data fusion processing module to fuse and normalize the multi-source heterogeneous data to obtain multi-source fused data. Finally, the system establishes a mechanical-thermal-electrical multi-physics simulation through a digital twin construction module. The system employs a coupled model to achieve high-fidelity simulation of equipment operating status. It also dynamically identifies equipment operating status through a feature extraction module and an operating mode segmentation module, accurately distinguishing between stable operation and transitional conditions. Furthermore, the system quantifies the coupling relationships and propagation paths of multiple risk factors through a risk propagation analysis module, providing early warnings of cascading failure risks. Finally, the system comprehensively assesses equipment health status and predicts remaining service life through a health assessment module and a lifespan prediction module, providing a scientific basis for preventative maintenance. Finally, the system generates precise maintenance plans and resource allocation instructions through a decision optimization module, achieving closed-loop management throughout the entire process from condition monitoring to maintenance execution. This significantly improves overall equipment reliability, extends equipment lifespan, and optimizes maintenance resource allocation. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the dynamic risk assessment system based on multi-field coupling modeling in this embodiment. Detailed Implementation

[0017] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0018] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0019] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0020] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] Please see Figure 1 As shown, this is a schematic diagram of the dynamic risk assessment system based on multi-field coupling modeling in this embodiment. The system includes: The multi-source data acquisition module is used to collect vibration spectrum data, temperature field distribution data, oil quality parameters, temperature monitoring data, pressure monitoring data and flow monitoring data to obtain multi-source heterogeneous data; A data fusion processing module is used to fuse and normalize the multi-source heterogeneous data to obtain multi-source fused data. The data fusion processing module is connected to the multi-source data acquisition module. A digital twin construction module is used to construct a dynamic digital twin using a digital twin construction method. The digital twin construction module is connected to the data fusion processing module. The feature extraction module is used to input the multi-source fused data into the dynamic digital twin to obtain the operating mode identifier and physical feature vector output by the dynamic digital twin. The feature extraction module is connected to the digital twin construction module. A mode segmentation module is used to dynamically segment the operating modes of the device based on the operating mode identifier and physical feature vector, and obtain the mode segmentation result. The operating mode segmentation module is connected to the feature extraction module. The risk propagation analysis module is used to analyze the coupling relationship and propagation path of multiple risk factors in the device based on the modality division results, and to obtain the risk transmission path. The risk propagation analysis module is connected to the operation modality division module. A health assessment module is used to assess the health status of the equipment based on the operating mode and feature vector and the risk transmission path, and to obtain a health level assessment result. The health assessment module is connected to the risk propagation analysis module. The lifespan prediction module is used to predict the remaining effective lifespan of the device based on the health level assessment results, and to obtain a remaining lifespan estimate. The lifespan prediction module is connected to the health assessment module. The decision optimization module is used to generate maintenance decision schemes and resource allocation instructions based on the health level assessment results and the remaining life prediction. The decision optimization module is connected to the life prediction module.

[0022] Specifically, the dynamic risk assessment system based on multi-field coupling modeling is applied to the health management platform of key power plant equipment. The system collects multi-source monitoring data, constructs digital twins for multi-physics field simulation, dynamically identifies operating modes, analyzes risk propagation paths, assesses health status, and predicts remaining lifespan, ultimately generating precise maintenance decisions. This achieves accurate assessment and prediction of equipment status, improving equipment operational reliability and maintenance efficiency. The system comprehensively acquires multi-dimensional monitoring data such as vibration, temperature, and oil levels through a multi-source data acquisition module, providing a complete data foundation for health assessment. The system also utilizes data fusion processing... The system employs a processing module to fuse and normalize the multi-source heterogeneous data, resulting in multi-source fused data. It also establishes a mechanical-thermal-electrical multi-physics coupling model through a digital twin construction module, enabling high-fidelity simulation of equipment operating conditions. Furthermore, the system dynamically identifies equipment operating states through a feature extraction module and an operating mode segmentation module, accurately distinguishing between stable operation and transitional conditions. A risk propagation analysis module quantifies the coupling relationships and propagation paths of various risk factors, providing early warnings of cascading failure risks. Additionally, a health assessment module and a lifespan prediction module comprehensively evaluate equipment health status and predict remaining service life, providing a scientific basis for preventative maintenance. Finally, a decision optimization module generates precise maintenance plans and resource allocation instructions, achieving closed-loop management from condition monitoring to maintenance execution, significantly improving overall equipment reliability, extending equipment lifespan, and optimizing maintenance resource allocation.

[0023] Specifically, the multi-source data acquisition module collects vibration spectrum data by deploying vibration sensors, acquires temperature field distribution data by using an infrared thermal imager, acquires oil quality parameters by using an oil condition sensor, and integrates temperature monitoring data, pressure monitoring data, and flow monitoring data from the DCS system and SIS system through an industrial protocol gateway, which together constitute the multi-source heterogeneous data.

[0024] Specifically, the multi-source data acquisition module acquires vibration spectrum data by deploying vibration sensors, which means using vibration sensors installed at key parts of the equipment to obtain a set of characteristic frequency components reflecting the mechanical operating state of the equipment; acquires temperature field distribution data by using an infrared thermal imager, which means using infrared thermal imaging equipment to perform non-contact scanning on the equipment surface to obtain two-dimensional temperature gradient information expressed in the form of a heat map; acquires oil quality parameters by using an oil condition sensor, which means using dedicated sensors to monitor in real time the physicochemical indicators characterizing the performance of lubricating or insulating oils, such as dielectric constant, moisture content, and particulate matter concentration; and integrates temperature monitoring data, pressure monitoring data, and flow monitoring data from the DCS system through an industrial protocol gateway, which means supporting multiple industrial communication protocols (such as Modbus, OPC) The UA conversion device uniformly collects the core process parameters (temperature, pressure, flow) used for process monitoring in the distributed control system (DCS); at the same time, it integrates temperature monitoring data, pressure monitoring data, and flow monitoring data from the SIS system, which means synchronously collecting similar parameters from independent measurement channels in the safety instrumented system (SIS) used for safety interlock protection through the same gateway; the multi-source heterogeneous data jointly formed refers to the final set of original data with differences in structure and semantics, including time-domain waveforms, image matrices, chemical parameters, and various industrial real-time monitoring parameters.

[0025] Specifically, the data fusion processing module performs fusion and normalization processing on the multi-source heterogeneous data to obtain multi-source fused data, including: Data cleaning is performed on multi-source heterogeneous data to remove missing values, outliers, and inconsistencies, resulting in cleaned multi-source heterogeneous data. The cleaned multi-source heterogeneous data is then standardized to obtain standardized multi-source heterogeneous data. Data integration technology is used to merge the standardized multi-source heterogeneous data to obtain merged multi-source heterogeneous data. Finally, the merged multi-source heterogeneous data is normalized to obtain multi-source fused data.

[0026] Specifically, the data cleaning of multi-source heterogeneous data refers to improving the quality of the original data by filling in missing values, correcting outliers, and unifying data units and naming conventions to obtain cleaned multi-source heterogeneous data. The standardization of the cleaned multi-source heterogeneous data involves using methods such as Z-score to eliminate dimensional differences between different feature data, converting them into a distribution with a mean of 0 and a standard deviation of 1, resulting in standardized multi-source heterogeneous data. Merging standardized multi-source heterogeneous data using data integration technology involves identifying the same entity through entity resolution and resolving data conflicts according to rules, integrating multi-source information into a unified view to obtain merged multi-source heterogeneous data. Normalization of the merged multi-source heterogeneous data involves using methods such as min-max normalization to linearly scale the data to the [0,1] interval, ultimately generating multi-source fusion data with a uniform scale that can be directly used for digital twin modeling and analysis.

[0027] Specifically, the digital twin construction method includes: Step S1: Model the geometric and physical properties of the equipment to obtain a parametric 3D model and a material property library; Step S2: Perform multiphysics coupling analysis based on the parametric 3D model and the material property library to obtain a multiphysics coupling finite element model; Step S3: Perform co-simulation integration of the multiphysics coupled finite element model and the system behavior model to obtain an initial digital twin; Step S4: Use real-time acquired multi-source fusion data to calibrate and drive the initial digital twin to obtain a dynamic digital twin.

[0028] Specifically, the modeling of the equipment's geometric and physical properties to obtain a parametric 3D model and material property library involves: constructing a parametric 3D model containing dimensions, tolerances, and assembly relationships based on the equipment's CAD drawings; simultaneously defining the material properties of each component constituting the equipment, including elastic modulus, density, specific heat capacity, thermal conductivity, and electromagnetic properties, forming a material property library; and performing multiphysics coupling analysis based on the parametric 3D model and material property library to obtain a multiphysics coupling finite element model involves: importing the output of step S1 into finite element analysis software, applying mechanical boundary conditions, thermal boundary conditions, and electrical excitation under actual working conditions, and calculating the distribution data of the stress field, temperature field, and electromagnetic field of the equipment during operation by solving the coupled partial differential equations, thereby forming a multiphysics coupling finite element model that reflects the mechanical-thermal-electrical coupling effect; and then performing co-simulation integration of the multiphysics coupling finite element model with the system behavior model to obtain an initial... The initial digital twin is specifically established as follows: A system behavior model representing the control logic and dynamic response of the equipment is established in the system simulation environment; through a co-simulation interface, the multi-physics coupled finite element model obtained in step S2 is connected to the system behavior model, so that the physical field state (e.g., temperature) can be fed back to the control system, and the control command (e.g., rotational speed) can drive the physical field simulation, forming a bidirectional data-interactive initial digital twin. The process of calibrating and driving the initial digital twin using real-time acquired multi-source fusion data to obtain a dynamic digital twin is specifically as follows: Multi-source fusion data (e.g., actual vibration and temperature data) from the data fusion processing module is input in real-time into the initial digital twin obtained in step S3; a parameter identification and state update algorithm is used to dynamically correct key parameters in the model (e.g., friction coefficient and thermal resistance), ensuring that the output of the digital twin is consistent with the real state of the physical entity, ultimately forming a dynamic digital twin that evolves synchronously with the physical entity and can be used for prediction.

[0029] Specifically, the feature extraction module inputs the multi-source fusion data into the dynamic digital twin to obtain the device features output by the dynamic digital twin, including: Step A1: Input the process parameters in the multi-source fusion data as boundary conditions into the dynamic digital twin to drive it to perform transient simulation calculations and obtain high-dimensional physical field data. Step A2: Perform time-frequency domain joint analysis on the high-dimensional physical field data and real-time sensor data to extract key physical features and form a physical feature vector; Step A3: Based on the physical feature vector, identify the operating status mode of the device through a pre-set fuzzy C-means clustering model to obtain the operating mode identifier.

[0030] Specifically, the process parameters in the multi-source fusion data are used as boundary conditions to input the dynamic digital twin, driving it to perform transient simulation calculations to obtain high-dimensional physical field data. Specifically, temperature, pressure, and flow process parameters are extracted from the multi-source fusion data and applied as time-varying boundary conditions to the multi-physics coupled finite element model of the dynamic digital twin. Transient simulation calculations are performed, and high-dimensional physical field data containing stress field, temperature field, and electromagnetic field distributions are output. The high-dimensional physical field data and real-time sensor data are subjected to joint time-frequency domain analysis to extract key physical features and form a physical feature vector. Specifically, the high-dimensional physical field data obtained in step A1 is synchronized with the vibration spectrum and infrared thermal imaging real-time sensor data in the multi-source fusion data; the wavelet packet transform algorithm is applied to extract time-frequency domain energy features, and key physical features such as stress concentration factor, temperature gradient, and electromagnetic asymmetry are calculated to jointly constitute the physical feature vector. Based on the physical feature vector, the operating state mode of the device is identified by the modal clustering algorithm to obtain the operating mode identifier. Specifically, the physical feature vector obtained in step A2 is input into a pre-set trained fuzzy C-means clustering model. The model dynamically divides the device operating state into stable operating mode, start-stop transition mode, and abnormal disturbance mode according to the similarity of the feature vectors, and outputs the corresponding operating mode identifier.

[0031] Specifically, the operation mode segmentation module dynamically segments the operation modes of the device based on the operation mode identifier and physical feature vector, and the resulting mode segmentation results include: Construct a dynamic feature matrix containing the operating mode identifier and physical feature vector; perform fuzzy clustering analysis on the dynamic feature matrix to calculate the membership degree of each sample point to each cluster center; determine the operating mode boundary based on the membership degree matrix, and generate a mode partitioning result containing the operating mode identifier, mode boundary time, and baseline feature vector.

[0032] Specifically, constructing a dynamic feature matrix containing the operating mode identifier and the physical feature vector involves: performing one-hot encoding on the operating mode identifier to form a mode encoding vector M; concatenating the mode encoding vector M with the physical feature vector P to construct a dynamic feature matrix X of dimension N×(K+L), where N is the number of samples, K is the physical feature dimension, L is the mode encoding dimension, and each row of the matrix represents a complete feature representation at a sampling time. The dynamic feature matrix is ​​subjected to fuzzy clustering analysis. The membership degree of each sample point to each cluster center is calculated as follows: The dynamic feature matrix X is clustered using the fuzzy C-means algorithm. The objective function J = ΣΣu{ij}^m ||xi - cj||^2 is iteratively optimized, where u{ij} represents the membership degree of the i-th sample to the j-th cluster center, m is the fuzzy factor, and cj is the cluster center. Finally, the sample membership matrix U = [u{ij}]{N×C} is obtained, where C is the preset number of modal categories. The modal partitioning result, which includes the modal identifier, modal boundary time, and reference feature vector, is determined based on the membership matrix. Specifically, the maximum value of each row in the membership matrix U is calculated. When max(u{ij}) > θhigh, the sample is assigned to the corresponding modality. When max(u{ij}) < θlow, it is marked as a transition state. The start time ts and end time te of each modality are recorded as the modal boundary time, and the centroid of each modal sample is used as the reference feature vector. Finally, the modal partitioning result, which includes the modal identifier, the modal boundary time set {ts, te}, and the reference feature vector set, is output.

[0033] Specifically, the risk propagation analysis module analyzes the coupling relationships and propagation paths of multiple risk factors in the device based on the modality segmentation results, and obtains the risk transmission paths including: Based on the operating mode identifier in the modality partitioning results, the set of dominant risk factors under the current operating condition is determined, a risk node topology is established, and the coupling strength between each risk node in the set of dominant risk factors is calculated through dynamic causal association analysis. A risk coupling matrix is ​​constructed, and based on the risk coupling matrix, a fault tree and Bayesian network fusion analysis method is used to quantify the risk propagation probability and generate a risk transmission path.

[0034] Specifically, based on the operating mode identifier in the modality division result, the set of dominant risk factors under the current working condition is determined, and the risk node topology is established as follows: according to the operating mode identifier, a preset risk mode library is matched to determine the dominant mechanical wear risk node Rm, insulation aging risk node Re, and scaling risk node Rf under this mode, and the connection topology between risk nodes is established based on the physical structure of the equipment. Through dynamic causal correlation analysis, the coupling strength between each risk node in the set of dominant risk factors is calculated, and the risk coupling matrix is ​​constructed by extracting the feature cluster center coordinates in the modality division results, combining them with historical fault data, and using the Granger causality test method to calculate the interaction strength between risk nodes, forming a risk coupling matrix Cij that includes mechanical-electrical-thermal cross-physical field coupling relationships. Based on the aforementioned risk coupling matrix, a fault tree and Bayesian network fusion analysis method is used to quantify the risk propagation probability and generate the risk transmission path. Specifically, the risk coupling matrix Cij is used as the conditional probability distribution input of the Bayesian network, and the risk propagation probability is calculated using the following formula: P(Rj|Ri) = Cij×exp(-λ×dij); Where Ri represents the source risk node, Rj represents the target risk node, dij represents the physical distance between risk nodes, and λ is the attenuation coefficient. The final output is a key risk transmission path map containing propagation probability weights. The source risk node refers to the fault risk unit that is the starting point or upstream node in the risk transmission chain, that is, the risk factor that first causes an anomaly or initially triggers a fault. The target risk node refers to the downstream node affected by the source risk node in the risk transmission chain, that is, the risk factor that may subsequently cause anomalies or faults due to the propagation of upstream risks. The physical distance between risk nodes refers to the actual interval length between the source risk node and the target risk node in the physical location in the equipment spatial layout or functional topology, or the equivalent logical distance calculated according to the functional connection relationship. The attenuation coefficient is a model parameter used in the risk propagation probability calculation model to quantify the exponential attenuation rate of the risk impact intensity as the physical / logical distance increases. Its value needs to be calibrated according to the specific equipment type and historical fault data statistical analysis.

[0035] Specifically, the health assessment module assesses the health status of the equipment based on the operating mode and feature vector and the risk transmission path, and obtains the health level assessment results, including: Based on the physical feature vector and operating mode identifier in the operating mode and feature vector, and combined with the risk propagation probability in the risk transmission path, the comprehensive health index of the equipment is calculated; the comprehensive health index and real-time equipment operating data are input into the health status classification model to obtain a preliminary health level; based on the key risk node information in the risk transmission path, the preliminary health level is corrected to obtain the final health level assessment result.

[0036] Specifically, based on the physical feature vector and operating mode identifier in the operating mode and feature vector, and combined with the risk propagation probability in the risk transmission path, the comprehensive health index of the equipment is calculated as follows: Key feature parameters such as effective vibration value, temperature gradient, and oil contamination degree are extracted from the physical feature vector; baseline values ​​for each parameter are determined according to the operating mode identifier; and the risk propagation probability is introduced as a weighting correction factor. The comprehensive health index (HI) is then calculated using a health assessment function. HI = 1 - Σ(wn × |xn - x{n0}| / x{n0}) × Prisk; where wn is the weight coefficient of the nth feature parameter, xn is the current value of the parameter, x{n0} is the baseline value of the parameter under the corresponding operating mode, and Prisk is the comprehensive risk probability from the risk transmission path; The comprehensive health index and real-time equipment operation data are input into the health status classification model to obtain the preliminary health level. Specifically, the comprehensive health index calculated in step S1 and the real-time data of stress concentration and temperature field uniformity from the digital twin are input into the health status classification model based on support vector machine, and the preliminary health level that conforms to the "condition maintenance guidelines" is output, including four levels: normal, attention, abnormal and serious. Based on the key risk node information in the risk transmission path, the preliminary health level is corrected to obtain the final health level assessment result. Specifically, the key risk nodes in the risk transmission path that are in an active state and their propagation path lengths are extracted. When there is a key risk with a cross-physical field propagation path length less than the threshold Lmin, the preliminary health level is raised by one severity level. Finally, a health level assessment result containing health level identification and risk warning information is output.

[0037] Specifically, the lifespan prediction module predicts the remaining effective lifespan of the device based on the health level assessment results, and the remaining lifespan estimate includes: Based on the health index value and status classification identifier in the health level assessment results, a device degradation trajectory sequence is constructed; a Wiener process is used to establish a random degradation model to identify parameters of the degradation trajectory sequence; based on the random degradation model, the probability distribution of remaining lifespan is calculated using the first arrival time concept; combined with the risk warning information in the health level assessment results, the probability distribution of remaining lifespan is corrected to obtain a remaining lifespan estimate.

[0038] Specifically, based on the health index value and status classification identifier in the health level assessment results, the construction of the equipment degradation trajectory sequence involves: extracting the historical health index value HI(t) and its corresponding timestamp from the health level assessment results; combining this with the status classification identifier; marking the change in health status from "normal" to "attention" as the degradation start point t0; and constructing the degradation trajectory sequence {HI(t0), HI(t1), ..., HI(t2)} from t0 to the current time. n )}; A stochastic degradation model is established using the Wiener process. Parameter identification of the degradation trajectory sequence is specifically performed as follows: A degradation model is established based on the Wiener process: X(t) = X0 + λt + σB(t), where X(t) represents the degradation amount at time t, X0 represents the initial degradation state, λ is the drift coefficient, σ is the diffusion coefficient, and B(t) represents standard Brownian motion. The degradation trajectory sequence is then fitted using the maximum likelihood estimation method to obtain parameter estimates (λ^, σ^). Based on the aforementioned stochastic degradation model, the probability distribution of remaining lifetime is calculated using the first-arrival time concept as follows: A failure threshold D is set. Based on the first-arrival time concept, the remaining lifetime T follows an inverse Gaussian distribution: T ~ IG(μ,λ), where μ = (D - X0) / λ^, λ = (D - X0)² / σ^². The probability density function f(t) and cumulative distribution function F(t) of the remaining lifetime are calculated using this distribution. Combining the risk warning information in the health level assessment results, the remaining life probability distribution is corrected to obtain the remaining life prediction. Specifically, when there is a high-risk warning in the health level assessment results, the failure threshold D is adjusted based on the propagation speed of the risk transmission path, and D' = κD is used for correction, where κ is the risk correction coefficient (0 < κ < 1). Finally, the remaining life prediction containing the expected value of remaining life E[T] and the confidence interval is obtained.

[0039] Specifically, the decision optimization module generates maintenance decision schemes and resource allocation instructions based on the health level assessment results and the remaining lifespan estimate, including: Based on the status classification identifier in the health level assessment results and the expected value of remaining life in the remaining life estimation, a maintenance priority assessment matrix is ​​established; according to the maintenance priority assessment matrix and resource constraints, a mixed integer programming optimization model is constructed; the branch and bound algorithm is used to solve the mixed integer programming model to generate a preliminary maintenance decision scheme; based on equipment correlation and risk transmission path, the preliminary maintenance decision scheme is collaboratively optimized to generate a maintenance decision scheme and resource allocation instructions.

[0040] Specifically, based on the status classification identifier in the health level assessment results and the expected value of remaining life in the remaining life estimation, the maintenance priority assessment matrix is ​​established as follows: the equipment health status is assigned a score of 1-4 according to four levels: "normal, attention, abnormal, and serious". The remaining life is divided into four intervals according to length and assigned a score of 1-4. The maintenance priority score P = w1×S + w2×L is constructed, where S is the health status score, L is the remaining life score, and w1 and w2 are weighting coefficients, forming an assessment matrix that includes the maintenance priority of each piece of equipment. Based on the maintenance priority evaluation matrix and resource constraints, the mixed-integer programming optimization model is constructed as follows: An optimization model with the objective of minimizing total cost is established: min Σ(c 1i ×x i + c 2i ×y i + c 3i ×z i ), where c 1i For the maintenance cost of the i-th device, c 2i For spare parts procurement costs, c 3i To cover downtime losses, x i y i z i The decision variables are 0-1, and the constraints include resource limitations such as total budget B, manpower hours T, and spare parts inventory S. The branch and bound algorithm is used to solve the mixed integer programming model to generate a preliminary maintenance decision scheme. The branch and bound algorithm is used to solve the established optimization model to obtain the optimal solution set that satisfies the resource constraints, including the list of equipment that needs to be maintained, the maintenance time window arrangement, and the types and quantities of spare parts required, thus forming a preliminary maintenance decision scheme. Based on equipment correlation and risk transmission path, the preliminary maintenance decision scheme is collaboratively optimized to generate maintenance decision scheme and resource allocation instructions. Specifically, the preliminary maintenance decision scheme considers the functional correlation between equipment and the coupling relationship in the risk transmission path. When multiple related equipment need maintenance at the same time, they are merged into the same maintenance work order to obtain the maintenance decision scheme. At the same time, a resource allocation instruction containing a spare parts procurement list, personnel scheduling plan and maintenance time arrangement is generated based on the maintenance decision scheme, and this instruction is fed back to the monitoring system to adjust the monitoring strategy of the relevant equipment.

[0041] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A dynamic risk assessment system based on multi-field coupling modeling, characterized in that, The method comprises the following steps: A multi-source data acquisition module is used to acquire vibration spectrum data, temperature field distribution data, oil quality parameters, temperature monitoring data, pressure monitoring data and flow monitoring data, thereby obtaining multi-source heterogeneous data; A data fusion processing module is used to fuse and normalize the multi-source heterogeneous data, thereby obtaining multi-source fusion data; A digital twin construction module is used to construct a dynamic digital twin through a digital twin construction method; A feature extraction module is used to input the multi-source fusion data into the dynamic digital twin, thereby obtaining a running mode identifier and a physical feature vector output by the dynamic digital twin; A running mode division module is used to dynamically divide the running mode of the equipment according to the running mode identifier and the physical feature vector, thereby obtaining a mode division result; A risk propagation analysis module is used to analyze the coupling relationship and propagation path of multiple risk factors in the equipment based on the mode division result, thereby obtaining a risk transmission path; A health assessment module is used to assess the health status of the equipment based on the running mode and the feature vector and the risk transmission path, thereby obtaining a health grade assessment result; A life prediction module is used to predict the remaining effective life of the equipment based on the health grade assessment result, thereby obtaining a remaining life prediction; A decision optimization module is used to generate a maintenance decision scheme and a resource allocation instruction based on the health grade assessment result and the remaining life prediction.

2. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The multi-source data acquisition module acquires vibration spectrum data by deploying a vibration sensor, acquires temperature field distribution data by deploying an infrared thermal imager, acquires oil quality parameters by deploying an oil state sensor, and acquires temperature monitoring data, pressure monitoring data and flow monitoring data from DCS systems and SIS systems through an industrial protocol gateway, thereby forming the multi-source heterogeneous data.

3. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The data fusion processing module fuses and normalizes the multi-source heterogeneous data, thereby obtaining multi-source fusion data, which specifically comprises the following steps: Data cleaning is performed on the multi-source heterogeneous data to process missing values, abnormal values and inconsistencies in the multi-source heterogeneous data, thereby obtaining cleaned multi-source heterogeneous data; the cleaned multi-source heterogeneous data is standardized, thereby obtaining standardized multi-source heterogeneous data; the standardized multi-source heterogeneous data is integrated through data integration technology, thereby obtaining integrated multi-source heterogeneous data; and the integrated multi-source heterogeneous data is normalized, thereby obtaining multi-source fusion data.

4. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The digital twin construction method comprises the following steps: Step S1: modeling the geometric and physical properties of the equipment to obtain a parameterized three-dimensional model and a material attribute library; Step S2: performing multi-physical field coupling analysis based on the parameterized three-dimensional model and the material attribute library to obtain a multi-physical field coupling finite element model; Step S3: performing collaborative simulation integration of the multi-physical field coupling finite element model and a system behavior model to obtain an initial digital twin; Step S4: calibrating and driving the initial digital twin using real-time acquired multi-source fusion data to obtain a dynamic digital twin.

5. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The feature extraction module inputs the multi-source fusion data into the dynamic digital twin, thereby obtaining a running mode identifier and a physical feature vector output by the dynamic digital twin, which comprises the following steps: Step A1, input the process parameters in the multi-source fusion data as boundary conditions into the dynamic digital twin, drive it to perform transient simulation calculation, and obtain high-dimensional physical field data; Step A2, perform time-frequency domain joint analysis on the high-dimensional physical field data and real-time sensor data, extract key physical features, and form a physical feature vector; Step A3, based on the physical feature vector, identify the operating state mode of the equipment through a pre-set fuzzy C-means clustering model, and obtain an operating mode identifier.

6. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The operating mode division module divides the operating mode of the equipment according to the operating mode identifier and the physical feature vector, and obtains a mode division result including: A dynamic feature matrix containing the operating mode identifier and the physical feature vector is constructed; fuzzy clustering analysis is performed on the dynamic feature matrix to calculate the membership degree of each sample point to each cluster center; the operating mode boundary is determined based on the membership matrix, and a mode division result containing the operating mode identifier, the mode boundary time, and the reference feature vector is generated.

7. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The risk propagation analysis module analyzes the coupling relationship and propagation path of multiple risk factors in the equipment based on the mode division result, and obtains a risk transmission path including: Based on the operating mode identifier in the mode division result, the dominant risk factor set under the current working condition is determined, the risk node topology is established, the coupling strength between each risk node in the dominant risk factor set is calculated through dynamic causal correlation analysis, the risk coupling matrix is constructed, and based on the risk coupling matrix, the fault tree and Bayesian network fusion analysis method is used to quantify the risk propagation probability, and the risk transmission path is generated.

8. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The health assessment module assesses the health status of the equipment based on the operating mode and feature vector and the risk transmission path, and obtains a health grade evaluation result including: Based on the physical feature vector and the operating mode identifier in the operating mode and feature vector, and combining the risk propagation probability in the risk transmission path, a comprehensive health index of the equipment is calculated; the comprehensive health index and real-time equipment operation data are input into a health state classification model to obtain a preliminary health grade; based on the key risk node information in the risk transmission path, the preliminary health grade is corrected to obtain the final health grade evaluation result.

9. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The life prediction module predicts the remaining effective life of the equipment based on the health grade evaluation result, and obtains a remaining life estimate including: Based on the health index value and state classification identifier in the health grade evaluation result, a device degradation trajectory sequence is constructed; a random degradation model is established using a Wiener process to perform parameter identification on the degradation trajectory sequence; based on the random degradation model, the probability distribution of the remaining life is calculated through the concept of first passage time; combined with the risk warning information in the health grade evaluation result, the probability distribution of the remaining life is corrected to obtain the remaining life estimate.

10. The dynamic risk assessment system based on multi-field coupling modeling of claim 1, wherein, The decision optimization module generates a maintenance decision scheme and resource allocation instruction based on the health grade evaluation result and the remaining life estimate including: Based on the state classification identifier in the health level evaluation result and the residual life expectancy value in the residual life prediction, a maintenance priority evaluation matrix is established; a mixed integer programming optimization model is constructed according to the maintenance priority evaluation matrix and a resource constraint condition; a branch and bound algorithm is used to solve the mixed integer programming model, and a preliminary maintenance decision scheme is generated; based on equipment correlation and a risk transmission path, the preliminary maintenance decision scheme is optimized in coordination, and a maintenance decision scheme and a resource allocation instruction are generated.

Images