Elevator fault real-time monitoring and early warning method, system and terminal

Abstract

The invention relates to an elevator fault real-time monitoring and early warning method, system and terminal, and belongs to the technical field of elevator monitoring. The monitoring and early warning method comprises the steps that multi-source sensing data of an elevator are collected, and timestamp synchronization is conducted on the multi-source sensing data; calculating a comprehensive risk value of the elevator based on the multi-source sensing data after timestamp synchronization; judging whether the comprehensive risk value is greater than a risk threshold value or not; if yes, Granger causal test is operated, an abnormal causal event is searched for from a pre-constructed causal graph, and relevant data corresponding to the abnormal causal event is packaged and sent to the elevator digital twinborn body; fault acceleration simulation is carried out in the digital twinborn body, and an abnormal root is positioned; and a health assessment report of the elevator is generated according to the simulation process data and the abnormal root. The method has the beneficial effects that the actual running state of the elevator is truly reflected, so that the running safety and reliability of the elevator are improved.

Description

Technical Field

[0001] This invention relates to the technical field of elevator monitoring, and in particular to a method, system and terminal for real-time monitoring and early warning of elevator faults. Background Technology

[0002] With the acceleration of urbanization and the increasing number of high-rise buildings, elevators, as an indispensable vertical transportation tool, are seeing their usage frequency and coverage expand continuously. By 2020, my country had over 8 million elevators, and this number continues to grow. However, elevator safety accidents still occur frequently, making elevator safety and reliability a focus of widespread public concern. Residents have raised higher requirements for elevator safety, comfort, and safety supervision. Therefore, effectively monitoring elevator malfunctions, providing timely warnings, and ensuring the safety of passengers have become urgent problems to be solved.

[0003] Currently, various elevator fault monitoring and early warning technologies have emerged in the market to address elevator safety issues. Some traditional methods employ external sensor alarms, installing sensors in key parts of the elevator to acquire operational status information such as speed and displacement. When the data detected by the sensors exceeds a preset normal range, an alarm is triggered. Other systems rely on simple statistical analysis of historical elevator operation records and fault data, setting fixed rules and thresholds to determine whether an elevator malfunction is likely.

[0004] Traditional sensor-based elevator fault monitoring methods suffer from a high false alarm rate. This is because sensors often acquire unofficial "second-hand information," which is easily affected by various factors such as external environmental influences and the stability of the equipment itself. This leads to inaccurate detection information that fails to accurately reflect the actual operating status of the elevator. For example, slight vibrations or disturbances may cause sensors to misinterpret them as faults, issuing unnecessary alarms. This not only increases the workload of maintenance personnel but also reduces the reliability and trustworthiness of the system. Summary of the Invention

[0005] In order to accurately reflect the actual operating status of elevators and improve the safety and reliability of elevator operation, this invention provides a method, system and terminal for real-time monitoring and early warning of elevator faults.

[0006] In a first aspect, the present invention provides a method for real-time monitoring and early warning of elevator malfunctions, employing the following technical solution: A method for real-time monitoring and early warning of elevator malfunctions includes: Collect multi-source sensor data of the elevator and synchronize the multi-source sensor data with timestamps; The comprehensive risk value of the elevator is calculated based on the multi-source sensor data after timestamp synchronization. Determine whether the overall risk value is greater than the risk threshold; If so, then run the Granger causality test, find the abnormal causal events from the pre-built causal graph, and package the relevant data corresponding to the abnormal causal events and send them to the elevator digital twin; Accelerated fault simulation is performed within the digital twin to pinpoint the root cause of the anomaly; Based on the simulation process data and the root causes of the anomalies, a health assessment report for the elevator is generated.

[0007] By employing the aforementioned technical solution, the comprehensive risk value of the elevator is calculated based on synchronized data, enabling a quantitative assessment of the overall elevator operating status and providing a clear understanding of the elevator's risk level. By determining the relationship between the comprehensive risk value and the risk threshold, potential anomalies during elevator operation can be detected promptly. When the comprehensive risk value exceeds the risk threshold, a Granger causality test is performed, and abnormal causal events are identified from the causal graph. This allows for in-depth analysis of potential causal relationships leading to elevator anomalies, providing clues for accurately locating the root cause of the fault. The relevant data corresponding to the abnormal causal events are packaged and sent to the elevator's digital twin. Leveraging the highly realistic nature of the digital twin, fault simulation is accelerated in a virtual environment, quickly locating the root cause of the anomaly and significantly shortening troubleshooting time. Finally, a health assessment report for the elevator is generated based on the simulation data and the root cause of the anomaly. This provides elevator maintenance personnel with comprehensive and detailed information on the elevator's operating status, facilitating the development of scientific and reasonable maintenance plans. This improves the safety and reliability of elevator operation, reduces the probability of malfunctions, and minimizes safety hazards and economic losses caused by elevator failures.

[0008] Optionally, the steps for calculating the overall risk value of the elevator include: Feature extraction was performed on the synchronized multi-source sensor data to obtain several key feature indicators related to the elevator's operating status. For each key characteristic indicator, calculate the individual risk value of that key characteristic indicator based on its historical normal operation data and the preset safety threshold range; The elevator's overall risk value is obtained by weighting and summing all individual risk values ​​using a pre-set weighted fusion algorithm. The weight of each individual risk value is dynamically adjusted according to the importance of different key characteristic indicators for the safe operation of the elevator.

[0009] By adopting the above technical solution, multi-source heterogeneous raw sensor data can be transformed into comprehensive quantitative indicators that can intuitively reflect the overall risk level of the elevator. Through step-by-step feature extraction, individual risk assessment, and weighted fusion, not only are the independent risk contributions of each key component and operating parameter considered, but also their correlation and comprehensive impact on overall safety are taken into account. This avoids the possibility of misjudgment or omission of a single indicator, providing a reliable quantitative basis for subsequent anomaly detection and early warning, and ensuring the objectivity and accuracy of elevator risk assessment.

[0010] Optionally, the step of performing accelerated fault simulation and locating the root cause of the anomaly in the digital twin includes: Data related to abnormal causal events are injected into the digital twin as initial boundary conditions; An accelerated simulation algorithm is used to reproduce the fault evolution process in the digital twin, simulate the response of different components under abnormal excitation, and record the time-series curves of the state parameters of each component. By comparing cause-effect graphs with simulation data and combining them with graph neural networks, fault propagation paths can be identified. Calculate the contribution of each potential fault source in the fault propagation path, and determine the fault source with a contribution exceeding a preset threshold as the root cause of the anomaly. Perform parameter sensitivity analysis on the root cause of the anomaly to verify whether it is the only or main cause of the fault, eliminate interfering factors with weak correlation, and finally output a clear result of the anomaly root cause location.

[0011] By adopting the above technical solution, firstly, by injecting data related to abnormal causal events as initial boundary conditions into the digital twin, the simulated environment can closely resemble actual elevator fault scenarios. Next, an accelerated simulation algorithm is used to reproduce the fault evolution process, simulating the responses of different components under abnormal excitation and recording the time-series curves of state parameters, significantly shortening fault investigation time. Then, by comparing the causal graph with the simulation data and combining it with a graph neural network to identify the fault propagation path, the spread of the fault in the elevator system can be clearly displayed, clarifying the specific path from the fault's origin to its impact on other components. This helps to comprehensively understand the scope and mechanism of the fault's influence, avoiding focusing only on surface faults while ignoring potential deeper problems. Based on this, the contribution of each potential fault source in the fault propagation path is calculated, and fault sources with contribution exceeding a preset threshold are identified as abnormal root causes. This accurately identifies the key factors leading to the fault, avoiding blind investigation and improving the accuracy and efficiency of fault location. Finally, parameter sensitivity analysis was performed on the anomaly root cause to verify whether it was the sole or primary cause of the fault, eliminate weakly correlated interfering factors, and ultimately output a clear anomaly root cause location result. This further ensured the accuracy and reliability of fault location, avoided misjudging some accidental or secondary factors as the root cause of the fault, and provided a scientific and accurate basis for elevator maintenance and repair, thereby effectively improving the safety and reliability of elevators and reducing the probability of fault occurrence and maintenance costs.

[0012] Optionally, the steps for generating a health assessment report include: Based on the comparison between the simulation process data and the normal operation benchmark model, a multi-index fusion evaluation algorithm is adopted to quantify the real-time health status of each key subsystem of the elevator. The health status includes health index and status level. By combining historical operating data, current health status, and simulated failure evolution trends, a time series prediction model is used to predict the remaining useful life (RUL) of key vulnerable components and the probability of specific types of failures occurring in the short term. By integrating the results of anomaly root cause location, health index, remaining service life and short-term failure probability, combined with elevator maintenance procedures and expert knowledge base, specific maintenance or repair suggestions are generated, and maintenance priorities are marked according to the severity of the failure and the urgency of its occurrence. Fill the predefined health assessment report template with the health status assessment results, prediction results, maintenance or repair recommendations and priorities to form a complete health assessment report.

[0013] By adopting the above technical solution, firstly, based on the comparison between simulated process data and normal operation benchmark models, a multi-index fusion evaluation algorithm is used to quantify the real-time health status of each key subsystem of the elevator, deriving a health index and status level. This allows relevant personnel to intuitively and accurately grasp the current actual operating status of each subsystem, avoiding the uncertainty caused by subjective judgment. Secondly, combining historical operating data, current health status, and simulated fault evolution trends, a time series prediction model is used to predict the remaining useful life (RUL) of key vulnerable components and the probability of specific types of faults occurring in the short term. This provides forward-looking prediction of elevator faults, effectively preventing elevator shutdowns due to sudden faults, greatly ensuring the continuous and stable operation of the elevator, and improving the passenger experience. Thirdly, integrating the anomaly root cause location results, health index, remaining useful life, and short-term fault probability, and combining elevator maintenance procedures and expert knowledge base, specific maintenance or repair suggestions are generated. Maintenance priorities are marked according to the severity and urgency of the fault impact, providing clear and targeted guidance for elevator maintenance work, avoiding blind maintenance and resource waste. Finally, the health status assessment results, predictions, maintenance or repair recommendations, and priorities are filled into a predefined health assessment report template to form a complete health assessment report, allowing the elevator's health information to be presented systematically and in a standardized manner. This standardized report format greatly facilitates communication and exchange among different personnel, thereby better managing and maintaining the elevator, comprehensively improving the overall safety and reliability of the elevator, and reducing operating costs.

[0014] Optionally, the real-time monitoring and early warning method for elevator faults further includes: After generating the health assessment report, the report is pushed to the elevator IoT platform, triggering the following actions: In case of an emergency, a red work order is automatically generated and sent directly to the maintenance personnel's mobile terminal, and a voice alarm is activated. Preventative maintenance requests generate yellow work orders that are synchronized to the property management dispatch system.

[0015] By adopting the above technical solutions, the automatic generation of red work orders for emergency malfunctions, directly connecting to maintenance personnel's mobile terminals and activating voice alarms, significantly improves emergency response speed. When an elevator malfunctions, maintenance personnel can receive detailed fault information and handling suggestions immediately, and the voice alarms ensure that they don't miss important information even in busy or noisy environments. The generation of yellow work orders for preventative maintenance needs, synchronized with the property management dispatch system, reflects the forward-looking and proactive nature of elevator management. Property management departments can plan maintenance work in advance based on the work order content, rationally allocate human and material resources, and avoid resource shortages and inefficiencies caused by ad-hoc maintenance arrangements.

[0016] Optionally, the real-time monitoring and early warning method for elevator faults further includes: A dynamic risk threshold update model is constructed, and the risk threshold is periodically optimized based on historical fault data, elevator service life, operating frequency and environmental parameters.

[0017] By adopting the above technical solutions, constructing the model, and periodically optimizing the risk threshold, the accuracy, effectiveness, and practicality of the real-time monitoring and early warning method for elevator faults can be significantly improved, providing strong support for the safe operation and efficient management of elevators.

[0018] Optionally, the real-time monitoring and early warning method for elevator faults further includes: An online adaptive learning mechanism is constructed to dynamically optimize the cause-effect graph structure and simulation parameter library based on real-time feedback of maintenance results and actual fault data.

[0019] By adopting the above technical solutions, the online adaptive learning mechanism can capture these changing information through real-time feedback of maintenance results and actual fault data. Dynamically optimizing the cause-effect graph structure can more accurately reflect the causal relationships between faults, enabling the monitoring system to more accurately identify abnormal causal events and avoid misjudgments or omissions due to inaccurate causal relationships, thereby improving the accuracy of fault monitoring. Optimizing the simulation parameter library allows for faster and more accurate simulation of faults in the digital twin, making the simulation process more closely resemble actual conditions, thus further improving the safety and stability of elevator operation.

[0020] Secondly, this invention provides a real-time monitoring and early warning system for elevator malfunctions, employing the following technical solution: A real-time elevator fault monitoring and early warning system includes: The data acquisition module is used to collect multi-source sensor data of the elevator and to timestamp and synchronize the multi-source sensor data. The data processing module is used to calculate the comprehensive risk value of the elevator based on the multi-source sensor data synchronized with the timestamp, and to determine whether the comprehensive risk value is greater than the risk threshold. If so, the Granger causality test is run to find abnormal causal events from the pre-built causal graph, and the relevant data corresponding to the abnormal causal events are packaged and sent to the elevator digital twin. The simulation control module is used to perform accelerated fault simulation in the digital twin to locate the root cause of the anomaly; The report generation module is used to generate a health assessment report for the elevator based on the simulation process data and the root causes of the anomalies.

[0021] Thirdly, the present invention provides a terminal, which adopts the following technical solution: A terminal, comprising: The memory contains a real-time monitoring and early warning program for elevator malfunctions. The processor is used to execute the program stored in the memory to implement the steps of the above-described method for real-time monitoring and early warning of elevator malfunctions.

[0022] In summary, the present invention has at least the following beneficial effects: Calculating the elevator's comprehensive risk value based on synchronized data allows for a quantitative assessment of the elevator's overall operational status, providing a clear understanding of the risk level. By determining the relationship between the comprehensive risk value and a risk threshold, potential anomalies during elevator operation can be detected promptly. When the comprehensive risk value exceeds the risk threshold, a Granger causality test is performed, and abnormal causal events are identified from the causal graph. This process delves into the potential causal relationships leading to elevator anomalies, providing clues for accurately locating the root cause of the fault. The relevant data corresponding to the abnormal causal events are packaged and sent to the elevator's digital twin. Leveraging the highly realistic nature of the digital twin, fault simulation is accelerated in a virtual environment, enabling rapid identification of the root cause and significantly reducing troubleshooting time. Finally, a health assessment report is generated based on the simulation data and the root cause of the anomaly. This report provides elevator maintenance personnel with comprehensive and detailed information on the elevator's operational status, facilitating the development of scientific and reasonable maintenance plans. Ultimately, this improves the safety and reliability of elevator operation, reduces the probability of malfunctions, and minimizes safety hazards and economic losses caused by elevator failures. Attached Figure Description

[0023] Figure 1 This is a first flowchart of an embodiment of the method of this application; Figure 2 This is a second flowchart of an embodiment of the method of this application; Figure 3 This is a third flowchart of an embodiment of the method of this application; Figure 4 This is the fourth flowchart of an embodiment of the method of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will be described in conjunction with the appendices in the embodiments of the present invention. Figure 1 - Appendix Figure 4 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The first embodiment of this invention discloses a method for real-time monitoring and early warning of elevator malfunctions. (Refer to...) Figure 1 The monitoring and early warning methods include S110-S160: S110 collects multi-source sensor data from the elevator and timestamps and synchronizes the multi-source sensor data. S120 calculates the comprehensive risk value of the elevator based on multi-source sensor data synchronized with timestamps; S130, determine whether the overall risk value is greater than the risk threshold; S140, if so, run the Granger causality test, find the abnormal causal events from the pre-built causal graph, and package the relevant data corresponding to the abnormal causal events and send them to the elevator digital twin; S150 performs accelerated fault simulation in a digital twin to pinpoint the root cause of anomalies; S160 generates a health assessment report for the elevator based on simulation process data and the root causes of anomalies.

[0026] Specifically, during the S110 data acquisition and synchronization phase, the elevator's existing infrared, speed, and vibration sensors are utilized. An edge computing gateway collects data from these three sensors in real-time at a sampling frequency of 1kHz. The infrared sensors record door opening and closing time sequences, the speed sensors acquire the elevator's speed and acceleration data, and the vibration sensors capture the amplitude of vibrations in the car or tracks. A timestamp alignment mechanism based on the Network Time Protocol (NTP) is employed to unify the raw data from different sensors to a millisecond-level timeline, ensuring time consistency for subsequent feature extraction. The synchronized data is transmitted to the central terminal via LoRa or 5G modules, while sensor values ​​are simultaneously displayed in real-time on a local screen inside the elevator, enabling parallel processing of user visualization and backend processing.

[0027] When S140 determines that the comprehensive risk value exceeds the preset threshold, for example, when the speed overshoot risk value is 120%, the vibration RMS risk value is 80%, and the gate response time risk value is 50%, the comprehensive risk value = 120% × 0.4 + 80% × 0.35 + 50% × 0.25 = 91.5%. Since the risk threshold is set to 80%, 91.5% is greater than 80%, so the Granger causality test is initiated. Based on the pre-constructed causal graph (a knowledge graph trained on historical fault data, containing causal links such as "speed anomaly → emergency stop triggering" and "vibration exceeding limits → bearing wear"), the current abnormal feature data (such as a sudden increase in speed fluctuation variance) is input, and abnormal causal events with statistical significance (P < 0.05) (such as "speed sensor data drift → emergency stop false triggering") are selected.

[0028] Reference Figure 2 In S120, the steps for calculating the overall risk value of the elevator include S210-S230: S210, extract features from the synchronized multi-source sensor data to obtain several key feature indicators related to the elevator's operating status; S220: For each key characteristic indicator, calculate the individual risk value of that key characteristic indicator based on its historical normal operation data and the preset safety threshold range. S230 uses a preset weighted fusion algorithm to sum all individual risk values ​​to obtain the elevator's comprehensive risk value; the weight of each individual risk value is dynamically adjusted according to the importance of different key characteristic indicators for the safe operation of the elevator.

[0029] Specifically, in the S210 feature extraction stage, time-series features such as door operator response time, number of door openings and closings, and obstacle detection frequency are extracted from infrared sensor data; acceleration and deceleration gradients, speed fluctuation variance, and emergency stop trigger threshold deviation are calculated from speed sensor data through a sliding window; and vibration sensor data are decomposed using wavelet transform to obtain the energy proportion and peak factor of three frequency bands (0-50Hz, 50-200Hz, and 200-500Hz), forming a high-dimensional feature set containing nine key feature indicators.

[0030] In the calculation of the S220 individual risk value, a normal distribution model of each feature is constructed based on the normal operation data of the past 12 months. For example, the 95% confidence interval of the speed fluctuation variance is set as the safety threshold range. The risk score is calculated by the formula "Exceeding the upper limit: Individual risk value = (Current feature value - upper threshold) / (Upper threshold - lower threshold), If within the range: 0". Since the obstacle detection frequency of the infrared sensor is directly related to the personal safety of passengers, a stricter preset threshold range needs to be set, such as triggering risk calculation if it exceeds 5 times in a single day.

[0031] When integrating the S230 comprehensive risk value, the Analytic Hierarchy Process (AHP) is used in conjunction with the experience of maintenance experts. The speed sensor characteristics (affecting operational safety) are dynamically assigned a weight of 40%, the vibration sensor characteristics (related to the health of mechanical structures) a weight of 35%, and the infrared sensor characteristics (related to the door operator system) a weight of 25%. The comprehensive risk value is obtained by weighted summation and is displayed intuitively on the elevator display screen in the form of "Risk Index: XX".

[0032] Reference Figure 3 In S150, the steps for accelerating fault simulation and locating the root cause of anomalies in the digital twin include S310-S350: S310, inject data related to abnormal causal events as initial boundary conditions into the digital twin; S320 employs an accelerated simulation algorithm to reproduce the fault evolution process in a digital twin, simulate the response of different components under abnormal excitation, and record the time-series curves of the state parameters of each component. The S330 identifies fault propagation paths by comparing cause-effect graphs with simulation data and combining them with graph neural networks. S340, calculate the contribution of each potential fault source in the fault propagation path, and determine the fault source with the contribution exceeding the preset threshold as the abnormal root cause. S350 performs parameter sensitivity analysis on the root cause of the anomaly to verify whether it is the only or main cause of the fault, eliminates interfering factors with weak correlation, and finally outputs a clear result of the anomaly root cause location.

[0033] Specifically, S310 injects the original sensor data of the abnormal event (such as a continuous 5-second speed fluctuation curve) into the elevator digital twin. This twin is built based on Unity3D and integrates physical models such as the door operator, traction system, and guide rails, supporting multibody dynamics simulation.

[0034] The S320 uses a GPU-accelerated finite element simulation algorithm to reproduce the fault evolution process at 10 times the real-time speed. For example, it simulates the time-series curves of parameters such as the change of traction machine current and the amplitude of car horizontal vibration under the state of speed sensor drift, and simultaneously visualizes the fault propagation dynamics on the digital twin interface.

[0035] The S330 combines causal graphs with simulation data and uses a graph neural network (GNN) to identify fault propagation paths: the state parameters of each component obtained from the simulation (such as bearing temperature and wire rope tension) are used as inputs to the GNN, the nodes of the causal graph are used as priors of the graph structure, and the dependencies between nodes are learned through an attention mechanism to output the propagation path of "speed sensor drift → traction machine load unevenness → car sway → vibration sensor over-limit".

[0036] The S340 uses the Shapley value algorithm to calculate the contribution of each node in the path, compares the contribution of speed sensor drift (e.g., 85%) with a preset threshold (50%), and determines it as the root cause of the anomaly.

[0037] S350 uses parameter sensitivity analysis to verify the impact of the speed sensor sampling frequency on the probability of emergency stop by fixing other variables and adjusting it only (e.g., sensitivity coefficient 0.92). It eliminates weak correlation factors such as guide rail installation error (sensitivity coefficient 0.15) and finally displays "Root cause of anomaly: speed sensor drift (confidence level 98%)" on the elevator display screen.

[0038] Reference Figure 4 S160, the steps for generating a health assessment report based on simulation process data and root causes of anomalies include S410-S440: S410, based on the comparison between simulated process data and normal operation benchmark model, adopts a multi-index fusion evaluation algorithm to quantify the real-time health status of each key subsystem of the elevator. The health status includes health index and status level. S420, combining historical operating data, current health status, and simulated failure evolution trends, uses a time series prediction model to predict the remaining useful life (RUL) of key vulnerable components and the probability of specific types of failures occurring in the short term. S430 integrates the results of anomaly root cause location, health index, remaining service life and short-term failure probability, and combines elevator maintenance procedures and expert knowledge base to generate specific maintenance or repair suggestions, and marks maintenance priorities according to the severity of the failure and the urgency of its occurrence. S440 fills the predefined health assessment report template with the health status assessment results, prediction results, maintenance or repair recommendations and priorities to form a complete health assessment report.

[0039] Specifically, in the S410 stage, a multi-index fusion evaluation method based on principal component analysis (PCA) and dynamic time warping (DTW) can be adopted. First, the characteristic spatial distribution of multi-dimensional parameters such as vibration frequency, motor current, door operator action sequence, and traction machine temperature of the elevator under standard operating conditions is extracted using the normal operation benchmark model to construct a low-dimensional health reference manifold. Then, the real-time collected simulation process data is mapped to the same feature space. By calculating the weighted deviation between Euclidean distance and Mahalanobis distance, and introducing the fuzzy comprehensive evaluation method, the health of key subsystems such as hydraulic system, control system, and safety protection device is quantitatively scored. Finally, the health index (HDI) in the range of 0 to 1 is output, and four status levels of "healthy", "sub-healthy", "mildly degraded" and "severely abnormal" are divided according to the preset threshold.

[0040] In the S420 stage, to improve the accuracy and timeliness of remaining useful life (RUL) prediction, a time series joint prediction framework integrating Long Short-Term Memory (LSTM) network and Weibull proportional hazards model can be constructed. On the one hand, LSTM is used to capture nonlinear degradation trajectories in historical operating data, especially the hidden evolution patterns of gradual failures such as brake wear, wire rope fatigue, and increased guide shoe clearance. On the other hand, a degradation state transition model based on the Wiener process is established by combining the component failure database obtained from accelerated degradation tests, and the current health index is used as the hidden state input to realize the probabilistic inference of the future performance degradation path of key vulnerable components. On this basis, a Bayesian update mechanism is further integrated to dynamically correct the prior distribution according to the latest monitoring data, so as not only predicting the expected value of RUL and its confidence interval, but also simultaneously calculating the conditional probability of typical failures such as inaccurate leveling, door lock failure, and emergency stop triggering within the next 72 hours.

[0041] By deeply integrating the aforementioned quantitative analysis results with qualitative knowledge, a maintenance suggestion generation engine based on rule-case hybrid reasoning is constructed. This engine first accesses the root cause localization results of S410 anomalies (e.g., "encoder signal drift causing position detection error" or "contactor contact oxidation causing control delay"), using these as fault semantic tags to match typical fault mode libraries in the expert knowledge base. Then, combining the degree of degradation reflected by the current health index, the aging stage indicated by the RUL prediction value, and the risk urgency reflected by the short-term failure probability, it calls upon a preset IF-THEN rule set (e.g., "if brake pad RUL < 30 days and failure probability > 60%, immediate replacement is recommended"). Simultaneously, a case retrieval mechanism is used to search for successfully handled maintenance records under similar historical conditions. Finally, based on national elevator maintenance standards such as GB / T18775 and TSGT5002, the engine verifies whether the suggested content meets mandatory requirements and marks maintenance priorities according to three levels: "emergency stop," "major hidden danger," and "general defect."

[0042] In the S440 phase, the system organizes the aforementioned multi-source heterogeneous information and injects it into a standardized health assessment report template. This template uses XMLSchema to define structured fields, including but not limited to basic equipment information, test environment description, health radar charts for each subsystem, RUL trend curves, failure probability heatmaps, root cause analysis summaries, maintenance recommendation lists, and priority indicators. Utilizing Natural Language Generation (NLG) technology, the system can automatically generate coherent and fluent technical commentary paragraphs, such as: "The current health index of the traction system is 0.78, indicating a sub-healthy state. The main reasons are high bearing temperature and increased vibration energy. Combined with LSTM prediction results, the remaining life of the main shaft bearing is approximately 45±7 days, and the probability of an overheating alarm within the next 48 hours is 52%. It is recommended to arrange a special inspection during the next scheduled maintenance window and replace spare parts in advance." The report is ultimately output in both PDF and JSON formats. The former is used for manual review and archiving, while the latter facilitates integration with the elevator IoT platform for automated task dispatch.

[0043] Real-time monitoring and early warning methods for elevator malfunctions also include: After generating the health assessment report, the report is pushed to the elevator IoT platform, triggering the following actions: Emergency faults automatically generate red work orders that are directly sent to maintenance personnel's mobile terminals and trigger voice alarms; preventative maintenance needs generate yellow work orders that are synchronized to the property management dispatch system.

[0044] Specifically, the health assessment report is automatically pushed to the elevator IoT platform, triggering actions: emergency faults (such as red work orders) are directly generated and sent to the maintenance personnel's mobile terminals (such as APP), and voice alarms are activated at the same time; preventive maintenance needs (such as yellow work orders) are synchronized to the property management dispatch system; at the same time, the display screen provided by the customer displays the sensor sensing content (such as infrared door opening and closing status), fault information and maintenance priority in real time, ensuring that passengers can see it, and transmits the same content to the central terminal and the maintenance center terminal, so that the maintenance center can receive the elevator address and fault details, and prepare tools and parts accordingly, and arrange personnel for efficient response.

[0045] In addition, methods for real-time monitoring and early warning of elevator malfunctions also include: A dynamic risk threshold update model is constructed, and the risk threshold is periodically optimized based on historical fault data, elevator service life, operating frequency and environmental parameters. An online adaptive learning mechanism is constructed to dynamically optimize the cause-effect graph structure and simulation parameter library based on real-time feedback of maintenance results and actual fault data.

[0046] Specifically, a dynamic risk threshold update model is constructed. Based on historical fault data, elevator service life, operating frequency (extracted from sensor logs), and environmental parameters (such as temperature and humidity), the risk threshold is periodically optimized (e.g., updated monthly) using regression analysis or machine learning algorithms (such as random forest) to ensure that the threshold adapts to elevator aging or changes in usage patterns.

[0047] An online adaptive learning mechanism is constructed, which dynamically optimizes the causal graph structure (through graph neural network retraining) and simulation parameter library (such as adjusting the friction coefficient model of digital twin) based on real-time feedback maintenance results (obtained from maintenance work orders) and real fault data (such as sensor readings after maintenance), forming a closed-loop optimization to continuously improve diagnostic accuracy.

[0048] Based on the above method embodiments, the second embodiment of the present invention discloses an elevator fault real-time monitoring and early warning system. The elevator fault real-time monitoring and early warning system of this embodiment can implement any of the above-described elevator fault real-time monitoring and early warning methods, and the specific working process of each module in the elevator fault real-time monitoring and early warning system can be referred to the corresponding process in the above method embodiments.

[0049] For ease of understanding, an example is as follows: A real-time elevator fault monitoring and early warning system includes: The data acquisition module is used to collect multi-source sensor data from the elevator and to synchronize the multi-source sensor data with timestamps. The data processing module is used to calculate the comprehensive risk value of the elevator based on the multi-source sensor data synchronized with the timestamp, and to determine whether the comprehensive risk value is greater than the risk threshold. If so, the Granger causality test is run to find abnormal causal events from the pre-built causal graph, and the relevant data corresponding to the abnormal causal events are packaged and sent to the elevator digital twin. The simulation control module is used to accelerate fault simulation in a digital twin and locate the root cause of anomalies. The report generation module is used to generate a health assessment report for the elevator based on simulation process data and the root causes of anomalies.

[0050] A third embodiment of the present invention provides a terminal. As one implementation of the terminal, the terminal may include: a memory and a processor; wherein, The memory is used to store the real-time monitoring and early warning program for elevator malfunctions; The processor is used to execute the program stored in the memory to implement the steps of the above-mentioned elevator fault real-time monitoring and early warning method.

[0051] The memory can communicate with the processor via a communication bus, which can be an address bus, a data bus, a control bus, etc.

[0052] Additionally, the memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device.

[0053] Furthermore, the processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0054] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for real-time monitoring and early warning of elevator failure, characterized in that, The method comprises the following steps: Collecting multi-source sensor data of the elevator and synchronizing the time stamps of the multi-source sensor data; Calculating a comprehensive risk value of the elevator based on the multi-source sensor data synchronized by time stamps; Determining whether the comprehensive risk value is greater than a risk threshold value; If yes, performing Granger causality test, searching for an abnormal causality event from a pre-constructed causality graph, and packaging and sending relevant data corresponding to the abnormal causality event to an elevator digital twin; Performing fault acceleration simulation in the digital twin to locate the abnormal root cause; Generating a health assessment report of the elevator according to simulation process data and the abnormal root cause.

2. The method of real-time monitoring and early warning of elevator failure according to claim 1, characterized in that, The step of calculating the comprehensive risk value of the elevator comprises: Extracting features from the synchronized multi-source sensor data to obtain a plurality of key feature indicators related to the running state of the elevator; For each key feature indicator, calculating a single risk value of the key feature indicator based on historical normal operation data and a preset safety threshold range; Summing all single risk values by a preset weighted fusion algorithm to obtain the comprehensive risk value of the elevator; the weight corresponding to each single risk value is dynamically adjusted according to the importance degree of different key feature indicators to the safe operation of the elevator.

3. The method of real-time monitoring and early warning of elevator failure according to claim 1, characterized in that, The step of performing fault acceleration simulation in the digital twin to locate the abnormal root cause comprises: Injecting the abnormal causality event related data as an initial boundary condition into the digital twin; Reproducing the fault evolution process in the digital twin by using an acceleration simulation algorithm, simulating the response of different components under abnormal excitation, and recording the state parameter time curve of each component; Identifying the fault propagation path by comparing the causality graph with the simulation data and combining the graph neural network; Calculating the contribution degree of each potential fault source in the fault propagation path, and determining the fault source whose contribution degree exceeds a preset threshold as the abnormal root cause; Performing parameter sensitivity analysis on the abnormal root cause to verify whether it is the only or main cause of the fault, excluding weakly related interference factors, and finally outputting a clear abnormal root cause positioning result.

4. The method of real-time monitoring and early warning of elevator failure according to claim 3, characterized in that, The step of generating the health assessment report comprises: Based on the comparison between the simulation process data and the normal operation benchmark model, using a multi-index fusion evaluation algorithm to quantify the real-time health status of each key subsystem of the elevator, including the health degree index and the state level; Using a time series prediction model to predict the remaining useful life (RUL) of the key vulnerable components and the probability of occurrence of a specific type of fault in the short term, combining the historical operation data, the current health status and the simulated fault evolution trend; Integrating the abnormal root cause positioning result, the health degree index, the remaining useful life and the short-term fault probability, combining the elevator maintenance regulations and the expert knowledge base to generate specific maintenance or repair recommendations, and marking the maintenance priority according to the fault impact severity and urgency; Filling the health status evaluation results, the prediction results, the maintenance or repair recommendations and the priority into a pre-defined health assessment report template to form a complete health assessment report.

5. The method of real-time monitoring and early warning of elevator failure according to claim 1, characterized in that, The elevator fault real-time monitoring and early warning method further comprises: After generating the health assessment report, pushing the health assessment report to an elevator Internet of Things platform and triggering the following actions: Emergency failure automatically generates a red work order to pass through the maintenance personnel mobile terminal and starts the voice alarm; Preventive maintenance needs generate a yellow work order to synchronize to the property management dispatching system.

6. The method of real-time monitoring and early warning of elevator failure according to claim 1, characterized in that, The elevator failure real-time monitoring and early warning method further comprises: A dynamic risk threshold updating model is constructed to periodically optimize the risk threshold based on historical failure data, elevator service life, operation frequency and environmental parameters.

7. The method of real-time monitoring and early warning of elevator failure according to claim 5, characterized in that, The elevator failure real-time monitoring and early warning method further comprises: An online adaptive learning mechanism is constructed to dynamically optimize the causal graph structure and simulation parameter library based on real-time feedback maintenance results and real failure data.

8. A real-time monitoring and early warning system for elevator failure, characterized in that, The elevator failure real-time monitoring and early warning method according to any one of claims 1-7 comprises: A data acquisition module for acquiring multi-source sensor data of the elevator and synchronizing the time stamp of the multi-source sensor data; A data processing module for calculating the comprehensive risk value of the elevator based on the time-stamped multi-source sensor data, and determining whether the comprehensive risk value is greater than the risk threshold; if so, performing Granger causality test, finding abnormal causal events from the pre-constructed causal graph, and packaging and sending the related data corresponding to the abnormal causal events to the elevator digital twin; An analog control module for performing failure acceleration simulation in the digital twin to locate the abnormal root cause; A report generation module for generating a health assessment report of the elevator according to the simulation process data and the abnormal root cause.

9. A terminal, characterized by comprising: It comprises: A memory storing an elevator failure real-time monitoring and early warning program; A processor for executing the program stored on the memory to realize the steps of the elevator failure real-time monitoring and early warning method according to any one of claims 1-7.

Images