An elevator operation hidden danger accurate identification system and method fusing big data analysis

Abstract

The present application belongs to the technical field of elevator hidden danger identification, and particularly relates to an elevator operation hidden danger accurate identification system and method fusing big data analysis. The present application performs time alignment and operation stage division on multi-source data such as control system, car operation, door machine system, braking system and vibration signal, so that different source data are analyzed under unified working condition dimension, effectively improving data comparability and analysis accuracy. By introducing multi-time scale decomposition mode and dynamic coupling matrix construction mechanism, the change characteristics of each subsystem of the elevator under different time scales are jointly described, the identification ability of complex operation state is improved, the operation state, propagation path and node influence are structuredly expressed by constructing the elevator operation state atlas, the hidden danger propagation process is transformed from experience judgment to graph structure, and the systematicness and explainability of hidden danger identification are improved.

Description

Technical Field

[0001] This invention belongs to the field of elevator hazard identification technology, specifically relating to a precise elevator operation hazard identification system and method that integrates big data analysis. Background Technology

[0002] With the continuous advancement of urbanization, elevators, as indispensable vertical transportation equipment in high-rise buildings and public places, have operational safety that directly affects the safety of personnel passage and the stability of equipment operation. Elevator systems typically consist of a control system, a car operation system, a door operator system, a braking system, and various status monitoring units. During long-term, high-frequency operation, they are susceptible to various factors such as mechanical wear, electrical fluctuations, load changes, and environmental vibrations, which can gradually lead to potential operational anomalies or latent faults.

[0003] In actual operation, elevators exhibit distinct phased characteristics, such as the start-stop phase, the acceleration and constant-speed operation phase, and the braking phase. The response characteristics and energy consumption of each subsystem differ significantly across these phases. Furthermore, the strong coupling and dynamic changes of multi-source operational data over time result in a complex, multi-dimensional temporal evolution of elevator operation. Therefore, multi-source data fusion analysis and operational status identification are crucial for improving elevator safety.

[0004] Existing elevator hazard identification methods mostly rely on a single data source or fixed thresholds for judgment, making it difficult to comprehensively characterize the collaborative operation status and hazard propagation process among multiple elevator subsystems. This can easily lead to misjudgments or omissions under complex operating conditions. Furthermore, existing methods lack the ability to fuse and analyze multi-source operational data, lack a unified mechanism for characterizing multi-scale temporal features and differences across different operational stages, and their utilization of historical hazard data is relatively simple, failing to establish an effective structured memory and reuse mechanism. This hinders the continuous identification and re-triggering analysis of similar hazards, thus limiting the accuracy and adaptability of hazard identification. Summary of the Invention

[0005] The purpose of this invention is to provide a system and method for accurately identifying elevator operation hazards by integrating big data analysis. This system can achieve multi-source data fusion analysis, characterization of hazard propagation process, and reuse of historical hazards, thereby improving the accuracy and stability of elevator operation hazard identification.

[0006] The specific technical solution adopted by this invention is as follows: A method for accurately identifying potential elevator malfunctions by integrating big data analytics includes: Acquire multi-source operation data of the target elevator within a preset operating cycle, and obtain a phased operation dataset based on the multi-source operation data; Multi-scale temporal features are obtained by running the dataset in stages, and a dynamic coupling matrix is ​​constructed. An elevator operation status map is constructed based on multi-scale temporal features and dynamic coupling matrix. The hazard propagation path and node influence weight are obtained based on the elevator operation status map. Dynamic risk propagation results are generated based on the hazard propagation path and node influence weight. The theoretical and actual energy consumption data during elevator operation are obtained, an energy conservation deviation value is generated, the hidden danger propagation path and node influence weight in the elevator operation status map are dynamically corrected, the corrected causal propagation map is obtained, and the dynamic risk propagation results are updated. Construct a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights; Based on the hazard memory weight, the hazard propagation path in the corrected causal propagation map is filtered to obtain the hazard matching path; The hazard matching path is matched with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are superimposed and corrected. Based on the corrected dynamic risk propagation results, the hazard location results, risk level and early warning information are output.

[0007] In a preferred embodiment, multi-source operational data of the target elevator within a preset operating cycle is acquired, and a phased operational dataset is obtained based on the multi-source operational data, including: Acquire multi-source operational data of the target elevator within a preset operating cycle; Multi-source operational data undergoes data preprocessing, which includes outlier removal, missing data completion, and data format standardization to form multi-source operational data. Multiple timestamps are obtained from multi-source runtime data, and time alignment is performed to construct a time runtime data sequence. Based on the time-based operation data sequence, the elevator operation status change feature points are identified, and the elevator operation process is divided into operation stages according to the elevator operation status change feature points, resulting in the start-stop stage, constant speed operation stage and braking stage, and the staged operation dataset is generated.

[0008] In a preferred embodiment, the multi-source operating data includes control system information, car operation information, door operator system information, braking system information, and vibration signal information.

[0009] In a preferred embodiment, multi-scale temporal features are obtained based on a phased running dataset, and a dynamic coupling matrix is ​​constructed, including: The time series decomposition process will be performed on each data point in the phased running dataset to obtain short-period components, medium-period components and long-period components. Based on the short-cycle component, medium-cycle component and long-cycle component, obtain the response change of each data in the phased running dataset within the same time window; Based on the response change, obtain the synchronous change value of each data point in the phased running dataset within the same time window; Obtain the synchronization change matrix corresponding to each time window based on the synchronization change values; Obtain the cumulative change value across the time window based on the synchronous change matrix; The coupling strength matrix is ​​obtained based on the cumulative change values ​​across the time window; Based on the coupling strength matrix, obtain the numerical distribution results of each data point in the phased running dataset at different time scales, and generate a dynamic coupling matrix based on the numerical distribution results.

[0010] In a preferred embodiment, an elevator operation status map is constructed based on multi-scale temporal features and a dynamic coupling matrix. The hazard propagation path and node influence weights are obtained based on the elevator operation status map. Dynamic risk propagation results are then generated based on the hazard propagation path and node influence weights, including: Based on multi-scale temporal features and dynamic coupling matrix, the state transmission values ​​and node response strengths between each data point in the phased operation dataset are obtained, and an elevator operation status map is generated based on the state transmission values ​​and node response strengths. Based on the elevator operation status map, the propagation path sequence and path transmission coefficient between each data point in the phased operation data are obtained, and the hazard propagation path is obtained based on the propagation path sequence and path transmission coefficient. Based on the elevator operation status map and the hazard propagation path, the cumulative impact value of each data in the phased operation dataset in the propagation path is obtained, and the node impact weight is obtained based on the cumulative impact value. Based on the propagation path of the hidden danger and the influence weight of the nodes, the corresponding risk propagation intensity is obtained, and dynamic risk propagation results are generated based on the risk propagation intensity.

[0011] In a preferred embodiment, theoretical and actual energy consumption data during elevator operation are obtained to generate an energy conservation deviation value. This deviation is then used to dynamically correct the hazard propagation paths and node influence weights in the elevator operation status map, resulting in a corrected causal propagation map. Finally, the dynamic risk propagation results are updated, including: Based on the control system data, car operation data, door operator system data, and braking system data during elevator operation, theoretical energy consumption data during elevator operation is obtained. Based on control system data, car operation data, door operator system data, braking system data, and vibration signal data, the actual energy consumption data during elevator operation is obtained. The energy conservation deviation value is obtained by comparing theoretical energy consumption data with actual energy consumption data. Based on the energy conservation deviation value, the hidden danger propagation path in the elevator operation status map is corrected to obtain the corrected hidden danger propagation path, and the node influence weight is corrected to obtain the corrected node influence weight. Based on the corrected hazard propagation path and the corrected node influence weights, obtain the corrected causal propagation map; Based on the corrected causal propagation map, the dynamic risk propagation results are updated to obtain the updated dynamic risk propagation results.

[0012] In a preferred embodiment, a hazard memory database is constructed to obtain historical hazard characteristics and generate corresponding hazard memory weights, including: Obtain historical elevator operation data, and then obtain a historical potential hazard sample set based on the historical elevator operation data; Based on the historical hazard sample set, operational characteristic data corresponding to historical hazard events are obtained, and historical hazard characteristics are obtained. These characteristics include the historical spread range, historical spread depth, and historical risk duration. Obtain a set of hazard characteristics based on historical hazard features; Construct a hazard memory bank based on the hazard feature set; Obtain memory impact values ​​based on historical hazard characteristics; Based on the memory impact value, a corresponding hidden danger memory weight is generated.

[0013] In a preferred embodiment, hazard propagation paths in the corrected causal propagation map are filtered based on hazard memory weights to obtain hazard matching paths, including: Based on the hazard memory weights, obtain the weight distribution results corresponding to each historical hazard feature in the hazard memory database; Based on the weight distribution results and the corrected causal propagation map, multiple candidate hazard propagation paths in the corrected causal propagation map are obtained; Based on multiple candidate hazard propagation paths and their corresponding hazard memory weights, obtain the path matching value for each candidate hazard propagation path; Based on the path matching value, multiple candidate hazard propagation paths are sorted to obtain a path matching sequence; Based on the path matching sequence, candidate hazard propagation paths that meet the preset path conditions are filtered to obtain hazard matching paths.

[0014] In a preferred embodiment, the hazard matching path is matched against a hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are overlaid and corrected. Based on the corrected dynamic risk propagation results, the hazard location result, risk level, and early warning information are output, including: Based on the hazard matching path, the corresponding path feature sequence is obtained, which includes node order information, path propagation intensity information, and path duration information. Path similarity values ​​are obtained based on path feature sequences and historical hazard features; The path similarity value is compared with a preset threshold. When the path similarity value exceeds the preset threshold, the hazard correction condition is determined to be met. Based on the hazard correction conditions, the updated dynamic risk propagation results are corrected to obtain the corrected dynamic risk propagation results. Based on the revised dynamic risk propagation results, obtain the corresponding data on the location of potential hazards and the intensity of their propagation impact; Based on the location data of the potential hazard and the intensity data of its propagation and impact, the system generates hazard location results, risk levels, and early warning information.

[0015] This invention also provides a system for accurately identifying elevator operation hazards by integrating big data analysis, used in the aforementioned method for accurately identifying elevator operation hazards by integrating big data analysis, comprising: The multi-source data module is used to acquire multi-source operation data of the target elevator within a preset operating cycle, and to acquire a phased operation dataset based on the multi-source operation data; The matrix construction module acquires multi-scale temporal features based on the phased execution of the dataset and constructs a dynamic coupling matrix; The propagation results module constructs an elevator operation status map based on multi-scale temporal features and a dynamic coupling matrix, obtains the hazard propagation path and node influence weight based on the elevator operation status map, and generates dynamic risk propagation results based on the hazard propagation path and node influence weight. The risk update module is used to acquire theoretical and actual energy consumption data during elevator operation, generate energy conservation deviation values, dynamically correct the hidden danger propagation paths and node influence weights in the elevator operation status map, obtain the corrected causal propagation map, and update the dynamic risk propagation results. The hazard weighting module is used to build a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights. The hazard path module filters the hazard propagation paths in the corrected causal propagation map based on the hazard memory weight, and obtains the hazard matching path; The hazard results module is used to match the hazard matching path with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are superimposed and corrected, and the hazard location results, risk level and early warning information are output based on the corrected dynamic risk propagation results.

[0016] The technical effects achieved by this invention are as follows: This invention aligns and divides multi-source data, including those from the control system, car operation, door operator system, braking system, and vibration signals, into time-based and operational phases. This allows data from different sources to be analyzed under a unified operating condition dimension, effectively improving data comparability and analytical accuracy. By introducing a multi-timescale decomposition method and a dynamic coupling matrix construction mechanism, it achieves a joint characterization of the changing characteristics of each elevator subsystem at different time scales, enhancing the ability to identify complex operating states. Furthermore, by constructing an elevator operating state map, it provides a structured expression of operating states, propagation paths, and node influences, transforming the hazard propagation process from experience-based judgment into a graphical structure, thus improving the systematic nature and interpretability of hazard identification. This invention introduces a deviation constraint between theoretical and actual energy consumption to dynamically correct the influence weights of hazard propagation paths and nodes, making the analysis results closer to the actual operating state and reducing the impact of modeling bias. By constructing a hazard memory database and introducing hazard memory weights, it achieves continuous accumulation and recall of historical hazard patterns, improving the ability to identify recurring or similar hazards. Through the similarity judgment between hazard matching paths and historical hazard characteristics, it achieves the re-identification and superimposed correction of potential hazards, improving the advance detection and sensitivity of hazard identification. Through the continuous updating and correction of dynamic risk propagation results, it achieves unified output of hazard location results, risk levels, and early warning information, improving the overall efficiency of elevator operation safety management. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method provided by the present invention; Figure 2 This is a system module diagram provided by the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0021] Furthermore, the present invention will be described in detail with reference to the schematic diagrams. When describing the embodiments of the present invention in detail, the schematic diagrams are merely examples for ease of explanation and should not limit the scope of protection of the present invention.

[0022] Please see the appendix Figure 1 As shown, a method for accurately identifying potential elevator malfunctions by integrating big data analytics is provided, including: S1. Obtain multi-source operation data of the target elevator within a preset operating cycle, and obtain a phased operation dataset based on the multi-source operation data; S2. Obtain multi-scale temporal features based on the phased running dataset and construct a dynamic coupling matrix; S3. Construct an elevator operation status map based on multi-scale temporal features and dynamic coupling matrix, obtain the hidden danger propagation path and node influence weight based on the elevator operation status map, and generate dynamic risk propagation results based on the hidden danger propagation path and node influence weight. S4. Obtain theoretical and actual energy consumption data during elevator operation, generate energy conservation deviation value, dynamically correct the hidden danger propagation path and node influence weight in the elevator operation status map, obtain the corrected causal propagation map, and update the dynamic risk propagation results. S5. Construct a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights; S6. Based on the hazard memory weight, the hazard propagation path in the corrected causal propagation map is screened to obtain the hazard matching path; S7. Match the hazard matching path with the hazard memory database. When the matching similarity exceeds the preset threshold, overlay and correct the updated dynamic risk propagation results, and output the hazard location results, risk level and early warning information based on the corrected dynamic risk propagation results.

[0023] As described in steps S1 to S7 above, multi-source operation data of the target elevator within a preset operating cycle is acquired. This multi-source operation data is then time-aligned and divided into operating stages to form a phased operation dataset. This achieves consistent data representation and partitioned management under different operating conditions. Based on the phased operation dataset, time series decomposition and multi-scale feature extraction are performed on various types of operation data. A dynamic coupling matrix is ​​constructed by combining the response change information of each data source within different time windows, thereby obtaining the dynamic change characteristics of each subsystem at different time scales. An elevator operation state map is constructed based on the multi-scale time series features and the dynamic coupling matrix. Each data source is abstracted as a node in the map, and nodes and their corresponding values ​​are constructed using state transmission values ​​and node response intensity. This study identifies the propagation channels between elevators, obtains the propagation paths and node influence weights of potential hazards, and generates dynamic risk propagation results. This achieves a structured representation of the hazard propagation process. Theoretical and actual energy consumption data during elevator operation are introduced. Energy conservation deviation values ​​are used to dynamically correct the hazard propagation paths and node influence weights in the state graph, eliminating the deviation between ideal and actual operation. The causal propagation structure and dynamic risk propagation results are updated to improve the physical consistency of the risk propagation results. A hazard memory bank is constructed, and historical hazard characteristics are extracted based on historical elevator operation data. Corresponding hazard memory weights are generated to characterize the intensity, scope, and persistence of historical hazards, thus forming a sustainable and more... A new set of historical hazard knowledge is used to filter paths in the corrected causal propagation graph based on hazard memory weights. Hazard propagation paths matching historical hazard characteristics are extracted, resulting in hazard matching paths. This allows for the constraint filtering of the current propagation structure by historical hazard patterns. The hazard matching paths are then compared with the hazard memory database for similarity. When the similarity exceeds a preset threshold, a hazard correction mechanism is triggered, superimposing and correcting the updated dynamic risk propagation results. Finally, hazard location results, risk levels, and warning information are output, enabling accurate identification and graded warning of elevator operation hazards. By performing time alignment and operational stage segmentation on multi-source data from the control system, car operation, door operator system, braking system, and vibration signals, data from different sources are unified... Processing data under different operating conditions effectively improves data comparability and analytical accuracy. By introducing multi-timescale decomposition and a dynamic coupling matrix construction mechanism, it achieves a joint characterization of the changing characteristics of various elevator subsystems at different time scales, enhancing the ability to identify complex operating states. By constructing an elevator operating state map, it structurally expresses the operating state, propagation path, and node influence, transforming the hazard propagation process from experience-based judgment into a graphical structure, improving the systematic nature and interpretability of hazard identification. By introducing a deviation constraint between theoretical and actual energy consumption, it dynamically corrects the weights of hazard propagation paths and node influences, making the results closer to the actual operating state and reducing the impact of modeling bias. Finally, it constructs a hazard memory database and introduces hazard memory weights.This system enables the continuous accumulation and recall of historical hazard patterns, improving the ability to identify recurring or similar hazards. By judging the similarity between hazard matching paths and historical hazard characteristics, it allows for the re-identification and overlay correction of potential hazards, enhancing the proactiveness and sensitivity of hazard identification. Furthermore, through continuous updating and correction of dynamic risk propagation results, it achieves unified output of hazard location results, risk levels, and early warning information, thereby improving the overall efficiency of elevator operation safety management.

[0024] In a preferred embodiment, multi-source operation data of the target elevator within a preset operating cycle is acquired, and a phased operation dataset is obtained based on the multi-source operation data, including: S101. Obtain multi-source operation data of the target elevator within a preset operating cycle; S102. Perform data preprocessing on the multi-source operational data, including abnormal data removal, missing data completion, and data format standardization, to form multi-source operational data. S103. Obtain multiple timestamp information based on multi-source running data, perform time alignment processing, and construct a time running data sequence; S104. Identify elevator operation status change feature points based on time operation data sequence, and divide the elevator operation process into operation stages according to the elevator operation status change feature points to obtain the start-stop stage, constant speed operation stage and braking stage, and summarize to generate a staged operation dataset.

[0025] It should be noted that the multi-source operating data includes control system information, car operation information, door operator system information, braking system information, and vibration signal information.

[0026] As described in steps S101 to S104 above, the preset operating cycle is determined by the scheduling parameters in the elevator operation management system. This preset operating cycle is the time interval for continuous collection of elevator operation data, such as a continuous 7-day operating cycle or a fixed daily operating period (e.g., 06:00 to 22:00). Within this preset operating cycle, multi-source operating data is acquired through various acquisition units deployed in the elevator system. Among these, control system information is acquired through the operating command data output by the elevator control motherboard, including floor commands, speed commands, and acceleration / deceleration control signals; car operation information is acquired through position sensors and speed encoders installed on the car or traction system; and door operator system information is acquired through the door operator opening / closing status detection unit. The system obtains braking action and release status information through the brake state feedback loop, and vibration signal information is obtained through vibration sensors installed on the guide rails, car bottom, or traction structure to acquire structural vibration response data. For example, during the elevator's journey from the 1st to the 10th floor, the control system outputs speed curve change commands, the car encoder synchronously records displacement changes, and the vibration sensors synchronously collect acceleration changes, thus forming time-synchronized multi-source operating data. This multi-source operating data undergoes preprocessing to improve data consistency and effectiveness, including three processes: abnormal data removal, missing data completion, and data format standardization. Abnormal data removal is achieved by setting physical constraints, such as setting reasonable limits for car speed data. Within a given time interval, data exceeding the rated speed limit or exhibiting instantaneous abrupt changes are discarded. Missing data is filled using interpolation methods based on time proximity points. For example, when a vibration sensor experiences a short-term disconnection, linear or moving average compensation is performed using sampled data from before and after the connection point. Data format standardization is achieved through a unified data structure, converting heterogeneous data output from different devices into a unified time series format. This includes standardizing the structure for timestamps, data types, and data values, thereby eliminating the impact of differences in sampling frequencies and data formats. For instance, 10ms-level data from the control system and 1ms-level data from the vibration signal are mapped to the same time representation. Various timestamp information, including control system timestamps, is extracted from preprocessed multi-source operational data. The elevator system uses a unified time reference, employing timestamps for car operation, door operator system, and vibration signals. This time alignment eliminates time discrepancies caused by differences in sensor sampling frequencies. During alignment, high-frequency vibration data is aggregated within fixed time windows (e.g., 1ms or 5ms time windows for mean or feature value extraction) and synchronously mapped with low-frequency control data (e.g., 10ms or 100ms sampling). This constructs a data sequence on a unified time axis. The aligned data is then arranged chronologically to form a time-based operational data sequence. This sequence reflects the synchronous evolution of various system states over time during continuous elevator operation.For example, if information on car speed, door operator status, and vibration amplitude is simultaneously available at the same time point, the elevator's operational status change characteristic points are identified based on the time-series data. This is achieved through comprehensive analysis of speed change trends, acceleration change trends, and vibration response changes. For instance, it identifies the acceleration inflection point when the car speed gradually increases from 0 to the rated speed, the deceleration inflection point when it begins to decrease from a stable speed, and the state switching point when the braking system is triggered. Furthermore, it combines vibration signal abrupt change characteristics to enhance identification accuracy. After identifying the state change characteristic points, the entire operation process is divided into intervals based on these points, dividing the elevator operation process into start-stop phases and constant speed phases. The elevator operation is divided into three phases: operation and braking. The start-stop phase includes acceleration and deceleration. The constant-speed operation phase corresponds to the stable operating range, and the braking phase corresponds to the period from braking intervention to complete stop. Multi-source operational data within the time intervals corresponding to each phase are collected and processed to form a phased operational dataset. For example, data from startup to reaching rated speed is included in the start-stop phase dataset. This achieves structured separation of data under different operating conditions. By dividing the elevator operation process into start-stop, constant-speed, and braking phases, data from different operating conditions are processed independently, avoiding feature interference caused by the mixing of different operating states and improving the accuracy of hazard identification.

[0027] In a preferred implementation, multi-scale temporal features are obtained based on a phased running dataset, and a dynamic coupling matrix is ​​constructed, including: S201. Perform time series decomposition on each data point in the phased running dataset to obtain short-period components, medium-period components and long-period components. S202. Based on the short-cycle component, medium-cycle component and long-cycle component, obtain the response change of each data in the phased running dataset within the same time window; S203. Obtain the synchronous change value of each data point in the phased running dataset within the same time window based on the response change amount; S204. Obtain the synchronization change matrix corresponding to each time window based on the synchronization change values; S205. Obtain the cumulative change value across the time window based on the synchronous change matrix; S206. Obtain the coupling strength matrix based on the cumulative change values ​​across the time window; S207. Based on the coupling strength matrix, obtain the numerical distribution results of each data in the phased running dataset at different time scales, and generate a dynamic coupling matrix based on the numerical distribution results.

[0028] As described in steps S201 to S207 above, data is read from the phased operation dataset to obtain the operation data corresponding to the start-stop phase, constant speed operation phase, and braking phase. This operation data includes operating speed, motor drive current, and drive voltage data collected by the control system; displacement and acceleration data collected by the car operation system; door opening and closing speed and door drive current data collected by the door operator system; braking force change data collected by the braking system; and vibration amplitude and frequency data collected by the vibration acquisition device. Since the data during elevator operation simultaneously includes instantaneous changes, periodic changes, and long-term trend changes, it needs to be segmented according to the time scale. The data processing involves establishing time series for each type of operational data based on continuous sampling time. For example, using a sampling interval of 100 milliseconds, 5000 data points are collected within a complete operational cycle to form the corresponding time series. Different time scales are defined based on the duration of data changes: data changes lasting less than 5 seconds and with a rapid rate of change are classified as short-period components, such as the sudden increase in motor current at elevator startup; data changes lasting between 5 and 60 seconds are classified as medium-period components, such as the changes caused by speed adjustments during elevator car operation; and data lasting more than 60 seconds or gradually changing across multiple operational cycles are classified as long-period components, such as those related to engine control. The wear and tear of the power system leads to a gradual decrease in braking force. By decomposing the data, each operational data point is characterized by short-term, medium-term, and long-term features. This avoids the problem of overlooking transient anomalies or long-term degradation trends due to single-time-scale analysis. Time windows are established according to the elevator's operational phases. For example, the start-stop phase is set as the first time window (0-10 seconds), the constant-speed operation phase as the second time window (10-50 seconds), and the braking phase as the third time window (50-60 seconds). The changes in the short-term, medium-term, and long-term components of each operational data point within each time window are obtained. Furthermore, the same data is analyzed... The magnitude of changes between different time points is statistically analyzed. For example, within the braking phase window, the braking pressure data shows a decrease from 8.5 MPa to 7.8 MPa, resulting in a braking pressure change of 0.7 MPa; the vibration signal amplitude increases from 0.12 g to 0.32 g, resulting in a vibration change of 0.20 g; and the drive current decreases from 15 A to 10 A, resulting in a current change of 5 A. Since the units of different data are different, the changes are further converted into standardized response change values. For example, they are proportionally converted according to the historical normal operating range, resulting in a braking pressure response change value of 0.35, a vibration response change value of 0.60, and a current response change value of 0.40. Through processing, the response change of each data point in the phased running dataset within the same time window can be obtained, enabling comparison of data from different sources on a unified scale. This allows for the acquisition of the direction and magnitude of response changes for two data objects within the same time window. For example, if the braking pressure changes by 0.35 and the vibration amplitude changes by 0.60, the degree of synchronization can be determined by judging whether the direction of change is consistent and whether the amplitudes show a corresponding trend. When two data points increase or decrease simultaneously and their amplitudes are similar, the synchronization value is high. When the two data points change in opposite directions or have significantly different amplitudes, the synchronization value is low. For example, in the braking phase, if the braking force decreases by 0.35, the response change can be compared to the previous data point. When the vibration amplitude increases by 0.60, historical hazard sample statistics show that a decrease in braking force and an increase in vibration usually occur simultaneously. Therefore, a higher synchronous change value is assigned, such as 0.85. However, the change in the gantry crane operating current is 0.10, which differs significantly from the change in braking force, so the synchronous change value is only 0.25. The synchronous change value is derived from the degree of consistency in direction and the similarity in amplitude between the response changes. By comparing multiple data points one by one, the synchronous change results between each data point are formed. The synchronous change values ​​between multiple data points are stored in a matrix according to time windows, so that the synchronous change relationship between different operating data forms structured data. The data objects in the matrix are determined, such as control system data, car operation data, and gantry crane system data. Data, braking system data, and vibration signal data are used as row and column data in a matrix, respectively. The corresponding synchronous change values ​​between two data points are filled into their respective positions. For example, if the synchronous change value between braking system data and vibration signal data is 0.85 within a certain braking phase time window, then 0.85 is filled into the corresponding matrix position; if the synchronous change value between braking system data and door operator system data is 0.20, then it is filled into the corresponding position, forming a synchronous change matrix corresponding to a time window. When the elevator operates continuously, generating multiple time windows, multiple synchronous change matrices are generated. For example, if one operating cycle is divided into 10 time windows, 10 sets of synchronous change matrices are obtained, transforming the originally discrete data change relationships into a structured matrix. Structured matrix data enables unified analysis of data changes across different time windows, obtaining synchronous change matrices corresponding to multiple consecutive time windows. Then, continuous statistics are performed on the same data combination location. For example, in five consecutive operating windows, the synchronous change values ​​between the braking system and the vibration signal are 0.60, 0.65, 0.72, 0.80, and 0.86, respectively. The synchronous change values ​​from multiple windows are accumulated to obtain a cumulative change value of 3.63. Subsequently, an average is calculated based on the number of windows, yielding an average cumulative change value of 0.726. This value reflects the degree of continuous synchronous change between the two data points during continuous operation. For example, if the cumulative change value of a certain data combination consistently exceeds 0...If the value is 70, it indicates that the data combination may exhibit a persistent abnormal correlation trend. By accumulating data across time windows, the impact of single abnormal fluctuations on the analysis results can be avoided, improving the ability to identify gradual hidden dangers. The cumulative change value between each data set is used as a basic indicator and adjusted according to the operational phase. For example, during the start-stop phase, due to the large variation in motor starting current, the cumulative change value between the control system data and motor operation data needs to be focused on. During the braking phase, because changes in braking force affect the vibration state, the cumulative change value between the braking system data and vibration signal data is a key analysis object. Furthermore, the cumulative change value is converted into a coupling strength value within the range of 0 to 1, where 0 indicates a weak correlation between the two data sets, and 1 indicates a highly synchronous change. For example, the coupling strength between the braking system data and vibration signal data is calculated to be 0.82, indicating a strong correlation, while the coupling strength between the gantry crane system data and vibration signal data is 0.30, indicating a low correlation. Arranging the coupling strengths corresponding to all data combinations forms a coupling strength matrix. This matrix clarifies the important correlation between different operational data, allowing for the acquisition of the corresponding short-cycle components. The coupling strength matrix, the coupling strength matrix corresponding to the medium-period component, and the coupling strength matrix corresponding to the long-period component were obtained. The coupling strength distribution of each data combination at different time scales was statistically analyzed. For example, the coupling strengths of braking system data and vibration signal data at short, medium, and long periods were 0.90, 0.82, and 0.75, respectively. The influence ratio at different time scales was determined based on the statistical results of historical hazard samples. For instance, a statistical analysis of 1000 elevator fault cases revealed that medium-period changes contributed the most to hazard assessment; therefore, the influence ratios for short, medium, and long periods were set at 30%, 40%, and 3%, respectively. The coupling strength at different time scales is fused according to this ratio. For example, the short-cycle coupling strength (0.90) multiplied by 30% yields 0.27, the medium-cycle coupling strength (0.82) multiplied by 40% yields 0.328, and the long-cycle coupling strength (0.75) multiplied by 30% yields 0.225. These three are added together to obtain the dynamic coupling strength of 0.823. The dynamic coupling strength of all data combinations is calculated in the same way, and the results are arranged to generate a dynamic coupling matrix. This dynamic coupling matrix can reflect the changes and correlations between various elevator operating data at different operating stages and time scales in real time.

[0029] In a preferred embodiment, an elevator operation status map is constructed based on multi-scale temporal features and a dynamic coupling matrix. The hazard propagation path and node influence weights are obtained based on the elevator operation status map. Dynamic risk propagation results are generated based on the hazard propagation path and node influence weights, including: S301. Based on the multi-scale temporal features and dynamic coupling matrix, obtain the state transmission values ​​and node response intensity between each data in the phased operation dataset, and generate an elevator operation status map based on the state transmission values ​​and node response intensity. S302. Based on the elevator operation status map, obtain the propagation path sequence and path transmission coefficient between each data in the phased operation data set, and obtain the hidden danger propagation path based on the propagation path sequence and path transmission coefficient. S303. Based on the elevator operation status map and the hazard propagation path, obtain the cumulative impact value of each data in the phased operation data in the propagation path, and obtain the node impact weight based on the cumulative impact value. S304. Based on the hazard propagation path and node influence weight, obtain the corresponding risk propagation intensity, and generate dynamic risk propagation results based on the risk propagation intensity.

[0030] As described in steps S301 to S304 above, multi-scale temporal features and a dynamic coupling matrix are obtained. The multi-scale temporal features are derived from short-cycle, medium-cycle, and long-cycle components obtained by time-series decomposition of various types of operational data in the phased operational dataset. The operational data includes control system data, car operation data, door operator system data, braking system data, and vibration signal data. For example, in one elevator operation cycle, the control system collects motor drive current, operating speed command, and feedback speed data; the car operation system collects car position, operating acceleration, and leveling deviation data; the door operator system collects door opening and closing current, door operator operating time, and door lock feedback status data; and the braking system collects... Braking force detection data, braking response time data, and vibration detection module data of car vibration amplitude and frequency are collected. After decomposition processing, each type of operating data forms corresponding multi-scale variation data. For example, braking system pressure data in the current braking stage yields short-cycle variation values ​​of 0.75, medium-cycle variation values ​​of 0.60, and long-cycle variation values ​​of 0.35; vibration signal data yields short-cycle variation values ​​of 0.80, medium-cycle variation values ​​of 0.65, and long-cycle variation values ​​of 0.45. Based on the multi-scale temporal characteristics and dynamic coupling matrix, the state transfer values ​​between different operating data are obtained, that is, the coupling strength value between any two operating data in the dynamic coupling matrix is ​​read. This coupling strength value originates from the aforementioned synchronous changes. The matrix and cumulative change values ​​across time windows are used. For example, a coupling strength of 0.85 between braking system data and vibration signal data indicates a high degree of synchronous change between the two during the current operating phase. The direction of state transfer is determined based on the temporal relationship between the changes in the two data points in the multi-scale time series characteristics. For instance, if a decrease in braking system pressure is detected followed by an increase in vibration signal amplitude within 0.3 seconds, then the state is determined to be transferred from the braking system data to the vibration signal data. Combining the coupling strength value and response time yields the state transfer value. For example, if the coupling strength is 0.85 and the time influence coefficient corresponding to the response time is 0.9, the combined state transfer value is approximately 0.76. This value describes the effect of one operating data change on another. The degree of influence of operational data is determined by obtaining the node response intensity based on the multi-scale variation amplitude of each data point in the phased operational dataset and the state transfer values ​​between it and other data. The degree of change of a particular operational data point after being affected by other data is statistically analyzed. For example, vibration signal data is affected by braking system data, car operation data, and door operator system data, with corresponding influence values ​​of 0.68, 0.45, and 0.32, respectively. The cumulative influence values ​​are accumulated to obtain the cumulative response value of the vibration signal data, which is 1.45. Normalization is then performed based on the cumulative response values ​​of all operational data nodes. For example, if the sum of the cumulative response values ​​of all data is 5.2, then the vibration signal data node response intensity is 1.45 divided by 5.2, resulting in approximately 0.28. After acquiring the state transmission values ​​and node response intensities, each operational data point in the phased operation dataset is used as a state node, the state transmission values ​​are used as connection parameters between nodes, and the node response intensities are used as node attribute information to construct an elevator operation state map. For example, the control system data node, car operation data node, door operator system data node, braking system data node, and vibration signal data node are connected. The connection parameter between the braking system data node and the vibration signal data node is 0.76, and the vibration signal data node response intensity is 0.28. This forms an elevator operation state map containing information on data nodes, state transmission direction, and degree of influence. The elevator operation status map allows elevator operation status to be described not by a single detection parameter, but by the dynamic transmission process between multiple operation data. This enables the identification of the process by which potential hazards spread from their origin to other operational stages, improving the accuracy of subsequent risk propagation analysis. It obtains node connection information from the elevator operation status map. For example, if a braking system data node connects to a vibration signal data node, and the vibration signal data node further connects to a car operation data node, it indicates that when an abnormal change occurs in the braking system, this change may further affect the car operation status through changes in the vibration signal. Based on the node connection direction, a propagation path sequence is generated, for example, obtaining… The propagation path sequence of "braking system data, vibration signal data, and car operation data" is obtained. The state transmission value corresponding to each stage of the transmission process in the path is read, and the overall propagation capability is calculated comprehensively. For example, if the state transmission value between braking system data and vibration signal data is 0.80, and the state transmission value between vibration signal data and car operation data is 0.70, then the path transmission coefficient of this propagation path is calculated to be approximately 0.56 through continuous transmission influence. Based on the connection relationships of all nodes in the elevator operation status diagram, all possible propagation paths are traversed and analyzed to obtain multiple candidate propagation paths. For example, the path transmission coefficient of "braking system data, vibration signal data, and car operation data" is obtained. The system uses multiple paths, including "door machine system data," "control system data," and "car operation data," and calculates the path transmission coefficient for each path. Candidate paths are then filtered based on preset path selection criteria. These criteria are derived from historical elevator fault data. For example, 1000 historical fault samples are collected, and the data propagation process before and after the fault occurs is analyzed. When the path transmission coefficient is greater than 0.70, the probability of that path appearing in historical faults reaches over 80%. Therefore, 0.70 is set as the high-impact path selection threshold. When the path transmission coefficient is 0.78, that path is identified as a potential hazard propagation path; when the path transmission coefficient is 0...If the value is 35, then the path is considered to have a weak impact and is not considered a major hazard propagation path. This approach allows for the determination of the anomaly's propagation direction from the perspective of state change propagation, rather than simply locating the current anomaly data, thus improving the depth and reliability of hazard analysis. The data node information contained in all hazard propagation paths is read, and the state transmission value of each node in different paths is obtained. For example, in three different hazard propagation paths, vibration signal data nodes appear in the paths "Brake System Data, Vibration Signal Data, Car Operation Data," "Brake System Data, Vibration Signal Data, Door Operator System Data," and "Control System Data, Vibration Signal Data, Car Operation Data," respectively, with corresponding path impact values ​​of 0.75, 0.68, and 0. 0.72. The cumulative influence values ​​of the same node across multiple propagation paths are accumulated to obtain the cumulative influence value of the node. For example, the cumulative influence value of the vibration signal data node is 0.75 + 0.68 + 0.72, which is 2.15. The cumulative influence values ​​of other nodes, such as the braking system data node, control system data node, and car operation data node, are calculated in the same way to obtain the sum of the cumulative influence values ​​of all nodes. For example, the braking system node is 1.80, the vibration signal node is 2.15, the car operation node is 1.05, and the control system node is 1.00, for a total of 6.00. The proportion of each node's cumulative influence value to the total influence value is then calculated. For example, the influence weight of the vibration signal node is 2.15 divided by 6.00, resulting in 0. The influence weight of the braking system node is 0.36, and the influence weight set of each operational data node is formed. This allows us to identify the operational link with the greatest impact during the current hazard propagation process. For example, if the vibration signal node has the highest weight, it indicates that the mechanical vibration change has a high degree of influence during the current anomaly propagation process and can be used as a key detection target to improve the accuracy of hazard location. The path transmission coefficient corresponding to each hazard propagation path is obtained, and the node influence weight of all nodes in that path is read. For example, if a hazard propagation path is "braking system data, vibration signal data, car operation data", the corresponding path transmission coefficient is 0.75, where the influence weight of the braking system node is 0.30, and the influence weight of the vibration signal node is... The weight is 0.45, and the influence weight of the car operation node is 0.25. The risk propagation intensity is obtained by comprehensively processing the path transmission coefficient and the node influence weight. For example, combining the path transmission coefficient of 0.75 with the total value of the path node influence weight of 1.0, the risk propagation intensity of the current propagation path is 0.75. The risk level classification range is set according to historical fault level data. This range is obtained by statistically analyzing historical fault samples. For example, 500 sets of elevator fault records are collected. Faults with a risk propagation intensity of less than 0.40 usually only affect a single component and are defined as low risk. When the risk propagation intensity is between 0.40 and 0.70, it usually affects more than two operating links and is defined as medium risk. The risk propagation intensity exceeds 0.At a risk level of 70, a multi-stage diffusion trend is typically observed, defined as high risk. Therefore, when the current risk propagation intensity is 0.75, it is classified as a high-risk level. Dynamic risk propagation results are generated based on the propagation path, affected nodes, and risk level corresponding to the risk propagation intensity. For example, the output might be "The braking system anomaly is the starting point, propagating to the car's operating state via vibration signals; the current propagation intensity is 0.75, and the risk level is high risk." This achieves the transformation from anomaly detection to risk propagation analysis, enabling not only the identification of the current anomaly location but also the prediction of its potential impact range.

[0031] In a preferred embodiment, theoretical and actual energy consumption data during elevator operation are acquired, an energy conservation deviation value is generated, and the hazard propagation path and node influence weights in the elevator operation status map are dynamically corrected to obtain a corrected causal propagation map. The dynamic risk propagation results are then updated, including: S401. Based on the control system data, car operation data, door operator system data, and braking system data during elevator operation, obtain the theoretical energy consumption data during elevator operation. S402. Based on control system data, car operation data, door operator system data, braking system data, and vibration signal data, obtain the actual energy consumption data during elevator operation; S403. Obtain the energy conservation deviation value based on the theoretical energy consumption data and the actual energy consumption data; S404. Based on the energy conservation deviation value, the hidden danger propagation path in the elevator operation status diagram is corrected to obtain the corrected hidden danger propagation path, and the node influence weight is corrected to obtain the corrected node influence weight. S405. Based on the corrected hazard propagation path and the corrected node influence weights, obtain the corrected causal propagation map; S406. Based on the corrected causal propagation map, update the dynamic risk propagation results to obtain the updated dynamic risk propagation results.

[0032] As described in steps S401 to S406 above, the control system data provides elevator operation command information (such as speed curves, acceleration commands, and floor information), the car operation data provides actual displacement and speed change information, the door operator system data provides door opening and closing action status information, and the braking system data provides braking intervention timing information. The theoretical energy consumption data is obtained by comprehensively calculating the potential energy change corresponding to the car's lifting height, the kinetic energy change corresponding to the operating speed, and the drive power requirement corresponding to the control commands. For example, during the elevator's journey from the 1st to the 10th floor, the ideal lifting energy is calculated based on the car's mass, floor height difference, and acceleration curve, and this is superimposed with the basic energy required for door operator opening and closing to form the theoretical energy consumption value. The basic energy consumption is obtained by integrating the motor current, voltage, and operating power data. Energy consumption is calculated by combining the real-time power output corresponding to changes in the car's running speed with the actual energy consumption. Mechanical friction, track irregularities, and structural impact losses reflected in the vibration signal are also introduced as additional energy terms to obtain actual energy consumption data. For example, in the case of slight wear on the guide rail, the vibration signal amplitude increases, corresponding to an increase in additional energy loss, thus making the actual energy consumption higher than the theoretical energy consumption. The difference or ratio between the theoretical and actual energy consumption data is calculated to obtain the energy conservation deviation value. For example, the absolute deviation value is obtained by subtracting the theoretical energy consumption from the actual energy consumption, or the relative deviation value is calculated by the ratio of the actual value to the theoretical value. For example, when the theoretical energy consumption is 100 units and the actual energy consumption is 120 units, the energy conservation deviation value is 20 or 1.The deviation ratio of 2 represents the degree of energy deviation caused by mechanical losses, system anomalies, or environmental factors during elevator operation. When the energy conservation deviation value is high, the state transmission intensity in the corresponding path is weakened or strengthened. For example, the reliable propagation weight of a high-deviation path is reduced, and the node influence weight is adjusted. For example, the weight of nodes with abnormal vibration or braking is increased. The correction method can be implemented by scaling or segmented mapping. For example, when the deviation value exceeds the threshold, the path weight is multiplied by the correction coefficient to obtain the corrected hidden danger propagation path and the corrected node influence weight, making the graph structure closer to the actual energy operation state. By remapping the corrected path structure and node weights into the graph structure, the corrected causal propagation graph is obtained. For example, the corrected path is used as the edge structure in the graph, and the node influence weight is used as the node attribute value. At the same time, the edge weights and node weights are updated to form a new causal propagation graph structure. This graph can reflect the system propagation logic after correction by energy conservation constraints. This system refines the risk propagation path and node contributions to better reflect actual operational conditions. Based on changes in node and path weights in the graph, the original risk propagation intensity is recalculated. For example, propagation paths significantly affected by the correction are reassigned risk propagation intensity. The risk results are then updated over time, allowing risk values ​​to dynamically change with the operational process. For instance, when a braking system anomaly causes an increase in energy deviation, the corresponding path's risk propagation intensity increases synchronously, generating updated dynamic risk propagation results to reflect real-time risk status changes. An additional energy loss term is introduced through vibration signals, allowing latent faults such as mechanical wear and impact anomalies to be reflected in the energy deviation, improving the sensitivity of hazard identification. Dynamic correction of hazard propagation paths is achieved through energy conservation deviation, automatically adjusting the weights of high-deviation paths to reduce the impact of false propagation paths and improve path reliability. Finally, a node influence weight correction mechanism dynamically strengthens the influence of key nodes such as vibration anomalies and braking anomalies in the graph, improving the accuracy of key node identification.

[0033] In a preferred embodiment, a hazard memory database is constructed to obtain historical hazard characteristics and generate corresponding hazard memory weights, including: S501. Obtain historical elevator operation data and obtain a historical hidden danger sample set based on the historical elevator operation data; S502. Obtain operational characteristic data corresponding to historical hazard events based on the historical hazard sample set, and obtain historical hazard characteristics, including historical propagation range, historical propagation depth and historical risk duration. S503. Obtain the hazard feature set based on historical hazard characteristics; S504. Construct a hazard memory bank based on the hazard characteristic set; S505. Obtain memory impact values ​​based on the characteristics of historical hazards; S506. Generate corresponding hidden danger memory weights based on memory impact values.

[0034] As described in steps S501 to S506 above, historical elevator operation data is acquired through the elevator operation monitoring system, remote data acquisition platform, and equipment maintenance management system. This historical elevator operation data includes control system data, car operation data, door operator system data, braking system data, vibration signal data, and historical fault record data generated by the elevator during its historical operating cycle. The control system data is acquired through the data recording interface inside the elevator main controller, including operating commands, operating speed feedback values, motor drive current, voltage change values, and control status indicators. The car operation data is acquired through devices installed on the car position detection device, speed detection device, and acceleration detection device, including car operating position, operating speed, and acceleration change. The system collects data on leveling and floor leveling errors, as well as door operator system data, through the door operator controller. This data includes door opening and closing time, door operator drive current, door lock feedback status, and door operator operation count. Braking system data is collected through the brake detection module, including braking force detection value, braking response time, and braking action status. Vibration signal data is collected by vibration sensors installed at the bottom of the car, near the guide rails, or at the traction machine location. This data includes vibration amplitude, vibration frequency, and vibration trend data. During implementation, a historical data collection cycle is set, for example, acquiring the target elevator's operating data for the past three years. A complete operating log is collected daily, and continuous operating data is collected for 30 minutes before and 30 minutes after an abnormal event occurs. If an elevator experiences a braking anomaly alarm on March 10th of a certain year, the system retrieves operational data from the historical database for the 30 minutes prior to that alarm and the 30 minutes after the fault is resolved. This data includes the braking pressure drop, changes in car deceleration, vibration amplitude changes, and control system alarm information. Based on fault identification information, maintenance records, and manual inspection results from the historical operational data, fault event correlation processing is performed. This involves matching operational data within the corresponding time range based on the fault occurrence time and combining all operational data corresponding to the same anomaly event to form a single historical potential hazard sample. For example, if the maintenance record shows an increased braking distance fault, then the braking system data and vibration information before and after the fault occurrence are used to identify the fault. The system data and car operation data are correlated to form a historical braking anomaly sample. When the maintenance record shows an abnormal door operator switch, the changes in door operator current, door lock status, and control system door opening and closing command data are combined to form a historical door operator anomaly sample. This forms a historical hazard sample set containing multiple historical hazard events. Each historical hazard sample includes the operating status before the anomaly occurred, the anomaly propagation process, and the final fault type information. The system reads the anomaly start time, anomaly end time, and corresponding data change process from each historical hazard sample, and then extracts the operating status change data at different time points before and after the anomaly occurred. For example, for a historical sample of a braking system anomaly, the braking response time is read to increase from 0.5 seconds to 1 second.Data such as the braking force detection value decreasing from 8 MPa to 6.8 MPa over 3 seconds, the vibration amplitude increasing from 0.12 g to 0.38 g, and the car deceleration changing from normal to abnormal values ​​are used as the operational characteristic data corresponding to this historical hidden danger event. The propagation range is determined based on the number of operational data categories involved in the abnormal data changes. For example, if a braking anomaly initially occurs in the braking system data, subsequently causing changes in the vibration signal, and further leading to a decrease in car operation stability, then this hidden danger involves three data categories: braking system, vibration signal, and car operation. Therefore, the propagation range of this historical hidden danger is determined to be three data nodes. The propagation range is determined based on the transmission of abnormal states between different operational data. The historical propagation depth is determined by hierarchical classification. For example, if a braking anomaly directly leads to increased vibration, the propagation depth is level one. If the braking anomaly leads to increased vibration, and the increased vibration further leads to deviation in car operation, the propagation process passes through two transmission levels, and the propagation depth is determined to be level two. The duration of historical risk is determined based on the time difference between the first occurrence of the anomaly and the time of recovery. For example, if a door operator anomaly is first detected at 9:20 AM and is restored at 11:50 AM after repair, the duration of this historical risk is 2.5 hours. Historical hazard characteristics such as propagation range, propagation depth, and risk duration are extracted for each historical hazard sample. For example, historical hazard sample H... The characteristics corresponding to historical hazard sample H001 are: propagation range (5 data nodes), propagation depth (level 3), and risk duration (6 hours). Historical hazard sample H002 has characteristics corresponding to: propagation range (2 data nodes), propagation depth (level 1), and risk duration (1 hour). Each historical hazard event is converted into structured historical hazard features. The historical hazard features corresponding to each historical hazard event in the historical hazard sample set are read, including propagation range, propagation depth, risk duration, and corresponding hazard type information. Then, the historical hazard features from different sources are formatted uniformly. For example, in some historical maintenance records, the propagation range may be recorded in text form as affecting the braking, vibration, and car modules. First, convert the data to a numerical form of 3. For example, if the risk duration in some records is recorded as 6 hours and 30 minutes, convert it to a time value of 6.5 hours. If the propagation depth is recorded as tertiary diffusion, convert it to a propagation depth value of 3. After standardizing the format, arrange the features corresponding to each historical hazard event according to a unified data structure, for example, forming the following hazard feature set: Hazard sample H001, corresponding to propagation range 5, propagation depth 3, risk duration 6 hours; Hazard sample H002, corresponding to propagation range 3, propagation depth 2, risk duration 3 hours; Hazard sample H003, corresponding to propagation range 1, propagation depth 1, risk duration 0.Within 5 hours, hazard information scattered across different historical records is converted into a structured feature set, enabling comparative analysis between different historical hazard events based on unified features. A data storage structure for the hazard memory bank is established, including at least the following fields: historical hazard number, hazard type, occurrence time, corresponding operational characteristics, propagation range, propagation depth, risk duration, and handling result. Then, each historical hazard feature in the feature set is written into the hazard memory bank according to its corresponding fields. For example, a historical hazard record is created, numbered H001, with hazard type "deterioration of braking performance," propagation range of 5, propagation depth of 3, risk duration of 6 hours, and maintenance measure "adjustment of braking mechanism." Another record, H002, with hazard type... The door operator is malfunctioning, with a propagation range of 2, a propagation depth of 1, and a risk duration of 1 hour. The maintenance measure is door operator current calibration. A mapping relationship is established between historical hazard events and hazard characteristics. This means that the corresponding operational characteristics and hazard evolution information can be quickly queried using historical hazard numbers. For example, when a current abnormal propagation path is detected to be similar to historical hazard H001, the propagation range, propagation depth, and risk duration information corresponding to H001 can be directly retrieved, forming an experience database of historical elevator operational hazards. When analyzing the current operating status, the development patterns of similar historical hazards can be referenced, improving the accuracy of hazard identification. The propagation range, propagation depth, and risk duration characteristics are read from each historical hazard record, and based on historical events... The importance of different characteristics in fault data is statistically analyzed. For example, analysis of 1000 historical fault samples reveals that hazards with greater propagation depth are more likely to cause multi-stage anomalies. Therefore, the propagation depth influence ratio is set at 40%. The propagation range reflects the number of affected devices, so the propagation range influence ratio is set at 30%. The risk duration reflects the degree of anomaly persistence, so the risk duration influence ratio is set at 30%. Historical hazard characteristics are standardized. For example, the propagation range standard range is set to 1 to 10 nodes; when a historical hazard's propagation range is 5, it is converted to 0.5. The propagation depth standard range is set to 1 to 5 levels; when the propagation depth is 3 levels, it is converted to 0.6. The risk duration standard range is set to 0 to 10 hours. When the duration is 6 hours, it is converted to 0.6. A comprehensive calculation is performed based on the standardization results and the corresponding impact ratio. For example, if a historical hazard H001 has three characteristics: propagation range 0.5, propagation depth 0.6, and duration 0.6, then the contribution value of propagation range is 0.15, the contribution value of propagation depth is 0.24, and the contribution value of duration is 0.18, resulting in a final memory impact value of 0.57. Each historical hazard receives a corresponding memory impact value, which reflects the reference value of historical hazard records in the current analysis process. The memory impact values ​​corresponding to all historical hazard records in the hazard memory bank are obtained; for example, the memory impact value for H001 is 0.57, the memory impact value for H002 is 0.42, and the memory impact value for H003 is 0.25. All memory impact values ​​are normalized so that the weight corresponding to each historical hazard record is within the range of 0 to 1. The memory impact value of a particular historical hazard is compared with the sum of the memory impact values ​​of all historical hazard records. For example, if the sum of the impact values ​​of three historical hazards is 1.24, then the memory weight of hazard H001 is 0.57 divided by 1.24, resulting in approximately 0.46; the weight of H002 is 0.34; and the weight of H003 is 0.20, forming a set of hazard memory weights, which is stored in the hazard memory database. During subsequent hazard path matching, when a current abnormal propagation path is detected to be similar to a historical hazard path, the importance of different historical hazard experiences is determined based on the hazard memory weights. For example, historical hazard H001 with a weight of 0.46 has a higher priority than historical hazard H003 with a weight of 0.20, thereby improving the accuracy of historical hazard matching results. This allows for dynamic adjustment of the influence degree of historical experiences based on the frequency, scope, and duration of historical hazard occurrences, achieving accurate hazard identification based on historical experience.

[0035] In a preferred embodiment, the hazard propagation paths in the corrected causal propagation map are filtered based on the hazard memory weight to obtain hazard matching paths, including: S601. Based on the hazard memory weights, obtain the weight distribution results corresponding to each historical hazard feature in the hazard memory database; S602. Based on the weight distribution results and the corrected causal propagation map, obtain multiple candidate hidden danger propagation paths in the corrected causal propagation map; S603. Based on multiple candidate hazard propagation paths and their corresponding hazard memory weights, obtain the path matching value for each candidate hazard propagation path; S604. Based on the path matching value, sort the multiple candidate hazard propagation paths to obtain the path matching sequence; S605. Based on the path matching sequence, filter candidate hazard propagation paths that meet the preset path conditions, and obtain hazard matching paths.

[0036] As described in steps S601 to S605 above, the historical hazard features and their corresponding hazard memory weights stored in the hazard memory bank are subjected to unified scale mapping and statistical grouping. For example, all historical hazard weight values ​​are normalized to map them to the 0-1 interval, and segmented statistically analyzed according to the weight interval to form a distribution structure within different weight intervals. Combined with the historical hazard feature categories (such as brake system anomalies, gantry crane system anomalies, or vibration anomalies), classification and summarization are performed to obtain the distribution of each category of hazards within different weight intervals. For example, if brake system hazards account for a higher proportion in the high weight interval, a skewed distribution result is formed. Paths are then determined based on the node connection structure and edge weight information in the corrected causal propagation graph. The process expands the path and, combined with the weight distribution results, constrains and filters the path expansion direction. For example, it prioritizes path branches related to node types corresponding to high-weight distributions, thereby generating multiple candidate hazard propagation paths. Starting with the elevator control system node, the path is expanded according to the order of edge weights from high to low in the graph. During the expansion process, weight distribution matching conditions are introduced, such as retaining only the path node connections that meet the weight distribution threshold conditions. This forms multiple sets of candidate hazard propagation paths, such as control system, car operation, vibration system or braking system, car operation, and door operator system. The path matching value is calculated based on the consistency of characteristics between the candidate hazard propagation path and historical hazard propagation paths, rather than being directly set. This involves acquiring path feature information corresponding to candidate hazard propagation paths, including path node order, propagation intensity, duration, and node influence weights. Historical path feature information corresponding to historical hazard propagation paths is retrieved from a hazard memory database. Each feature in the candidate hazard propagation path is compared with that in the historical hazard propagation paths to obtain the degree of matching in node order, propagation intensity, duration, and node influence. A comprehensive calculation is performed based on these matching results to obtain the path matching value. The differences in corresponding features between the candidate path and the historical path are then identified; for example, a node order consistency of 0.90, a propagation intensity consistency of 0.85, and a duration consistency of 0.85. The consistency level between nodes is 0.80, and the consistency level of node influence weight is 0.88. The overall calculation is based on the influence ratios obtained from historical hazard sample statistics. Node order, propagation intensity, duration, and node influence weight account for 40%, 30%, 15%, and 15% respectively. The final path matching value is obtained by multiplying each consistency level by its corresponding ratio and then summing the results. For example, the node order matching value (0.90) multiplied by 0.40 equals 0.36, the propagation intensity matching value (0.85) multiplied by 0.30 equals 0.255, the duration matching value (0.80) multiplied by 0.15 equals 0.12, and the node influence matching value (0.88) multiplied by 0.15 equals 0.132. The sum of these four results yields a path matching value of 0.867. The proportional parameter is derived from historical hazard sample statistics. For example, by collecting 1000 confirmed hazard events and statistically analyzing the correlation between different path characteristics and failure outcomes, it was found that the node propagation order has the greatest impact on hazard type judgment, so a higher proportion is set. Propagation intensity is the second most important factor, while duration and node influence weight are secondary judgment factors, and therefore lower proportions are set for each. The path matching values ​​corresponding to all candidate hazard propagation paths are numerically sorted, for example, arranged in descending order, thus forming a path priority sequence. The path matching values ​​of all candidate paths are used as the sorting criteria; when the path matching value is higher, the corresponding path has a higher priority. The ranking results are generated. For example, if path A has a matching value of 0.93, path B has 0.78, and path C has 0.61, the ranking results are A, B, and C. This ranking result is used as the path matching sequence. The preset path conditions are not fixed in advance, but are determined after statistical analysis of historical hazard propagation path data. That is, the propagation path data corresponding to historical elevator hazard events are obtained, including information such as historical propagation path length, number of path nodes, path propagation intensity, and path matching value. Statistical analysis is performed on confirmed historical hazard samples. For example, after obtaining 1000 sets of historical hazard samples, the propagation path characteristics corresponding to each set of hazard are statistically analyzed, and it is found that more than 90% of them... The effective hazard propagation path contains 2 to 5 consecutive nodes, and the path matching value is mainly distributed in the range of 0.75 to 0.95. Statistical analysis of path propagation intensity shows that the propagation intensity of the real hazard path is usually greater than 0.60, while the propagation intensity of the ordinary operational fluctuation path is usually lower than 0.50. Therefore, based on the statistical results, preset path conditions are determined, such as setting the number of path nodes to the range of 2 to 5 nodes, setting the path matching value threshold to 0.75, and setting the path propagation intensity threshold to 0.60. Candidate hazard propagation paths in the corrected causal propagation map are obtained, and it is determined whether each candidate path meets the preset path conditions. For example, when a candidate hazard propagation path contains... When a candidate path contains 3 nodes, has a path matching value of 0.86, and a path propagation strength of 0.72, it is considered to meet the preset path conditions because its number of nodes is between 2 and 5, and its path matching value is greater than 0.75 and its propagation strength is greater than 0.60. Therefore, it is selected as a hazard matching path. If a candidate path contains only 1 node or its path matching value is less than 0.75, it is deemed not to meet the screening conditions and is removed. This filters out structurally incomplete paths, ultimately obtaining hazard propagation paths that meet the conditions as hazard matching paths. Through a multi-path candidate generation and matching ranking mechanism, multiple propagation possibilities can be considered simultaneously, improving the identification coverage capability under complex working conditions.

[0037] In a preferred embodiment, the hazard matching path is matched with a hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are overlaid and corrected. Based on the corrected dynamic risk propagation results, the hazard location result, risk level, and early warning information are output, including: S701. Based on the hazard matching path, obtain the corresponding path feature sequence, wherein the path feature sequence includes node order information, path propagation intensity information and path duration information. S702. Obtain the path similarity value based on the path feature sequence and historical hidden danger features; S703. Compare the path similarity value with a preset threshold. When the path similarity value exceeds the preset threshold, the hazard correction condition is determined to be met. S704. Based on the hazard correction conditions, the updated dynamic risk propagation results are corrected to obtain the corrected dynamic risk propagation results. S705. Based on the revised dynamic risk propagation results, obtain the corresponding data on the location of the hidden danger and the intensity of its propagation impact; S706. Based on the location data of the hazard and the intensity data of its propagation and impact, generate the hazard location results, risk level and early warning information.

[0038] As described in steps S701 to S706 above, the node order, path propagation intensity, and path duration in the hazard matching path are extracted and encoded in a structured manner. Specifically, the hazard matching path is traversed sequentially to record the order in which nodes are visited, thus forming node order information. The overall path propagation intensity is calculated based on the state transmission values ​​corresponding to each edge in the path and the node influence weights, for example, through weighted multiplication or weighted summation, thus forming path propagation intensity information. Path duration information is extracted based on the path's continuous existence time or abnormal duration window, for example, by counting the length of time the path is continuously identified as a high-risk path in multiple consecutive time windows, thus forming a complete path feature sequence. For example, for path control systems, car operation, and vibration systems, the node order is the corresponding node sequence, the propagation intensity is calculated by comprehensively calculating the path edge weights, and the duration is an abnormal duration of 120 seconds. The current path feature sequence is then matched with historical hazard features stored in the hazard memory bank in multiple dimensions, including node order consistency comparison, path propagation intensity difference calculation, and path duration deviation calculation. The results of each dimension are then weighted and fused. The path similarity value is obtained. For example, if the node order is completely identical and the propagation intensity is close to the historical braking anomaly path, and the duration difference is small, the path similarity value is high. Conversely, if the node order or propagation intensity differs greatly, the similarity value decreases. This achieves quantitative matching between the current hazard propagation path and historical hazard patterns. The preset threshold is obtained statistically based on historical elevator operation data and corresponding hazard records, rather than being manually set. That is, historical hazard samples within a preset time range are obtained, the path feature sequences corresponding to each historical hazard are extracted, and the path features and historical hazard features are calculated. Path similarity (i.e., the path similarity value between two samples calculated using the cosine similarity formula) is used to statistically analyze the similarity values ​​of multiple historical samples. A threshold value that can distinguish between real hidden dangers and fluctuations in normal operation is selected as the preset threshold. For example, after analyzing 1000 sets of historical hidden danger samples, it was found that the similarity values ​​of real hidden danger samples are mainly concentrated above 0.80, while the similarity values ​​of normal operation are below 0.70. Therefore, 0.80 is used as the preset threshold. During the comparison process, the current path similarity value is compared with the threshold. For example, when the path similarity value is 0.87 and the preset threshold is 0.At 8:00, the current path is determined to be highly consistent with historical hazard patterns, thus triggering the hazard correction mechanism and entering the risk correction and result update process. Based on the original dynamic risk propagation results, a historical hazard superposition factor is introduced. For example, when the hazard correction conditions are met, the risk propagation intensity in the corresponding hazard propagation path is enhanced, or the influence weight of relevant nodes is re-weighted, so that recurring hazard paths occupy a higher weight in risk assessment. For instance, when a brake system-related hazard is identified again, the risk propagation intensity of its corresponding path is proportionally increased, reflecting the cumulative impact of historical hazard repetition on the current risk, thus affecting the corrected risk propagation structure. A reverse backtracking process is performed to determine the location of the hazard by identifying the node with the highest contribution, such as locating it to key nodes like the braking system or door operator system. Simultaneously, the propagation impact intensity is calculated based on the cumulative contribution values ​​of each node and edge weight in the propagation process. For example, the cumulative weight contribution of a node in multiple high-risk paths is statistically analyzed to obtain propagation impact intensity data, which characterizes the diffusion intensity and impact range of the hazard in the system. The risk propagation paths in the corrected risk propagation results are then traced backward to identify the operating node with the greatest impact during risk propagation, and the elevator component location corresponding to that node is taken as the location of the hazard. For example, in the elevator operation status analysis process, the following is obtained... If a risk propagation path is found to be "braking system, car operation system, abnormal vibration signal", and the braking system node has the highest cumulative impact value, then the location data corresponding to the braking system is used as the location data of the potential hazard. This location data comes from the elevator equipment structure information table, which is pre-stored in the system database. The table includes elevator component names, component numbers, installation locations, and systems they belong to, for example, "Brake A01, located on the traction machine side of the machine room, belonging to the braking system". The impact degree of each risk node in the current propagation path is calculated, and the node impact weight, path propagation intensity, and risk duration information are obtained. The node impact weight is derived from the elevator operation system... The path propagation intensity is derived from the path transmission coefficient in the hazard propagation path, and the risk duration is derived from the continuous anomaly detection time window. For example, in a certain operating cycle, the influence weight of the braking system node is 0.45, the influence weight of the car operation node is 0.30, and the influence weight of the vibration node is 0.25. The three together form the current risk propagation influence range. Combined with the path duration of 120 seconds, the result of the hazard propagation influence intensity is obtained. In order to avoid the risk judgment bias caused by a single factor, multiple influence data are comprehensively processed. For example, the node influence weight, path propagation intensity, and duration are set to 0.5, 0.3, and 0.The ratio of 2 is comprehensively calculated, with the ratio derived from the statistical results of historical hazard samples. For example, by collecting data on 1000 historical elevator malfunctions and statistically analyzing the correspondence between different influencing factors and the final failure results, the highest proportion of nodal influencing factors is found, thus assigning a higher weight to them. This yields specific data on the intensity of the propagation impact. For instance, an impact intensity of 0.86 indicates a high degree of current hazard propagation. Risk level classification rules are pre-established, determined based on historical failure cases and safety management requirements. For example, the propagation impact intensity is divided into three intervals: an impact intensity less than 0.40 is classified as low risk; an impact intensity greater than or equal to 0.40 and less than 0.75 is classified as medium risk; and an impact intensity greater than or equal to 0.75 is classified as high risk. The interval range can be obtained through statistical analysis of historical hazard data. For example, by collecting elevator malfunction records from the past three years, low-impact malfunctions account for approximately 60%, with their propagation impact intensity concentrated in the range of 0 to 0.40; general malfunctions account for approximately 30%, with their impact intensity concentrated in the range of 0.40 to 0.75; and serious malfunctions account for approximately 10%, with their impact intensity concentrated in the range of 0 to 0.75. Since the risk level often exceeds 0.75, a corresponding risk level classification standard is established. When the current propagation impact intensity is 0.86, the risk level of the current hidden danger is determined to be high according to the risk level rules. A corresponding information template is matched according to preset warning rules. For example, when the hidden danger location is in the braking system and the risk level is high, the high-risk warning template is invoked to generate information including the hidden danger location, anomaly type, risk level, and handling suggestions. For example, "An abnormal risk has been detected in the braking system, with a high risk level. It is recommended to immediately check the wear condition of the brakes and the braking force output." The warning rules are derived from equipment maintenance standards and historical fault handling records. For example, the maintenance database records that the corresponding inspection items for braking system anomalies include brake clearance, braking torque, and mechanical wear condition. The system automatically associates the corresponding inspection information based on the hidden danger location results, achieving a precise mapping from the risk propagation result to the specific equipment location. Compared to the traditional method of only outputting abnormal alarm information, this method can further clarify the source and scope of the hidden danger, enabling maintenance personnel to quickly locate abnormal components and take targeted measures, reducing the scope of manual investigation, lowering the probability of elevator malfunctions, and improving elevator operation safety and maintenance efficiency.

[0039] Based on the location data of the potential hazard, the specific location of the faulty subsystem or equipment is determined, and the risk level is classified according to the intensity of the propagation impact. For example, the intensity of the propagation impact is divided into three levels: low, medium, and high, thus forming a risk level output. Early warning rules are triggered according to the risk level and the intensity of the propagation impact. For example, when the risk level reaches high and the intensity of the propagation impact exceeds a preset threshold, an early warning message is generated, thus outputting a result such as the braking system abnormality risk level being high and triggering an early warning, achieving a unified output of hazard location and risk warning.

[0040] Please see the appendix Figure 2 As shown, the present invention also provides an elevator operation hazard identification system integrating big data analysis, used in the aforementioned elevator operation hazard identification method integrating big data analysis, comprising: The multi-source data module is used to acquire multi-source operation data of the target elevator within a preset operating cycle, and to acquire a phased operation dataset based on the multi-source operation data; The matrix construction module acquires multi-scale temporal features based on the phased execution of the dataset and constructs a dynamic coupling matrix; The propagation results module constructs an elevator operation status map based on multi-scale temporal features and a dynamic coupling matrix, obtains the hazard propagation path and node influence weight based on the elevator operation status map, and generates dynamic risk propagation results based on the hazard propagation path and node influence weight. The risk update module is used to obtain theoretical and actual energy consumption data during elevator operation, generate energy conservation deviation values, dynamically correct the hidden danger propagation path and node influence weight in the elevator operation status map, obtain the corrected causal propagation map, and update the dynamic risk propagation results. The hazard weighting module is used to build a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights. The hazard path module filters the hazard propagation paths in the corrected causal propagation map based on the hazard memory weight, and obtains the hazard matching path; The hazard results module is used to match the hazard matching path with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are superimposed and corrected, and the hazard location results, risk level and early warning information are output based on the corrected dynamic risk propagation results.

[0041] As described above, the multi-source data module synchronously collects operational data from the elevator control system, car operation system, door operator system, braking system, and vibration signal acquisition unit. The collected data is then uniformly timestamped and stored. The preset operating cycle is determined by the elevator operation scheduling cycle or a set statistical time interval, such as a 24-hour period or 1000 consecutive start-stop cycles. By aligning the multi-source operational data with time and identifying the operating status, the elevator operation process is divided into start-stop, constant-speed operation, and braking phases, thus forming a structured, phased operational dataset. The matrix construction module decomposes the phased operational data into multiple time scales, extracting short-cycle, medium-cycle, and long-cycle variation characteristics, and storing them within a unified time window. The system calculates the synchronization intensity and response change of each data source to form a coupling strength expression for cross-system interaction. The coupling strength under different time windows is accumulated and normalized to form a coupling strength matrix. Further, a distribution mapping is performed under different operating stages to generate a time-varying dynamic coupling matrix. This matrix characterizes the dynamic interaction intensity between elevator subsystems under different operating states. The propagation result module uses each subsystem as a graph node and the interaction intensity in the coupling matrix as edge weights to construct an elevator operating state graph. Potential hazard propagation paths are extracted through path traversal and weight selection. Node influence weights are calculated based on the contribution values ​​of nodes in the path, and the propagation intensity and node influence weights are jointly calculated. The risk propagation intensity is calculated to generate dynamic risk propagation results, achieving a structured expression of the elevator hazard propagation process. The risk update module calculates theoretical energy consumption based on control system data, car operation data, door operator system data, and braking system data, and calculates actual energy consumption by combining vibration signal data. It then calculates the energy conservation deviation between the two, mapping this deviation to path correction factors and node correction factors. This adjusts the edge weights and node weights in the elevator operation state graph, generating a corrected causal propagation graph. Based on this graph, the risk propagation results are updated to ensure the risk assessment results match the actual energy loss state. The hazard weighting module extracts historical hazard samples from historical elevator operation data and fault records, and then... The system extracts and structures features from the propagation range, depth, and duration of risks to form a set of historical hazard features. Different historical hazard features are weighted to generate hazard memory weights, which are then stored in a hazard memory database, forming a historical hazard knowledge structure. The hazard path module constrains and filters candidate propagation paths in the graph based on the hazard memory weights, and performs matching scoring based on path structure features. This selects a set of propagation paths highly consistent with historical hazard features as the hazard matching path output. The hazard result module calculates the similarity between the hazard matching path and the historical hazard features. When the similarity exceeds a preset threshold, a hazard correction mechanism is triggered, and the risk propagation results are enhanced and corrected accordingly.Subsequently, based on the corrected risk propagation results, high-contribution nodes are extracted as hazard locations, and risk levels are classified according to propagation intensity. Finally, corresponding early warning information is generated, achieving a closed-loop output of hazard identification and risk warning. By introducing multi-scale temporal features and a dynamic coupling matrix, the elevator's operating state can be accurately characterized at different time scales, improving the completeness of state representation. By introducing energy conservation bias to correct the propagation path and node weights, the risk assessment results can reflect the actual operating energy state, improving accuracy. Through a similarity threshold triggering mechanism, duplicate or similar hazards can be automatically identified and reactivated, improving the sensitivity of hazard identification. Through propagation path and node location output, the joint determination of hazard location and risk level is achieved, improving the accuracy of early warning.

[0042] And, a terminal for accurately identifying elevator operation hazards by integrating big data analysis, including: One or more processors; A storage device on which one or more programs are stored; When one or more programs are executed by one or more processors, the one or more processors can implement a method for accurately identifying potential elevator operation hazards by integrating big data analysis.

[0043] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for accurately identifying potential elevator malfunctions by integrating big data analytics, characterized in that, include: Acquire multi-source operation data of the target elevator within a preset operating cycle, and obtain a phased operation dataset based on the multi-source operation data; Multi-scale temporal features are obtained by running the dataset in stages, and a dynamic coupling matrix is ​​constructed. An elevator operation status map is constructed based on multi-scale temporal features and dynamic coupling matrix. The hazard propagation path and node influence weight are obtained based on the elevator operation status map. Dynamic risk propagation results are generated based on the hazard propagation path and node influence weight. The theoretical and actual energy consumption data during elevator operation are obtained, an energy conservation deviation value is generated, the hidden danger propagation path and node influence weight in the elevator operation status map are dynamically corrected, the corrected causal propagation map is obtained, and the dynamic risk propagation results are updated. Construct a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights; Based on the hazard memory weight, the hazard propagation path in the corrected causal propagation map is filtered to obtain the hazard matching path; The hazard matching path is matched with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are superimposed and corrected. Based on the corrected dynamic risk propagation results, the hazard location results, risk level and early warning information are output.

2. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, Acquire multi-source operational data of the target elevator within a preset operating cycle, and obtain phased operational datasets based on the multi-source operational data, including: Acquire multi-source operational data of the target elevator within a preset operating cycle; Multi-source operational data undergoes data preprocessing, which includes outlier removal, missing data completion, and data format standardization to form multi-source operational data. Multiple timestamps are obtained from multi-source runtime data, and time alignment is performed to construct a time runtime data sequence. Based on the time-based operation data sequence, the elevator operation status change feature points are identified, and the elevator operation process is divided into operation stages according to the elevator operation status change feature points, resulting in the start-stop stage, constant speed operation stage and braking stage, and the staged operation dataset is generated.

3. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 2, characterized in that, Multi-source operating data includes control system information, car operation information, door operator system information, braking system information, and vibration signal information.

4. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, Multi-scale temporal features are obtained based on a phased dataset run, and a dynamic coupling matrix is ​​constructed, including: The time series decomposition process will be performed on each data point in the phased running dataset to obtain short-period components, medium-period components and long-period components. Based on the short-cycle component, medium-cycle component and long-cycle component, obtain the response change of each data in the phased running dataset within the same time window; Based on the response change, obtain the synchronous change value of each data point in the phased running dataset within the same time window; Obtain the synchronization change matrix corresponding to each time window based on the synchronization change values; Obtain the cumulative change value across the time window based on the synchronous change matrix; The coupling strength matrix is ​​obtained based on the cumulative change values ​​across the time window; Based on the coupling strength matrix, obtain the numerical distribution results of each data point in the phased running dataset at different time scales, and generate a dynamic coupling matrix based on the numerical distribution results.

5. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, An elevator operation status map is constructed based on multi-scale temporal features and a dynamic coupling matrix. The hazard propagation path and node influence weights are obtained from this map. Dynamic risk propagation results are then generated based on these paths and weights, including: Based on multi-scale temporal features and dynamic coupling matrix, the state transmission values ​​and node response strengths between each data point in the phased operation dataset are obtained, and an elevator operation status map is generated based on the state transmission values ​​and node response strengths. Based on the elevator operation status map, the propagation path sequence and path transmission coefficient between each data point in the phased operation data are obtained, and the hazard propagation path is obtained based on the propagation path sequence and path transmission coefficient. Based on the elevator operation status map and the hazard propagation path, the cumulative impact value of each data in the phased operation dataset in the propagation path is obtained, and the node impact weight is obtained based on the cumulative impact value. Based on the propagation path of the hidden danger and the influence weight of the nodes, the corresponding risk propagation intensity is obtained, and dynamic risk propagation results are generated based on the risk propagation intensity.

6. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, The theoretical and actual energy consumption data during elevator operation are obtained, an energy conservation deviation value is generated, and the hazard propagation paths and node influence weights in the elevator operation status map are dynamically corrected to obtain the corrected causal propagation map. The dynamic risk propagation results are then updated, including: Based on the control system data, car operation data, door operator system data, and braking system data during elevator operation, theoretical energy consumption data during elevator operation is obtained. Based on control system data, car operation data, door operator system data, braking system data, and vibration signal data, the actual energy consumption data during elevator operation is obtained. The energy conservation deviation value is obtained by comparing theoretical energy consumption data with actual energy consumption data. Based on the energy conservation deviation value, the hidden danger propagation path in the elevator operation status map is corrected to obtain the corrected hidden danger propagation path, and the node influence weight is corrected to obtain the corrected node influence weight. Based on the corrected hazard propagation path and the corrected node influence weights, obtain the corrected causal propagation map; Based on the corrected causal propagation map, the dynamic risk propagation results are updated to obtain the updated dynamic risk propagation results.

7. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, Construct a hazard memory database, acquire historical hazard characteristics, and generate corresponding hazard memory weights, including: Obtain historical elevator operation data, and then obtain a historical potential hazard sample set based on the historical elevator operation data; Based on the historical hazard sample set, operational characteristic data corresponding to historical hazard events are obtained, and historical hazard characteristics are obtained. These characteristics include the historical spread range, historical spread depth, and historical risk duration. Obtain a set of hazard characteristics based on historical hazard features; Construct a hazard memory bank based on the hazard feature set; Obtain memory impact values ​​based on historical hazard characteristics; Based on the memory impact value, a corresponding hidden danger memory weight is generated.

8. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, Based on the hazard memory weight, the hazard propagation paths in the corrected causal propagation map are filtered to obtain hazard matching paths, including: Based on the hazard memory weights, obtain the weight distribution results corresponding to each historical hazard feature in the hazard memory database; Based on the weight distribution results and the corrected causal propagation map, multiple candidate hazard propagation paths in the corrected causal propagation map are obtained; Based on multiple candidate hazard propagation paths and their corresponding hazard memory weights, obtain the path matching value for each candidate hazard propagation path; Based on the path matching value, multiple candidate hazard propagation paths are sorted to obtain a path matching sequence; Based on the path matching sequence, candidate hazard propagation paths that meet the preset path conditions are filtered to obtain hazard matching paths.

9. The method for accurate identification of elevator operation hazards by integrating big data analysis according to claim 1, characterized in that, The hazard matching path is matched with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are overlaid and corrected. Based on the corrected dynamic risk propagation results, the hazard location results, risk level, and early warning information are output, including: Based on the hazard matching path, the corresponding path feature sequence is obtained, which includes node order information, path propagation intensity information, and path duration information. Path similarity values ​​are obtained based on path feature sequences and historical hazard features; The path similarity value is compared with a preset threshold. When the path similarity value exceeds the preset threshold, the hazard correction condition is determined to be met. Based on the hazard correction conditions, the updated dynamic risk propagation results are corrected to obtain the corrected dynamic risk propagation results. Based on the revised dynamic risk propagation results, obtain the corresponding data on the location of potential hazards and the intensity of their propagation impact; Based on the location data of the potential hazard and the intensity data of its propagation and impact, the system generates hazard location results, risk levels, and early warning information.

10. A system for accurately identifying elevator operation hazards by integrating big data analysis, applied to the method for accurately identifying elevator operation hazards by integrating big data analysis as described in any one of claims 1 to 9, characterized in that, include: The multi-source data module is used to acquire multi-source operation data of the target elevator within a preset operating cycle, and to acquire a phased operation dataset based on the multi-source operation data; The matrix construction module acquires multi-scale temporal features based on the phased execution of the dataset and constructs a dynamic coupling matrix; The propagation results module constructs an elevator operation status map based on multi-scale temporal features and a dynamic coupling matrix, obtains the hazard propagation path and node influence weight based on the elevator operation status map, and generates dynamic risk propagation results based on the hazard propagation path and node influence weight. The risk update module is used to acquire theoretical and actual energy consumption data during elevator operation, generate energy conservation deviation values, dynamically correct the hidden danger propagation paths and node influence weights in the elevator operation status map, obtain the corrected causal propagation map, and update the dynamic risk propagation results. The hazard weighting module is used to build a hazard memory database, obtain historical hazard characteristics, and generate corresponding hazard memory weights. The hazard path module filters the hazard propagation paths in the corrected causal propagation map based on the hazard memory weight, and obtains the hazard matching path; The hazard results module is used to match the hazard matching path with the hazard memory database. When the matching similarity exceeds a preset threshold, the updated dynamic risk propagation results are superimposed and corrected, and the hazard location results, risk level and early warning information are output based on the corrected dynamic risk propagation results.

Images