Wharf facility health monitoring method and system based on digital twinning and machine learning

Abstract

The invention provides a wharf facility health monitoring method and system based on digital twinning and machine learning, and relates to the field of wharf facility health monitoring, and the method comprises the steps: building a weight coefficient of a monitoring point network topological graph according to a synchronism change value of the same type of sensor data corresponding to adjacent monitoring points; determining an abnormal transmission path based on the weight coefficient; determining an abnormal state propagation rule of each abnormal transmission path in each preset abnormal state according to the historical monitoring data corresponding to each preset abnormal state; after abnormal sensor data of any monitoring point is detected, parameter predicted values of other monitoring points in the corresponding abnormal transmission path are calculated based on an abnormal state propagation rule; and if the parameter prediction values of the preset proportion number deviate from the corresponding actual monitoring values, generating early warning information. By implementing the method, false alarms caused by single-point data fluctuation summarizing in a traditional health monitoring mode can be dealt with, and the accuracy and reliability of abnormal state early warning are improved.

Description

Technical Field

[0001] This application relates to the field of health monitoring of terminal facilities, and particularly to a method and system for health monitoring of terminal facilities based on digital twin and machine learning. Background Art

[0002] With the rapid development of the port economy, the scale and complexity of terminal facilities have been continuously increasing, and their safe operation has an important impact on port logistics and trade activities. During the long-term operation process, terminal facilities are subjected to the combined action of various adverse factors such as marine environment corrosion, periodic loads, and emergencies, which are prone to structural damage and performance degradation. Therefore, it is necessary to establish a perfect health monitoring system to ensure the safe operation of terminal facilities.

[0003] In the related art, health monitoring of terminal facilities mainly uses a distributed sensor network to collect real-time data and perform threshold alarm. Specifically, in the implementation, various types of sensors such as stress-strain sensors, displacement sensors, and stability sensors are arranged at key structural parts of the terminal facilities to collect structural response data; the collected data is transmitted to the monitoring center, and abnormal detection is performed by setting fixed warning thresholds; when a certain monitoring parameter exceeds the warning threshold, the system issues an alarm signal, and professional personnel analyze and evaluate according to the current monitoring data.

[0004] However, the existing fixed threshold alarm method generally directly triggers an alarm when the data at a single monitoring point exceeds the warning threshold. Since terminal facilities are affected by various factors such as temperature changes and tidal effects during actual operation, the data fluctuation at a single monitoring point may be caused by temporary disturbances rather than real abnormal states. This single-point alarm mechanism is prone to false alarms, which not only causes unnecessary production stoppage for maintenance but also reduces the credibility of the early warning system. Summary of the Invention

[0005] This application provides a method and system for health monitoring of terminal facilities based on digital twin and machine learning to address the problem of how to improve the accuracy and reliability of abnormal state early warning.

[0006] In a first aspect, this application provides a method for health monitoring of terminal facilities based on digital twin and machine learning, which is applied to a health monitoring system of terminal facilities. The method includes: Collecting monitoring data of terminal facilities in real time according to a multi-type sensor network deployed at multiple preset monitoring points; Classifying the monitoring data within a preset time period according to parameter types to obtain multiple monitoring parameter subsets, where the monitoring parameter subsets include multiple sensor data of the same type; A weight coefficient of the monitoring point network topology map is established based on the synchronization change value of the same type of sensor data corresponding to adjacent monitoring points in each of the monitoring parameter subsets, and the monitoring point network topology map corresponds to the monitoring parameter subsets one by one; In the digital twin model corresponding to the dock facilities, a target path corresponding to a network topology path where adjacent weight coefficients are both greater than a first preset threshold is determined as an abnormal transmission path for the same type of sensor data; Based on the historical monitoring data corresponding to each preset abnormal state, an abnormal state propagation rule for each abnormal transmission path in each preset abnormal state is determined; When it is detected that the sensor data of any monitoring point exceeds the corresponding second preset threshold range, parameter prediction values of other monitoring points in the corresponding abnormal transmission path are calculated based on the abnormal state propagation rule; If it is detected that the difference between the parameter prediction values of a preset proportion and the corresponding actual monitoring values exceeds a third preset threshold, a warning message including the abnormal location, the influence range, and the development trend is generated.

[0007] Through the above embodiments, the system can identify the transmission path of the abnormal state and avoid false alarms caused by single-point data fluctuations by establishing a network topology map and weight coefficients based on the synchronization change of multi-type sensor data. The abnormal state propagation rule established in combination with historical data can predict the diffusion trend of the abnormal state. When an abnormality is detected, by comparing the difference between the predicted value and the actual monitoring value, it is possible to more accurately determine whether it is a real abnormality and give a warning message including the abnormal location, the influence range, and the development trend, improving the accuracy and reliability of the warning.

[0008] In some embodiments, before the step of collecting the monitoring data of the dock facilities in real time by the multi-type sensor network deployed at multiple preset monitoring points, it further includes: Construct a multi-dimensional feature layer in the digital twin model corresponding to the dock facilities, and the multi-dimensional feature layer includes a material property layer, a load distribution layer, an environmental corrosion layer, and a construction quality layer; Calculate the weighted superposition value of the feature data of different feature layers in each monitoring area and the corresponding weight coefficients to obtain a combined influence coefficient. The digital twin model includes multiple monitoring areas, and the weight coefficients are determined based on preset expert scores; Determine the importance level corresponding to each monitoring area according to the combined influence coefficient; Determine the target monitoring areas with the importance level higher than the preset level threshold as the preset monitoring points.

[0009] Through the above embodiments, the system can evaluate the importance levels of each monitoring area by constructing a multi-dimensional feature layer including material properties, load distribution, environmental corrosion, and construction quality in the digital twin model and calculating the combined influence coefficients of different features. Based on the importance levels, the layout of preset monitoring points is determined, ensuring the reasonable allocation of monitoring resources, enabling the monitoring system to more specifically cover key areas, and improving the monitoring efficiency and economy.

[0010] In some embodiments, after the step of collecting the monitoring data of the wharf facilities in real time according to the multi-type sensor networks deployed at a plurality of preset monitoring points, the method further includes: Constructing a multi-level monitoring data visualization layer in the digital twin model, the visualization layer includes a global monitoring layer and a local monitoring layer, the global monitoring layer displays the comprehensive monitoring indicators of each monitoring area, and the local monitoring layer displays the sensor data of the target area; When the change rate of the comprehensive monitoring indicator within a preset time period does not exceed a preset change threshold, dynamically display the change process of the comprehensive monitoring indicator on the global monitoring layer; When the change rate of the comprehensive monitoring indicator within a preset time period exceeds the preset change threshold, dynamically display the change process of the sensor data of the target area on the local monitoring layer.

[0011] Through the above embodiments, the system can display the monitoring data at different scales by constructing a multi-level visualization system including a global monitoring layer and a local monitoring layer. When the indicator change rate is within the normal range, the global information is displayed, and when it exceeds the threshold, it automatically switches to the local detail display, realizing the intelligent display of monitoring information. This hierarchical display mechanism not only ensures the grasp of the overall state but also can timely pay attention to the specific conditions of abnormal areas, improving the visualization effect and utilization efficiency of the monitoring data.

[0012] In some embodiments, the step of establishing the weight coefficient of the monitoring point network topology diagram according to the synchronous change value of the same type of sensor data corresponding to adjacent monitoring points in each monitoring parameter subset specifically includes: Calculating the data change rate of each adjacent monitoring point within a preset time window to obtain a change rate trend, where the data change rate is the change amplitude of the sensor data per unit time; Calculating an initial weight coefficient according to the similarity of the change rate trends, where the initial weight coefficient represents the basic synchronization degree of the sensor data of adjacent monitoring points; Correcting the initial weight coefficient according to the time difference of the mutation points of the sensor data corresponding to adjacent monitoring points within the preset time window to obtain a time-series weight coefficient; Adjust the temporal weight coefficient according to the structural connection relationship between adjacent monitoring points to obtain the final weight coefficient, and different structural connection relationships correspond to different preset adjustment strategies for the temporal weight coefficient.

[0013] Through the above embodiments, the system obtains the initial weight coefficient by calculating the similarity of the data change rate trends of adjacent monitoring points, and performs multiple corrections in combination with the time difference of mutation points and the structural connection relationship, establishing a complete set of weight coefficient calculation methods. This weight calculation method considering multiple influencing factors can more accurately reflect the correlation degree between monitoring points and provide a reliable basis for identifying the abnormal transmission path.

[0014] In some embodiments, the step of determining the abnormal state propagation rules of each abnormal transmission path under each preset abnormal state according to the historical monitoring data corresponding to each preset abnormal state specifically includes: Use a sliding time window to segment the historical monitoring data, extract the abnormal occurrence time sequences of each monitoring point on the abnormal transmission path in each time window, and establish an abnormal transmission sequence; Calculate the abnormal propagation time interval between each monitoring point based on the abnormal transmission sequence, and determine the abnormal diffusion rate according to the statistical distribution characteristics of the abnormal propagation time interval; Divide the abnormal transmission path into a fast propagation section and a slow propagation section according to the abnormal diffusion rate, and establish a segmented propagation model based on the sensor data of the fast propagation section and the slow propagation section; Interpolate and fit the abnormal data of any monitoring point on the abnormal transmission path according to the segmented propagation model to generate a propagation rule curve of the abnormal state, and the propagation rule curve represents the change relationship of the abnormal state intensity with the propagation distance.

[0015] Through the above embodiments, the system performs time window segmentation analysis on historical data, extracts the abnormal transmission sequence, calculates the propagation time interval, and establishes a segmented propagation model considering the difference in propagation rate. This method can accurately depict the diffusion law of the abnormal state in space and time, facilitating the subsequent prediction of the abnormal development trend.

[0016] In some embodiments, the step of generating a warning message including the abnormal location, the influence range, and the development trend if it is detected that the difference between the predicted values of a preset proportion of the parameters and the corresponding actual monitoring values exceeds a third preset threshold specifically includes: Calculate an abnormal diffusion index based on the difference between the predicted value of the parameter and the actual monitoring value, and the abnormal diffusion index is the ratio of the number of monitoring points exceeding the third preset threshold to the total number of monitoring points on the abnormal transmission path; Determine the anomaly level based on the relationship between the time change rate of the abnormal diffusion index and a preset diffusion threshold, where the anomaly level includes a fast diffusion type anomaly and a slow diffusion type anomaly; Divide the monitoring points with a distance less than a preset distance threshold and the same anomaly level into the same anomaly influence area, and generate a spatial distribution map of the anomaly influence area; Generate hierarchical early warning information based on the anomaly level and the spatial distribution map, where the hierarchical early warning information includes the anomaly level, the diffusion speed, the boundary coordinates of the influence area, and the maximum anomaly degree within each influence area.

[0017] Through the above embodiments, the system can classify anomalies into fast diffusion type and slow diffusion type by calculating the abnormal diffusion index and combining the time change rate, and divide the influence area according to the spatial distribution characteristics. This hierarchical early warning mechanism can not only reflect the severity of the anomaly, but also predict its development trend and influence range, providing a decision-making basis for emergency response.

[0018] In some embodiments, after the step of generating early warning information including the abnormal location, the influence range, and the development trend if it is detected that the difference between the predicted values of a preset proportion of the parameters and the corresponding actual monitoring values exceeds a third preset threshold, the method further includes: Count the failure types, failure locations, and failure frequencies of each component of the terminal facilities within a preset time period, and generate a failure statistics table; According to the failure statistics table, combine the same type of failures with a failure frequency exceeding a preset number and adjacent locations into a failure mode; Analyze the occurrence time sequence of each failure in the failure mode, establish an association relationship between the failures with a time interval less than a preset duration, and form a failure association network; Divide the detection priorities for the locations where the failure modes are located according to the association degrees of the failure nodes in the failure association network, where the association degree is the number of failures directly connected to the failure node, and different detection priorities correspond to different manual inspection periods; Generate a maintenance schedule based on the detection priorities and the corresponding manual inspection periods.

[0019] Through the above embodiments, the system can identify relevant failure modes by statistically analyzing the failure types, locations, and frequencies and establishing a failure association network. Divide the detection priorities based on the association degrees of the failure nodes and formulate corresponding inspection periods, realizing the scientific formulation of the maintenance plan. This maintenance strategy based on failure association analysis can improve the maintenance efficiency and reduce the maintenance cost.

[0020] In a second aspect, the present application provides a terminal facility health monitoring system, where the terminal facility health monitoring system includes: one or more processors and a memory; The memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions. The one or more processors call the computer instructions so that the wharf facility health monitoring system can implement a wharf facility health monitoring method based on digital twin and machine learning provided in the above embodiments, which will not be elaborated here.

[0021] In a third aspect, the present application provides a computer-readable storage medium, including instructions, which when running on a wharf facility health monitoring system, enable the wharf facility health monitoring system to implement a wharf facility health monitoring method based on digital twin and machine learning provided in the above embodiments, which will not be elaborated here.

[0022] In a fourth aspect, the present application provides a computer program product, which when running on a wharf facility health monitoring system, enables the wharf facility health monitoring system to implement a wharf facility health monitoring method based on digital twin and machine learning provided in the above embodiments, which will not be elaborated here.

[0023] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. By analyzing the synchronous changes of multi-type sensor data to establish a network topology map and weight coefficients, and combining historical data to construct abnormal state propagation rules, the transformation from single-point monitoring to networked monitoring is realized. This method can not only address the false alarm problem easily generated by traditional single-point threshold alarms, but also accurately identify the transmission path and diffusion trend of abnormal states, significantly improving the accuracy and reliability of early warnings.

[0024] 2. By constructing a multi-dimensional feature layer including material properties, load distribution, environmental corrosion, and construction quality in the digital twin model, and calculating the combined influence coefficient to evaluate the importance level of the monitoring area, the optimal allocation of monitoring resources is realized. This monitoring strategy based on multi-dimensional feature fusion ensures the effective coverage of the monitoring system for key areas and improves the overall monitoring efficiency.

[0025] 3. By establishing a fault statistical analysis and correlation network model, identifying relevant fault modes, and dividing the detection priorities based on the correlation degree of fault nodes, a data-driven intelligent maintenance system is formed. This method breaks through the limitations of traditional fixed-cycle maintenance, realizes a differential maintenance strategy based on fault correlation relationships, effectively improves the maintenance efficiency and reduces the maintenance cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1It is a schematic flowchart of a wharf facility health monitoring method based on digital twin and machine learning in an embodiment of the present application; Figure 2 It is another schematic flowchart of a wharf facility health monitoring method based on digital twin and machine learning in an embodiment of the present application; Figure 3 It is a schematic diagram of the structure of an entity device of a wharf facility health monitoring system in an embodiment of the present application. Detailed implementation manners

[0027] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and appended claims of the present application, the singular forms "a", "an", "the", "above-mentioned", "said", and "this" are also intended to include the plural forms, unless clearly indicated to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any and all possible combinations including one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or indicating relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0029] It should be noted that for the convenience of understanding, in the embodiments of the present application, the wharf facility health monitoring system may also be simply referred to as the "system", and this abbreviation will not affect the protection scope of the present application.

[0030] For the convenience of understanding, the method provided in this embodiment is described in terms of a process below. Please refer to Figure 1 , which is a schematic flowchart of a wharf facility health monitoring method based on digital twin and machine learning in an embodiment of the present application.

[0031] S101. Real-time collect the monitoring data of the wharf facilities according to the multi-type sensor network deployed at multiple preset monitoring points.

[0032] Among them, the preset monitoring points refer to the positions that are pre-determined on the wharf facilities (such as wharf structural members, key nodes, etc.) and need to be monitored key points; the multi-type sensor network refers to a monitoring network composed of a variety of different types of sensors (such as stress-strain sensors, displacement sensors, temperature sensors, corrosion monitoring sensors, etc.), which is used to represent the physical state data (such as stress, displacement, temperature, corrosion degree, etc.) of the wharf facilities collected in real time from multiple dimensions.

[0033] Specifically, the wharf facility health monitoring system obtains the physical parameter data of the wharf facility in real time according to the set sampling frequency through various sensors (such as strain gauges, inclinometers, accelerometers, etc.) pre-deployed at key positions of the wharf structure. These sensors are distributed in different areas according to the monitoring requirements (such as wharf pile foundations, beam slabs, berthing members, etc.) to form a monitoring network covering the entire facility.

[0034] In addition, after obtaining the monitoring data, the system can also construct a multi-level monitoring data visualization layer in the digital twin model. This visualization layer includes a global monitoring layer and a local monitoring layer. The global monitoring layer is used to display the comprehensive monitoring indicators of each monitoring area (such as aggregated values like average stress and cumulative displacement), and the local monitoring layer is used to display the sensor data of the target area (such as the real-time values of stress and displacement at specific points). When the change rate of the comprehensive monitoring indicator within a preset time period (such as 1 hour) does not exceed the preset change threshold (such as 5% / hour), the system dynamically displays the change process of the comprehensive indicator in the form of a heat map, dynamic curve, etc. in the global monitoring layer. For example, it reflects the high and low of the indicators in each area through color gradient, and the curve scrolls to show the historical trend. When the change rate of the comprehensive monitoring indicator exceeds the preset change threshold, the system automatically loads the sensor data of the target area (such as the area where the indicator mutates) in the local monitoring layer, and displays the change process of the sensor data in this area in ways such as highlighting the three-dimensional model points, real-time data list, and dynamic waveform diagram. For example, it draws the stress mutation curve in the past 30 minutes and marks the threshold range to assist the operation and maintenance personnel in deeply analyzing the abnormal details.

[0035] S102. Classify the monitoring data within the preset time period according to the parameter type to obtain multiple monitoring parameter subsets.

[0036] Specifically, the system screens and classifies all the monitoring data collected within the preset time period according to the type tags of the sensors (such as "stress sensor", "displacement sensor", etc.). For example, the data of all stress and strain sensors are classified into the "stress parameter subset", and the data of displacement sensors are classified into the "displacement parameter subset", and so on. During the classification process, the system automatically identifies the unique identifier of the sensor (such as MAC address, device number) and its corresponding parameter type to ensure accurate data classification. The classified data subsets are stored in different tables in the database for subsequent synchronous analysis and network topology construction according to the parameter type.

[0037] S103. Establish the weight coefficient of the monitoring point network topology graph based on the synchronous change value of the same type of sensor data corresponding to adjacent monitoring points in each monitoring parameter subset.

[0038] Specifically, for each subset of monitoring parameters (such as the subset of stress parameters), the system traverses all pairs of adjacent monitoring points and calculates the synchronization index of the same type of sensor data within a preset time window (such as the past 24 hours). For example, for two adjacent strain monitoring points A and B, the correlation coefficient of their strain data sequences is calculated as the initial synchronization value. Then, the initial value is corrected by combining the structural connection relationship between adjacent monitoring points (such as whether they are directly connected by components) and the time difference of data mutation points (such as how long after the data mutation at point A does a similar mutation occur at point B) to obtain the final weight coefficient. For example, if A and B are connected by the same beam and the data mutation time difference is less than 5 minutes, the weight coefficient is increased by 20%; if it is a non-directly connected structure, the weight coefficient is decreased by 10%. Finally, each parameter subset corresponds to a weighted network topology graph for subsequent identification of abnormal transmission paths.

[0039] Optionally, the system calculates the sliding window correlation coefficient of the data sequences of adjacent monitoring points (the window size is 1 hour), and takes the average correlation coefficient within the past 24 hours as the initial weight. Detect mutation points in the data sequence (such as points with a change amplitude exceeding 5%), calculate the time difference between mutation points of adjacent monitoring points. If the time difference is less than 10 minutes, the weight coefficient is increased by 0.1; if it is greater than 30 minutes, the weight coefficient is decreased by 0.2. Determine the connection type between adjacent monitoring points (such as rigid connection, hinge connection) according to the structural design drawings. The weight coefficient of rigid connection is multiplied by 1.2, and that of hinge connection is multiplied by 0.8, etc., which is not limited here.

[0040] Optionally, the system can also calculate the theoretical strain transfer coefficient of adjacent monitoring points under the action of unit load according to the finite element model of the wharf structure as the initial weight. Compare the difference between the actual monitoring data and the theoretical transfer coefficient. When the difference is less than 5%, the weight coefficient remains unchanged; when the difference is greater than 10%, the weight coefficient is adjusted by ±0.2 (according to the direction of the difference). Finally, considering the influence of environmental factors (such as temperature change) on the structural response, the weight coefficient is corrected through a temperature compensation algorithm. For example, when the temperature changes by 10°C, the weight coefficient is adjusted by 0.05, which is not limited here.

[0041] S104. In the digital twin model corresponding to the wharf facilities, determine the target path corresponding to the network topology path where the adjacent weight coefficients are all greater than the first preset threshold as the abnormal transmission path for the corresponding same type of sensor data.

[0042] Among them, the digital twin model refers to a virtual mapping body of dock facilities constructed through 3D modeling technology, which integrates information such as structural geometric parameters, material properties, and sensor layouts, and is used to represent the real-time mirror image and simulation analysis of physical entities; the network topology path refers to a combination of a series of connected edges from the starting point to the ending point in the network topology diagram of monitoring points, and is used to represent the transmission path of data or abnormal states between monitoring points; the abnormal transmission path refers to the path in the digital twin model where when an abnormality occurs at a certain monitoring point, the abnormal state may be transmitted to other monitoring points through strongly associated paths, and is used to represent the potential channel for abnormal diffusion.

[0043] Specifically, based on the network topology diagram corresponding to each subset of monitoring parameters, the system traverses all possible network topology paths in the digital twin model. For the adjacent edges on each path (i.e., the connections between adjacent monitoring points), check whether their weight coefficients are all greater than the first preset threshold. For example, if a certain path contains monitoring points A→B→C, where the weight coefficient of A-B is 0.7 and the weight coefficient of B-C is 0.8, both greater than the preset threshold of 0.6, then this path is determined as an abnormal transmission path. For different parameter types (such as stress, displacement), the system can generate corresponding abnormal transmission paths respectively and mark them with different colors or line types in the digital twin model for quick call during subsequent abnormal detection.

[0044] Optionally, for the network topology diagram of each subset of monitoring parameters, the system can use the Dijkstra algorithm or the Floyd-Warshall algorithm to calculate the shortest paths between all node pairs (with the weight coefficient as the edge weight). Screen out the paths where the weight coefficients of all edges in the path are greater than the first preset threshold as candidate abnormal transmission paths. Visualize these paths in the digital twin model and sort them according to the path length (number of edges). Since abnormal diffusion is usually faster for short paths, short paths (such as 2-3 hop paths) can be preferentially marked and displayed.

[0045] S105. Determine the abnormal state propagation rules of each abnormal transmission path under each preset abnormal state according to the historical monitoring data corresponding to each preset abnormal state.

[0046] Among them, the preset abnormal state refers to the abnormal types of dock facilities predefined by the system (such as structural cracks, increased corrosion, displacement exceeding the limit, etc.); the abnormal state propagation rule refers to a mathematical model or curve that describes the diffusion speed, intensity attenuation, and spatial distribution law of the abnormal state on the abnormal transmission path.

[0047] Specifically, for each preset abnormal state (such as structural damage caused by stress exceeding the limit), the system collects the corresponding historical monitoring data (including the time series data of each monitoring point when the abnormality occurs). For each abnormal transmission path, a sliding time window (such as a window size of 2 hours and a step size of 1 hour) is used to segment the historical monitoring data, and the transmission order of the abnormality on the path within each window is extracted (such as the abnormality at monitoring point A appears before that at monitoring point B), thereby establishing an abnormal transmission sequence. By calculating the time difference between the occurrences of abnormalities at adjacent monitoring points, the time interval distribution of abnormal propagation is statistically obtained (such as the average time interval is 30 minutes), so as to determine the abnormal diffusion rate of this path (such as 0.5 meters per minute). According to the diffusion rate, the path is divided into a fast propagation section (such as from the pile foundation to the bearing platform) and a slow propagation section (such as from the bearing platform to the beam and slab), and linear or exponential propagation models are established respectively. Finally, a propagation rule curve is generated through interpolation fitting. For example, the abnormal intensity of a certain path decays exponentially with the propagation distance, and the formula is: , where, is the initial abnormal intensity, and k is the decay coefficient.

[0048] Optionally, the system can use a recurrent neural network (RNN) to train the time series model of the abnormal transmission path. The input is the abnormal data of the starting monitoring point, and the output is the predicted data of other monitoring points on the path. The model parameters are the propagation rules. The key sections of abnormal propagation are identified through the attention mechanism. For example, in a certain path, the propagation from monitoring point B to C has the greatest impact on the overall trend, and a higher model weight is assigned to it. The propagation process of rare abnormal states can also be simulated using a generative adversarial network (GAN) to supplement the rules for extreme cases missing in the historical data, which is not limited here.

[0049] S106. When the sensor data of any monitoring point is detected to exceed the corresponding second preset threshold range, calculate the parameter prediction values of other monitoring points in the corresponding abnormal transmission path based on the abnormal state propagation rules.

[0050] Specifically, when the system detects that the data of a certain monitoring point (such as the stress sensor of pile foundation A) exceeds the second preset threshold (such as 30 MPa), first determine the abnormal transmission path to which the monitoring point belongs (such as the path of pile foundation A → pile cap B → beam and slab C screened by weight coefficients). Then, according to the abnormal state propagation rule corresponding to this path (such as the diffusion rate of stress anomaly is 0.5 MPa·m⁻¹·min⁻¹), combined with the current abnormal intensity (such as the stress of pile foundation A is 35 MPa, exceeding the threshold by 5 MPa) and the propagation time (such as 10 minutes have passed), calculate the parameter prediction values of other monitoring points (pile cap B, beam and slab C) on the path. For example, assume the distance between pile foundation A and pile cap B is 2 meters, and the diffusion rate is 0.5 MPa·m⁻¹·min⁻¹, then the predicted stress of pile cap B after 10 minutes is 50 Mpa. After the prediction values are calculated, the system compares them with the actual monitoring data in real time to determine whether the anomaly is truly spreading.

[0051] S107. If it is detected that the difference between the predicted values of a preset proportion of parameters and the corresponding actual monitoring values exceeds the third preset threshold, an early warning message including the abnormal location, influence range, and development trend is generated.

[0052] Specifically, the system calculates the differences between the predicted values and the actual monitoring values of all monitoring points on the abnormal transmission path, and counts the proportion of the number of monitoring points whose differences exceed the third preset threshold. For example, a certain path has a total of 5 monitoring points, and the differences of 3 of them exceed the threshold, accounting for 60% (reaching the preset proportion), then it is determined that there is a real abnormal diffusion. Then, according to the abnormal transmission path and the spatial positions of the monitoring points, determine the abnormal location (such as pile foundation A) and the influence range (such as the area where pile cap B and beam and slab C are located). By analyzing the time change rate of the abnormal diffusion index (the proportion of monitoring points exceeding the threshold) (such as increasing by 5% per minute), judge whether the abnormal level is a fast diffusion type or a slow diffusion type, and predict the future trend (such as the influence range will expand to the adjacent path within 30 minutes). Finally, generate a visual early warning message including the abnormal location coordinates, the boundary of the affected area, the diffusion speed, and the trend prediction for the next 2 hours, and push it to the terminal device of the operation and maintenance personnel.

[0053] In the above embodiment, the system can identify the transmission path of the abnormal state and avoid false alarms caused by single-point data fluctuations by establishing a network topology map and weight coefficients based on the synchronous change of multi-type sensor data. The abnormal state propagation rule established in combination with historical data can predict the diffusion trend of the abnormal state. When an anomaly is detected, by comparing the difference between the predicted value and the actual monitoring value, it can be more accurately determined whether it is a real anomaly, and an early warning message including the abnormal location, influence range, and development trend is given, improving the accuracy and reliability of the early warning.

[0054] The following is a further and more specific process description of the method provided in this embodiment. Please refer toFigure 2 , which is another process schematic diagram of a wharf facility health monitoring method based on digital twin and machine learning in the embodiment of this application.

[0055] S201. Construct a multi-dimensional feature layer in the digital twin model corresponding to the wharf facility.

[0056] This step is executed when the digital twin model is initially constructed after system initialization or when a wharf facility is newly built / transformed, and can be updated regularly according to facility maintenance records or environmental changes subsequently.

[0057] Specifically, the system creates four independent feature layers in the digital twin model based on the design drawings, construction records, and environmental monitoring data of the wharf facility. For example, in the material property layer, the compressive strength parameter of C30 concrete and the elastic modulus of steel are assigned to each pile foundation member; in the load distribution layer, the impact force distribution when a ship berths is defined according to the wharf operation log; in the environmental corrosion layer, the salt spray corrosion rate is set in combination with local meteorological data; in the construction quality layer, the ultrasonic inspection report of pile foundation construction is imported, and the areas with honeycombing and pockmarks are marked. Each feature layer is bound to the components in the three-dimensional model one by one through the Geographic Information System (GIS) to form a virtual model with multi-dimensional data integration.

[0058] Among them, the multi-dimensional feature layer refers to the virtual level divided according to different influencing factors in the digital twin model, which is used to represent the set of multi-dimensional physical attributes that affect the health state of the wharf facility, including the material property layer, the load distribution layer, the environmental corrosion layer, and the construction quality layer. The material property layer refers to the level that stores the material property data of each component of the wharf facility, including the concrete strength grade, the yield strength of steel, the material aging coefficient, etc., and is used to represent the influence of the performance characteristics of the material itself on the structural health; the load distribution layer refers to the level that records various load data borne by the wharf facility, including self-weight load, ship impact load, cargo stacking load, tidal load, etc., and is used to represent the action effect of external loads on the structure; the environmental corrosion layer refers to the level that characterizes the corrosion effect of the marine environment on the wharf facility, including data such as salt spray concentration, water level change frequency, and microbial erosion degree, and is used to represent the degradation of structural performance caused by environmental factors; the construction quality layer refers to the level that reflects the construction technology and quality during the construction process of the wharf facility, including concrete pouring defects, weld quality grades, foundation treatment depths, etc., and is used to represent the potential influence of construction legacy problems on the structural health.

[0059] S202. Calculate the weighted superposition value of the feature data and the corresponding weight coefficients in different feature layers in each monitoring area to obtain the combined influence coefficient.

[0060] Specifically, the system divides the digital twin model into multiple monitoring areas (for example, divides the wharf into 10 monitoring areas according to structural segments), and extracts the feature data of four feature layers for each monitoring area. For example, the feature data of a certain pile foundation monitoring area includes: the concrete strength of 25 MPa in the material property layer (the designed value is 30 MPa, and the strength is reduced due to aging), the peak value of ship impact force of 500 kN in the load distribution layer, the salt spray corrosion rate of 0.01 mm / year in the environmental corrosion layer, and the pile foundation integrity of 90% in the construction quality layer. According to the preset weight coefficients (material property 0.4, load distribution 0.3, environmental corrosion 0.2, construction quality 0.1), the combined influence coefficient is calculated.

[0061] S203. Determine the importance level corresponding to each monitoring area according to the combined influence coefficient, and determine the target monitoring areas with importance levels higher than the preset level threshold as the preset monitoring points.

[0062] Specifically, the system divides the importance levels according to the numerical range of the combined influence coefficient. For example, it is set that the combined influence coefficient > 200 is level I (high risk), 100 - 200 is level II (medium risk), < 100 is level III (low risk), and the preset level threshold is level II. The combined influence coefficient of a certain monitoring area is 180 (level II), which is higher than the threshold, so it is determined as a preset monitoring point. The system marks these areas in the digital twin model and generates a monitoring point deployment list, including the recommended sensor types (such as stress, displacement, and corrosion sensors need to be deployed in level I areas, and at least stress sensors need to be deployed in level II areas) and installation locations (such as the top and middle of high-risk pile foundations).

[0063] In the above embodiment, the system constructs a multi-dimensional feature layer including material properties, load distribution, environmental corrosion, and construction quality in the digital twin model, calculates the combined influence coefficient of different features, and can evaluate the importance levels of each monitoring area. Determining the layout of the preset monitoring points based on the importance levels ensures the reasonable allocation of monitoring resources, enables the monitoring system to more specifically cover key areas, and improves the monitoring efficiency and economy.

[0064] S204. Calculate the initial weight coefficient according to the similarity of the change rate trends of the data change rates of adjacent monitoring points within the preset time window.

[0065] Specifically, for each subset of monitoring parameters (such as the subset of displacement parameters), the system traverses all pairs of adjacent monitoring points (such as monitoring points A and B), and extracts the sequence of minute-level data change rates within a preset time window (such as the most recent 6 hours). For example, calculate the displacement change rate of monitoring point A every 10 minutes (unit: mm / h) to form a change rate sequence containing 36 data points. Through the dynamic time warping (DTW) algorithm or the Pearson correlation coefficient, calculate the similarity value of the change rate sequences of adjacent monitoring points (such as the correlation coefficient is 0.8), and use it as the initial weight coefficient. The higher this coefficient, the more consistent the data change trends of adjacent monitoring points and the stronger the basic synchronization.

[0066] S205. Adjust the initial weight coefficient based on the time difference between the mutation points of the corresponding sensor data of adjacent monitoring points within the preset time window and the structural connection relationship between adjacent monitoring points to obtain the final weight coefficient.

[0067] Specifically, the system first detects the mutation points of adjacent monitoring points within the preset time window (such as identifying mutation points by setting the mutation threshold to 3 times the standard deviation of the data), calculates the mutation point time difference Δt. If Δt is less than the preset threshold (such as 10 minutes), it is determined that the abnormal conduction is rapid, and the initial weight coefficient is increased by Δw1 (such as 0.1); if Δt is greater than the threshold, it is determined that the conduction is lagging, and the coefficient is reduced by Δw2 (such as 0.05), where both Δw1 and Δw2 are time series weight coefficients. Then, adjust the weight according to the structural connection relationship: the coefficient of rigid connection (such as the same continuous beam) is multiplied by the adjustment factor k1 = 1.2, the hinged connection is multiplied by k2 = 0.9, and the non-direct connection (such as indirectly connected through the foundation) is multiplied by k3 = 0.7. The final weight coefficient w is: w = (initial weight coefficient + Δw) × k For example, the initial weight is 0.8, the mutation time difference is 5 minutes (+0.1), and the rigid connection (×1.2), the final weight w is (0.8 + 0.1) × 1.2 = 1.08 (which can be normalized to 1).

[0068] In the above embodiment, the system obtains the initial weight coefficient by calculating the similarity of the data change rate trends of adjacent monitoring points, and performs multiple corrections in combination with the mutation point time difference and the structural connection relationship, establishing a complete set of weight coefficient calculation methods. This weight calculation method considering multiple influencing factors can more accurately reflect the degree of association between monitoring points and provide a reliable basis for identifying the abnormal transmission path.

[0069] S206. When determining the monitoring point of the abnormal sensor data and the parameter prediction values of other monitoring points in the corresponding abnormal transmission path, calculate the abnormal diffusion index based on the difference between the parameter prediction value and the actual monitoring value.

[0070] Specifically, the system first determines the abnormal initial monitoring point (such as pile foundation A) and its affiliated abnormal transmission path (such as A→B→C) through step S106. Calculate the predicted parameter values of other monitoring points (B, C) on the path according to the propagation rule (such as the predicted stress at point B is 45 MPa and at point C is 38 MPa). Then, obtain the actual monitoring values of these monitoring points (such as the actual value at point B is 40 MPa and at point C is 35 MPa), and calculate the difference (the difference at point B is 5 MPa and at point C is 3 MPa). If the third preset threshold is 4 MPa, then the difference at point B exceeds the threshold and the difference at point C does not. The abnormal diffusion index is 1 / 2 == 0.5 (there are 2 total monitoring points on the path and 1 exceeds the threshold).

[0071] S207. Determine the abnormal level according to the relationship between the time change rate of the abnormal diffusion index and the preset diffusion threshold, and divide the monitoring points with a distance less than the preset distance threshold and the same abnormal level into the same abnormal influence area, and generate hierarchical early warning information including the abnormal level and the spatial distribution map of the abnormal influence area.

[0072] Specifically, the system calculates the time change rate of the abnormal diffusion index. If the time change rate is greater than the preset diffusion threshold (such as 0.05 / minute), it is determined as a fast-diffusion type of abnormality; otherwise, it is a slow-diffusion type. Then, based on the spatial coordinates of the monitoring points, the monitoring points with a distance less than the preset distance threshold (such as 10 meters) and the same abnormal level are divided into the same area. For example, for monitoring points B (8 meters away from point A) and C (12 meters away from point A), if B is of the fast-diffusion type and C is of the slow-diffusion type, then B is separately divided into the fast-diffusion area and C is included in the slow-diffusion area. Finally, generate a spatial distribution map in the digital twin model, mark the fast-diffusion area in red, the slow-diffusion area in yellow, and mark the boundary coordinates and the maximum abnormal degree of each area (such as the difference at point B is 5 MPa).

[0073] In the above embodiment, the system can classify the abnormality into fast-diffusion type and slow-diffusion type by calculating the abnormal diffusion index and combining the time change rate, and divide the influence area according to the spatial distribution characteristics. This hierarchical early warning mechanism can not only reflect the severity of the abnormality, but also predict its development trend and influence range, providing a decision-making basis for emergency disposal.

[0074] S208. Count the failure types, failure locations, and failure frequencies of each component of the wharf facilities within a preset time period, and combine the same type of failures with failure frequencies exceeding the preset number and adjacent locations into failure modes.

[0075] Specifically, the system can extract the fault data within a preset time period from historical early warning records and manual fault reports, and classify and count them according to component types (such as pile foundations, beam slabs), fault types (such as corrosion, cracks), and locations (such as the area 10 - 20 meters from the quay front line) to form a fault statistics table.

[0076] The system traverses the records in the fault statistics table. For the same fault type (such as corrosion), it filters out the records with a fault frequency exceeding a preset number of times (such as ≥3 times / year). Then, it calculates the spatial distance between the fault locations of these records. If the distance is less than the preset distance threshold (such as 10 meters), these faults are combined into a fault mode. For example, 3 pile foundations (with spacings all less than 8 meters) in a certain area have all had 4 corrosion faults in the past year, meeting the frequency and location conditions. The system combines them into a "pile foundation corrosion cluster fault mode" and records the central location, influence range, and fault type of this mode.

[0077] S209. Analyze the occurrence time sequence of each fault in the fault mode, and establish an association relationship between the faults with a time interval less than the preset duration to form a fault association network.

[0078] Specifically, the system sorts the faults in each fault mode by the occurrence time and calculates the time interval between adjacent faults (such as fault A occurred on January 5, 2024, and fault B occurred on January 15, 2024, with an interval of 10 days). If the time interval is less than the preset duration (such as 30 days), a directed edge is established between these two faults (such as A→B, indicating that A occurred before B). By traversing all fault pairs, a fault association network including time associations is formed. For example, a certain fault mode contains 5 corrosion faults, and the intervals of 3 of these faults are all less than 20 days, forming a chain association network of A→B→C, indicating a possible trend of corrosion spreading gradually.

[0079] S210. Divide the detection priority of the location where the fault mode is located according to the association degree of each fault node in the fault association network, and generate a maintenance schedule in combination with the corresponding manual inspection period.

[0080] Specifically, the system calculates the correlation degree of each node in the fault correlation network (i.e., the number of directly connected faults of the fault node in the correlation network), sorts them from high to low according to the correlation degree, and divides them into three levels of detection priorities: high (correlation degree ≥ 3), medium (correlation degree = 2), and low (correlation degree ≤ 1). For example, a node with a correlation degree of 4 is classified as a high priority, and the corresponding manual inspection period is once a week; a node with a correlation degree of 2 is classified as a medium priority, and the period is once a month. The system generates a maintenance schedule according to the priority and period, including information such as the fault mode location, detection priority, next inspection time, and responsible engineer, and pushes it to relevant personnel through emails or system messages.

[0081] In the above embodiment, the system can identify relevant fault modes by establishing a fault correlation network through statistical analysis of fault types, locations, and frequencies. The detection priorities are divided based on the correlation degree of the fault nodes, and corresponding inspection periods are formulated, realizing the scientific formulation of the maintenance plan. This maintenance strategy based on fault correlation analysis can improve the maintenance efficiency and reduce the maintenance cost.

[0082] The dock facility health monitoring system of the embodiment of the present invention is applied to electronic devices. Figure 3 It shows a schematic architecture diagram of an electronic device suitable for implementing the embodiment of the present invention.

[0083] It should be noted that Figure 3 The shown electronic device is only an example and should not bring any limitation to the functions and usage scopes of the embodiment of the present invention.

[0084] Those of ordinary skill in the art can understand that all or part of the steps in the above various methods can be completed by instructions (computer programs), or the relevant hardware can be controlled by instructions (computer programs). The instructions can be stored in a computer-readable storage medium and loaded and executed by a processor. The electronic device of this embodiment includes a storage medium and a processor. Among them, multiple instructions are stored in the storage medium, and these instructions can be loaded by the processor to execute any step of the method provided by the embodiment of the present invention.

[0085] Specifically, the storage medium and the processor are electrically connected directly or indirectly to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more signal lines. The storage medium stores computer-executable instructions for implementing the data access control method, including at least one software function module that can be stored in the storage medium in the form of software or firmware. The processor executes various functional applications and data processing by running the software programs and modules stored in the storage medium. The storage medium can be, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. Among them, the storage medium is used to store programs, and the processor executes the programs after receiving the execution instructions.

[0086] Further, the software programs and modules in the above storage medium may further include an operating system, which may include various software components and / or drivers for managing system tasks (such as memory management, storage device control, power management, etc.), and may communicate with various hardware or software components to provide a running environment for other software components. The processor can be an integrated circuit chip with signal processing capabilities. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc., which can implement or execute the various methods, steps, and logic flow block diagrams disclosed in this embodiment. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0087] Since the instructions stored in the storage medium can execute the steps in any of the methods provided in the embodiments of the present invention, the beneficial effects of any of the methods provided in the embodiments of the present invention can be achieved. For details, see the previous embodiments and will not be repeated here.

[0088] As described above, it is only the preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A terminal facility health monitoring method based on digital twins and machine learning, applied to a terminal facility health monitoring system, characterized in that: The method comprises: Real-time monitoring data of terminal facilities is collected based on a multi-type sensor network deployed at multiple preset monitoring points; Classifying the monitoring data within a preset time period according to parameter types to obtain a plurality of monitoring parameter subsets, wherein the monitoring parameter subsets include a plurality of sensor data of the same type; A weight coefficient of a monitoring point network topology map is established according to the synchronous change values ​​of the same type of sensor data corresponding to adjacent monitoring points in each monitoring parameter subset, and the monitoring point network topology map corresponds to the monitoring parameter subset one by one; In the digital twin model corresponding to the terminal facility, the target path corresponding to the network topology path whose adjacent weight coefficients are all greater than the first preset threshold is determined as an abnormal transmission path corresponding to the same type of sensor data; Determine the abnormal state propagation rules of each abnormal transmission path under each preset abnormal state according to the historical monitoring data corresponding to each preset abnormal state; When it is detected that the sensor data of any monitoring point exceeds the corresponding second preset threshold range, the parameter prediction values ​​of other monitoring points in the corresponding abnormal transmission path are calculated based on the abnormal state propagation rule; If it is detected that the difference between a preset proportion of the parameter prediction values ​​and the corresponding actual monitoring values ​​exceeds a third preset threshold, early warning information including the abnormal location, impact range and development trend is generated.

2. The method according to claim 1, characterized in that Before the step of collecting monitoring data of the terminal facilities in real time based on the multi-type sensor network deployed at multiple preset monitoring points, the method further includes: Constructing a multi-dimensional feature layer in the digital twin model corresponding to the terminal facility, wherein the multi-dimensional feature layer includes a material property layer, a load distribution layer, an environmental corrosion layer, and a construction quality layer; Calculate the weighted superposition value of the feature data of different feature layers in each monitoring area and the corresponding weight coefficient to obtain a combined influence coefficient, wherein the digital twin model includes multiple monitoring areas, and the weight coefficient is determined based on a preset expert score; Determine the importance level corresponding to each of the monitoring areas according to the combined influence coefficient; The target monitoring area whose importance level is higher than the preset level threshold is determined as a preset monitoring point.

3. The method according to claim 1, characterized in that After the step of collecting monitoring data of the terminal facilities in real time based on the multi-type sensor network deployed at multiple preset monitoring points, the method further includes: Constructing a multi-level monitoring data visualization layer in the digital twin model, wherein the visualization layer includes a global monitoring layer and a local monitoring layer, wherein the global monitoring layer displays the comprehensive monitoring indicators of each monitoring area, and the local monitoring layer displays the sensor data of the target area; When the change rate of the comprehensive monitoring indicator within the preset time period does not exceed the preset change threshold, the change process of the comprehensive monitoring indicator is dynamically displayed at the global monitoring layer; When the change rate of the comprehensive monitoring index exceeds a preset change threshold within a preset time period, the change process of the sensor data in the target area is dynamically displayed in the local monitoring layer.

4. The method according to claim 1, characterized in that: The step of establishing the weight coefficient of the monitoring point network topology map according to the synchronous change values ​​of the adjacent monitoring points corresponding to the same type of sensor data in each of the monitoring parameter subsets specifically includes: Calculate the data change rate of each adjacent monitoring point within a preset time window to obtain a change rate trend, wherein the data change rate is the change amplitude of the sensor data per unit time; Calculating an initial weight coefficient according to the similarity of the change rate trends, wherein the initial weight coefficient represents a basic synchronization degree of sensor data at adjacent monitoring points; The initial weight coefficient is modified according to the time difference between the mutation points of the sensor data corresponding to the adjacent monitoring points in the preset time window to obtain the time series weight coefficient; The time series weight coefficient is adjusted according to the structural connection relationship between the adjacent monitoring points to obtain the final weight coefficient. The preset adjustment strategies of the time series weight coefficient corresponding to different structural connection relationships are different.

5. The method according to claim 1, characterized in that: The step of determining the abnormal state propagation rules of each abnormal transmission path under each preset abnormal state according to the historical monitoring data corresponding to each preset abnormal state specifically includes: The historical monitoring data is segmented using a sliding time window, the abnormal occurrence time sequence of each monitoring point on the abnormal transmission path within each time window is extracted, and an abnormal transmission sequence is established; Calculate the anomaly propagation time interval between each monitoring point based on the anomaly transmission sequence, and determine the anomaly diffusion rate according to the statistical distribution characteristics of the anomaly propagation time interval; Dividing the abnormal transmission path into a fast propagation section and a slow propagation section according to the abnormal diffusion rate, and establishing a segmented propagation model based on sensor data of the fast propagation section and the slow propagation section; The abnormal data of any monitoring point on the abnormal transmission path are interpolated and fitted according to the segmented propagation model to generate a propagation rule curve of the abnormal state, wherein the propagation rule curve represents the relationship between the intensity of the abnormal state and the propagation distance.

6. The method according to claim 1, characterized in that If the difference between the predicted parameter values ​​of a preset proportion and the corresponding actual monitoring values ​​exceeds a third preset threshold, the step of generating early warning information including the abnormal location, impact range and development trend specifically includes: Calculating an anomaly diffusion index based on the difference between the parameter prediction value and the actual monitoring value, wherein the anomaly diffusion index is the ratio of the number of monitoring points exceeding the third preset threshold to the total number of monitoring points on the anomaly transmission path; Determine an abnormality level according to the relationship between the time change rate of the abnormal diffusion index and a preset diffusion threshold, wherein the abnormality level includes a fast diffusion type abnormality and a slow diffusion type abnormality; The monitoring points whose distance is less than a preset distance threshold and whose abnormality level is the same are divided into the same abnormality impact area, and a spatial distribution map of the abnormality impact area is generated; Based on the abnormality level and the spatial distribution map, graded warning information is generated, where the graded warning information includes the abnormality level, diffusion speed, boundary coordinates of the affected area, and the maximum abnormality degree in each affected area.

7. The method according to claim 1, characterized in that After the step of generating warning information including abnormal location, impact range and development trend if the difference between the predicted parameter values ​​of a preset proportion and the corresponding actual monitoring values ​​exceeds a third preset threshold, the method further includes: Count the fault types, fault locations and fault frequencies of various components of terminal facilities within a preset time period and generate a fault statistics table; According to the fault statistics table, the same type of faults whose fault frequencies exceed a preset number and whose positions are adjacent to each other are combined into a fault mode; Analyze the occurrence sequence of each fault in the fault mode, establish a correlation relationship between faults whose occurrence time interval is less than a preset time length, and form a fault correlation network; According to the correlation degree of each fault node in the fault association network, the detection priority of the location of the fault mode is divided, and the correlation degree is the number of faults directly connected to the fault node. Different detection priorities correspond to different manual inspection cycles; A maintenance schedule is generated according to the detection priority and the corresponding manual inspection cycle.

8. A terminal facility health monitoring system, characterized in that: The terminal facility health monitoring system includes: one or more processors and memory; The memory is coupled to the one or more processors, and the memory is used to store computer program codes, wherein the computer program codes include computer instructions, and the one or more processors call the computer instructions to enable the terminal facility health monitoring system to perform the method according to any one of claims 1 to 7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instruction is executed on a terminal facility health monitoring system, the terminal facility health monitoring system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on a terminal facility health monitoring system, the terminal facility health monitoring system is enabled to perform the method according to any one of claims 1 to 7.

Images