Intelligent elevator safety fault early warning method and system based on deep learning

Abstract

The invention relates to the technical field of elevator fault early warning, in particular to an intelligent elevator safety fault early warning method and system based on deep learning. The method comprises the following steps that real-time multi-dimensional operation state parameters of the elevator are obtained; time sequence change characteristic mining is conducted on the multi-dimensional operation state parameters, and therefore elevator temperature and humidity time sequence state characteristics are obtained; acquiring a multidirectional elevator monitoring video stream; video interval sampling is carried out on the multidirectional elevator monitoring video stream, and a monitoring image frame sequence is constructed; performing three-dimensional topological structure layout analysis on the monitoring image frame sequence to obtain elevator three-dimensional topological structure layout data; building a three-dimensional topological geometric model of the elevator based on the layout data of the three-dimensional topological structure of the elevator; and carrying out part wear state analysis on the three-dimensional topological geometric model of the elevator, carrying out potential structure defect analysis, and extracting potential topological structure defect data. According to the invention, efficient and accurate elevator safety fault early warning is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator fault warning technology, and in particular to a deep learning-based intelligent elevator safety fault warning method and system. Background Art

[0002] With the rapid development of industrial automation and intelligence, elevators have been widely used as important transportation equipment in various buildings. The safety and reliability of elevators are related to the safety of people’s lives and property, and therefore have become a key and difficult problem that needs to be urgently solved in the current field of industrial automation.

[0003] In actual use, elevators often need to operate for a long time at a high intensity and are easily affected by various internal and external factors such as component wear, load changes, mechanical vibrations, etc. These factors can cause key elevator components such as motors and braking systems to have hidden dangers of failure, and ultimately cause elevator safety accidents. Traditional elevator fault diagnosis mainly relies on manual inspections and regular maintenance, which is inefficient and inaccurate, and it is difficult to detect potential safety hazards in a timely manner.

[0004] Therefore, there is an urgent need to develop an intelligent elevator safety fault warning method based on deep learning, which can fully perceive the operating status of the elevator, predict the occurrence of faults in advance, and take differentiated warning measures according to the type and severity of the fault to ensure the safe and stable operation of the elevator. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a deep learning-based intelligent elevator safety fault warning method and system to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a smart elevator safety fault early warning method based on deep learning, comprising the following steps: Step S1: obtaining real-time multi-dimensional operating state parameters of the elevator; mining the time series variation characteristics of the multi-dimensional operating state parameters to obtain the time series state characteristics of the elevator temperature and humidity; Step S2: obtaining a multi-directional elevator monitoring video stream; performing video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; Step S3: performing a three-dimensional topological structure layout analysis on the monitoring image frame sequence to obtain the three-dimensional topological structure layout data of the elevator; and constructing a three-dimensional topological geometric model of the elevator based on the three-dimensional topological structure layout data of the elevator; Step S4: Analyze the wear state of components on the three-dimensional topological geometric model of the elevator, and analyze potential structural defects, and extract potential topological structural defect data; Step S5: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating an elevator fault diffusion trajectory; Step S6: Performing an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; generating an adaptive fault warning signal based on the adaptive fault decision strategy.

[0007] The present invention helps the system to timely understand the operation status of the elevator by acquiring the multi-dimensional operation status parameters of the elevator in real time, improves the perception of potential faults, obtains the time series state characteristics of the elevator temperature and humidity by mining the time series change characteristics, helps the system to more comprehensively understand the evolution process of the elevator operation status, obtains multi-directional elevator monitoring video streams to provide a more comprehensive perspective, helps the system to monitor the elevator operation status and timely discover abnormal situations, constructs a monitoring image frame sequence by video interval sampling, provides visual support for subsequent fault warnings, obtains the three-dimensional topological structure layout data of the elevator by three-dimensional topological structure layout analysis, helps to build a more realistic elevator model, builds a three-dimensional topological geometric model of the elevator based on the topological structure layout data, provides the system with more intuitive elevator structure information, and analyzes the wear of elevator components. The damage state and potential structural defects help the system to monitor the health status of the elevator in real time and prevent the occurrence of potential faults. The extraction of potential topological structure defect data helps the system to more accurately identify the problems of the elevator and provide a basis for subsequent early warnings. The defect state evolution of the potential topological structure defect data is simulated according to the temperature and humidity time series state characteristics of the elevator to generate the elevator fault diffusion trajectory, which helps the system predict the development trajectory of the fault. Through the fault diffusion trajectory, the system can better understand the propagation path of the fault, which is conducive to taking preventive measures in time. The adaptive fault decision-making intelligently adjusts the response strategy according to the real-time situation, improves the efficiency and accuracy of the system's response to faults, and generates adaptive fault warning signals based on the adaptive fault decision-making strategy, which helps the system to issue alarms in time and take necessary measures to ensure the safety of the elevator.

[0008] Preferably, step S1 comprises the following steps: Step S11: obtaining real-time multi-dimensional operating status parameters of the elevator based on sensors; Step S12: performing time synchronization processing on the multi-dimensional operating status parameters to obtain the operating status parameters with consistent timestamps; Step S13: Perform temperature and humidity characteristic analysis on the time-stamp consistent operating status parameters to extract real-time temperature and humidity characteristic data; Step S14: Calculate the change amplitude of multiple time points on the real-time temperature and humidity characteristic data to generate temperature and humidity change amplitude data; Step S15: mining the time series variation characteristics of the temperature and humidity variation amplitude data to obtain the time series state characteristics of the elevator temperature and humidity.

[0009] The present invention obtains multi-dimensional operating status parameters in real time through sensors to help the system understand the operating status of the elevator in a timely manner and enhance the perception and monitoring capabilities of the real-time status of the elevator. Time synchronization processing can ensure that the multi-dimensional operating status parameters have consistent timestamps, helping the system avoid problems caused by time differences when analyzing and comparing data. Analyzing the temperature and humidity characteristics in the operating status parameters with consistent timestamps helps the system understand the real-time situation of the internal environment of the elevator and provides important reference data. Calculating the change amplitude of real-time temperature and humidity characteristic data at different time points can help the system understand the fluctuation of temperature and humidity and predict abnormal situations. Through the mining of time series change characteristics, the system analyzes the temperature and humidity change amplitude data more carefully, discovers potential rules and abnormal changes, and provides a more accurate basis for fault warning.

[0010] Preferably, step S2 comprises the following steps: Step S21: Obtain multi-directional elevator monitoring video stream; Step S22: performing video interval sampling on the multi-directional elevator monitoring video stream to obtain multiple single monitoring image frames; Step S23: reconstructing the details of the multiple single surveillance image frames frame by frame to obtain multiple detail reconstructed image frames; Step S24: performing image-by-image temporal coding processing on the plurality of detail reconstructed image frames to obtain a code for each image frame; Step S25: performing pixel size standardization processing on a plurality of detail reconstructed image frames to obtain a plurality of standardized detail reconstructed image frames; Step S26: sequentially encode and sequence a plurality of standardized detail reconstructed image frames according to the encoding of each image frame to construct a monitoring image frame sequence.

[0011] The present invention obtains multi-directional elevator monitoring video streams, and the system comprehensively monitors the internal and external environment of the elevator, thereby improving the perception ability of the elevator safety status. Video interval sampling helps to reduce the complexity of data processing, making subsequent processing more efficient. A single monitoring image frame is obtained by sampling, which saves storage space and ensures the capture of key information. Frame-by-frame detail reconstruction helps to improve the clarity and quality of the monitoring image frame, better display the details of the internal and external environment of the elevator, and reconstructs the image frame to restore the detail information of the original monitoring image, improve the accuracy and reliability of the monitoring data, image time series coding processing compresses image data, reduces storage space occupancy, and improves data transmission efficiency. Coding processing helps to speed up data processing speed and improve the real-time and response speed of the system. Pixel size standardization can ensure that the pixel sizes between different image frames are consistent, which is conducive to subsequent processing and comparative analysis. Standardization processing helps to improve the display effect and compatibility of images on different devices. Sequential coding serialization processing helps to integrate each image frame into a continuous monitoring image frame sequence, which is convenient for subsequent data processing and analysis. Constructing a monitoring image frame sequence allows the system to perform time series analysis to detect any abnormal situation in the internal and external environment of the elevator and perform safety fault warning.

[0012] Preferably, the specific steps of step S23 are: Extracting a first image frame based on a plurality of single surveillance image frames; Performing brightness enhancement processing on the first image frame to obtain a brightness enhanced image frame; Perform forward multi-layer convolution calculation on the brightness enhanced image frame to extract the low-level texture features of the image frame; Nonlinearly compress low-level texture features of the image frame to obtain mid-level features of the image frame; Perform deep local feature extraction on the mid-level features of the image frame to obtain spatial local convolution features; Perform temporal convolution arrangement on the low-level texture features of the image frame, the mid-level features of the image frame and the spatial local convolution features to obtain a multi-channel convolution feature sequence; Performing convolution detail reconstruction on the first image frame according to the multi-channel convolution feature sequence to obtain a first detail reconstructed image frame; The convolution detail reconstruction process is repeated on the multiple single monitoring image frames until all the single monitoring image frames are processed, thereby obtaining multiple detail reconstructed image frames.

[0013] The present invention extracts the first image frame as the basis for subsequent processing, which is used to analyze and process the entire surveillance video sequence. By enhancing the brightness of the image, the visual quality of the image is improved, making the image clearer and easier to analyze. Through forward multi-layer convolution calculation, the low-level texture features of the image are effectively extracted, and these features are crucial for subsequent feature extraction and analysis. Nonlinear compression helps to reduce the dimension of the features and highlight the key features in the image, thereby simplifying the subsequent processing steps. Through deep local feature extraction, the spatial local information of the image is better captured, which helps to identify important features in the image. Through convolution detail reconstruction, the detail information of the image is effectively restored, and the clarity and quality of the image are improved.

[0014] Preferably, step S3 specifically comprises the following steps: Step S31: identifying structural feature points of the monitoring image frame sequence and marking the three-dimensional structural feature points; Step S32: performing multi-angle feature matching on the three-dimensional structural feature points, thereby obtaining spatial matching data of the structural feature points; Step S33: Calculating the spatial position of the feature points based on the spatial matching data of the structural feature points to generate the spatial position coordinates of the structural feature points; Step S34: performing a three-dimensional topological structure layout analysis on the spatial position coordinates of the structural feature points to obtain three-dimensional topological structure layout data of the elevator; Step S35: performing key connection detection on the three-dimensional structural feature points to obtain structural connection points; Step S36: Based on the structural connection points, three-dimensional topological connection modeling is performed on the elevator three-dimensional topological structure layout data to construct a three-dimensional topological geometric model of the elevator.

[0015] The present invention marks three-dimensional structural feature points, and the system identifies key features of the internal and external structures of the elevator, which is helpful for subsequent structural analysis and fault warning. Identifying structural feature points helps to extract key information, including data on elevator structure, layout, etc., and provides more effective information for the system. Multi-angle feature matching improves the matching accuracy and obtains more reliable spatial matching data of structural feature points. By matching data, the association between structural feature points at different angles is established, which is helpful for the analysis and modeling of the overall structure. By calculating the spatial position, the position of the structural feature points in the three-dimensional space is determined, and accurate position information is provided for subsequent topological structure analysis. The generation of spatial position coordinates helps to fully describe the spatial layout of the structural feature points, which is helpful for the overall structure. The analysis provides complete data support. Through the three-dimensional topological structure layout analysis, the system can deeply understand the overall structural layout of the elevator and discover potential structural problems or abnormalities. Analysis based on layout data is helpful for structural optimization and improving the overall performance and safety of the elevator. Detecting key connection points helps to determine the important connection parts in the structure and provide important clues for subsequent modeling and analysis. Through the detection of connection points, potential structural connection problems or abnormalities can be discovered, and early warning and processing can be carried out. Establishing a three-dimensional topological connection model helps to fully describe the structural characteristics of the elevator and provide a more intuitive geometric model for the system. The three-dimensional topological geometric model can be used to simulate and analyze the elevator structure, and plays an important role in fault warning and maintenance.

[0016] Preferably, the specific steps of step S4 are: Step S41: performing surface optical flow fluctuation analysis on the three-dimensional topological geometric model of the elevator to obtain the elevator surface structure fluctuation data; Step S42: positioning elevator components based on the elevator surface structure fluctuation data to obtain component nodes at defective locations; Step S43: Analyze the component wear status of the component node at the defective position to generate component wear status data; Step S44: Perform potential structural defect analysis on the component wear state data to extract potential topological structural defect data.

[0017] The present invention obtains the fluctuation data of the elevator surface structure through optical flow fluctuation analysis, which is helpful to understand the surface condition of the elevator structure. The obtained fluctuation data is used to monitor the health status of the elevator structure and discover abnormal conditions in time. By locating components, the component nodes of the defective positions are determined, and key information for locating faults and problems is provided. The positioning of component nodes helps to quickly respond to and solve potential elevator fault problems and improve fault handling efficiency. The component wear status analysis is performed to generate wear status data, which is helpful to evaluate the health status of the components, discover the component wear status in advance for preventive maintenance, and reduce the occurrence of sudden faults. The component wear status data is analyzed for potential structural defects, which is helpful to identify potential structural defect problems. The potential defect data is extracted for the fault warning system of the intelligent elevator, which helps to prevent faults in advance and ensure the safety and reliability of the elevator operation.

[0018] Preferably, the specific steps of step S5 are: Step S51: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics to generate topological structure defect evolution data; Step S52: performing deep defect evolution learning on topological structure defect evolution data to obtain topological defect evolution rules; Step S53: predicting the elevator failure probability based on the topological defect evolution law, thereby obtaining the elevator failure probability; Step S54: Evolving the fault diffusion trajectory based on the elevator failure probability, thereby generating the elevator fault diffusion trajectory.

[0019] The present invention simulates the evolution of defect states through the time series state characteristics of temperature and humidity, predicts the evolution trend of topological structure defects, and generates topological structure defect evolution data, which is helpful for preventive maintenance, early intervention in potential faults, and reducing the likelihood of faults. The topological structure defect evolution data is analyzed through deep learning to obtain the topological defect evolution law, which is helpful to discover the pattern and law of defect evolution. The learned evolution law is used to improve the accuracy and reliability of elevator fault prediction. Fault probability prediction is performed based on the topological defect evolution law to obtain the elevator failure probability, which helps to evaluate the likelihood of fault occurrence. Accurate fault probability prediction optimizes the fault warning system and improves the accuracy and timeliness of the warning. Fault diffusion trajectory is generated according to the fault probability, which helps to visualize the propagation path and impact range of the fault. Through fault diffusion trajectory analysis, the impact of the fault on the elevator system is evaluated, which helps to formulate countermeasures and reduce losses.

[0020] Preferably, the specific steps of step S6 are: Step S61: performing fault type identification on the elevator fault diffusion trajectory to generate the elevator fault type; Step S62: Perform safety risk quantitative analysis on the elevator fault diffusion trajectory to obtain the elevator fault safety risk level; Step S63: performing an adaptive fault decision based on the elevator fault type and the elevator fault safety risk level to generate an adaptive fault decision strategy; Step S64: Perform fault warning based on the adaptive fault decision strategy and generate an adaptive fault warning signal.

[0021] The present invention identifies the characteristics in the elevator fault diffusion trajectory and accurately divides different types of faults, which helps to take targeted countermeasures. The generation of elevator fault types helps to quickly and accurately locate problems and improve maintenance efficiency and response speed. By quantitatively analyzing the safety risks of the elevator fault diffusion trajectory and evaluating the potential impact of the fault on the elevator system and users, the elevator fault safety risk level is obtained, which helps to take corresponding safety measures to ensure the safety of elevator operation. In combination with the fault type and the safety risk level, a targeted fault decision-making strategy is formulated to improve the effectiveness and efficiency of maintenance and repair. The adaptive fault decision allocates resources according to specific circumstances, optimizes maintenance plans, reduces maintenance costs and improves the reliability of the elevator system. The adaptive fault decision strategy generates a fault warning signal to warn of the occurrence of faults in advance and avoid potential risks. The implementation of the adaptive fault warning helps to improve the safety of the elevator system and ensure the safety of passengers and equipment.

[0022] In this specification, a deep learning-based intelligent elevator safety fault warning system is provided, which is used to execute the deep learning-based intelligent elevator safety fault warning method as described above, including: The time series status module is used to obtain the real-time multi-dimensional operating status parameters of the elevator; the time series change characteristics of the multi-dimensional operating status parameters are mined to obtain the time series status characteristics of the elevator temperature and humidity; An image processing module is used to obtain a multi-directional elevator monitoring video stream; perform video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; A topology structure module is used to perform a three-dimensional topology structure layout analysis on a monitoring image frame sequence to obtain the three-dimensional topology structure layout data of the elevator; and to construct a three-dimensional topology geometric model of the elevator based on the three-dimensional topology structure layout data of the elevator; The structural defect module is used to analyze the wear status of components of the elevator's three-dimensional topological geometric model, analyze potential structural defects, and extract potential topological structural defect data; The fault prediction module is used to simulate the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating the elevator fault diffusion trajectory; The fault warning module is used to make an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; and generate an adaptive fault warning signal based on the adaptive fault decision strategy.

[0023] The present invention can realize real-time monitoring of the elevator status by acquiring the real-time multi-dimensional operating status parameters of the elevator. By mining the time series change characteristics of the multi-dimensional operating status parameters, the time series status characteristics of the elevator temperature and humidity are obtained, which is helpful to identify abnormal states and trends, obtain multi-directional elevator monitoring video streams, provide a full range of monitoring vision, and construct monitoring image frame sequences through video interval sampling, which is helpful to analyze the elevator operating status and detect abnormal situations. The monitoring image frame sequence is subjected to three-dimensional topological structure layout analysis to obtain the elevator three-dimensional topological structure layout data, which helps to understand the elevator structure layout and the positions of key components. The elevator three-dimensional topological geometric model is constructed based on the three-dimensional topological structure layout data, which provides a basis for subsequent analysis. Data, analyze the component wear status of the elevator's three-dimensional topological geometric model, extract potential structural defect data, which helps to identify existing structural problems, conduct potential structural defect analysis, help identify potential structural problems and prevent failures, simulate the defect state evolution of potential topological structural defect data according to the elevator's temperature and humidity time series state characteristics, generate the elevator fault diffusion trajectory, which helps to predict the path of the failure, make adaptive fault decisions on the elevator fault diffusion trajectory, generate adaptive fault decision strategies, help formulate targeted fault handling plans, generate adaptive fault warning signals based on the adaptive fault decision strategies, issue alarms in time and take preventive measures to improve the safety and reliability of elevators. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of the steps of a deep learning-based intelligent elevator safety fault warning method of the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0026] The example of this application provides a method and system. The execution subjects of the method and system include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of audio and image management system, information management system, and cloud data management system.

[0027] See also Figures 1 to 4 The present invention provides a smart elevator safety fault early warning method based on deep learning, and the smart elevator safety fault early warning method based on deep learning comprises the following steps: Step S1: obtaining real-time multi-dimensional operating state parameters of the elevator; mining the time series variation characteristics of the multi-dimensional operating state parameters to obtain the time series state characteristics of the elevator temperature and humidity; Step S2: obtaining a multi-directional elevator monitoring video stream; performing video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; Step S3: performing a three-dimensional topological structure layout analysis on the monitoring image frame sequence to obtain the three-dimensional topological structure layout data of the elevator; and constructing a three-dimensional topological geometric model of the elevator based on the three-dimensional topological structure layout data of the elevator; Step S4: Analyze the wear state of components on the three-dimensional topological geometric model of the elevator, and analyze potential structural defects, and extract potential topological structural defect data; Step S5: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating an elevator fault diffusion trajectory; Step S6: Performing an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; generating an adaptive fault warning signal based on the adaptive fault decision strategy.

[0028] The present invention helps the system to timely understand the operation status of the elevator by acquiring the multi-dimensional operation status parameters of the elevator in real time, improves the perception of potential faults, obtains the time series state characteristics of the elevator temperature and humidity by mining the time series change characteristics, helps the system to more comprehensively understand the evolution process of the elevator operation status, obtains multi-directional elevator monitoring video streams to provide a more comprehensive perspective, helps the system to monitor the elevator operation status and timely discover abnormal situations, constructs a monitoring image frame sequence by video interval sampling, provides visual support for subsequent fault warnings, obtains the three-dimensional topological structure layout data of the elevator by three-dimensional topological structure layout analysis, helps to build a more realistic elevator model, builds a three-dimensional topological geometric model of the elevator based on the topological structure layout data, provides the system with more intuitive elevator structure information, and analyzes the wear of elevator components. The damage state and potential structural defects help the system to monitor the health status of the elevator in real time and prevent the occurrence of potential faults. The extraction of potential topological structure defect data helps the system to more accurately identify the problems of the elevator and provide a basis for subsequent early warnings. The defect state evolution of the potential topological structure defect data is simulated according to the temperature and humidity time series state characteristics of the elevator to generate the elevator fault diffusion trajectory, which helps the system predict the development trajectory of the fault. Through the fault diffusion trajectory, the system can better understand the propagation path of the fault, which is conducive to taking preventive measures in time. The adaptive fault decision-making intelligently adjusts the response strategy according to the real-time situation, improves the efficiency and accuracy of the system's response to faults, and generates adaptive fault warning signals based on the adaptive fault decision-making strategy, which helps the system to issue alarms in time and take necessary measures to ensure the safety of the elevator.

[0029] In the embodiment of the present invention, refer to Figure 1 , is a schematic diagram of the steps of a deep learning-based intelligent elevator safety fault warning method of the present invention. In this example, the steps of the deep learning-based intelligent elevator safety fault warning method include: Step S1: obtaining real-time multi-dimensional operating state parameters of the elevator; mining the time series variation characteristics of the multi-dimensional operating state parameters to obtain the time series state characteristics of the elevator temperature and humidity; In this embodiment, a variety of sensors are installed at key parts of the elevator (such as the elevator car, machine room, etc.) to collect the required operating status parameters in real time. A data acquisition system is designed and deployed to ensure that the data collected from each sensor can be transmitted to the central control system in real time. Wireless transmission, Ethernet or other communication protocols are used to ensure that the collected data can be stored in the database. A suitable database management system (such as MySQL, MongoDB) is used for data management for subsequent analysis. The real-time collected data is cleaned, including removing outliers and filling missing values, to ensure the accuracy and completeness of the data. A time series analysis method is used to extract features from temperature and humidity data. Commonly used methods include: sliding Moving window method: Apply sliding window technology to time series to calculate statistical features (such as mean, standard deviation, maximum, and minimum) in each time period. Frequency domain analysis: Analyze the frequency characteristics of temperature and humidity data through Fourier transform or wavelet transform, identify periodic changes, and use feature selection methods (such as correlation analysis and principal component analysis) to select features related to the elevator operation status, reduce redundant information, and improve the efficiency of subsequent analysis. Based on the extracted features, build an elevator temperature and humidity time series state feature model, use time series prediction models (such as ARIMA and LSTM) to capture the change pattern of temperature and humidity, visualize the time series change characteristics, and display the change trend of temperature and humidity through charts to facilitate intuitive analysis and decision-making.

[0030] Step S2: obtaining a multi-directional elevator monitoring video stream; performing video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; In this embodiment, multiple cameras are installed at different locations of the elevator (such as in the car, at the entrance, in the machine room, etc.) to ensure that all key areas of the elevator are covered. The camera should support high-definition video capture to ensure image quality. The camera settings, including resolution, frame rate, encoding format, etc., are configured to ensure the stability and smoothness of the video stream. A network camera (IP camera) or an analog camera is selected to transmit the video stream captured by the camera to the central monitoring system in real time through the network or other transmission methods (such as HDMI, USB), ensuring low latency and high bandwidth for data transmission. A video storage solution is configured in the monitoring system to ensure that the real-time video stream is stored locally or in the cloud for subsequent analysis and review. Video management software (such as NVR) is used to manage storage. Use a video processing library (such as OpenCV) to read real-time video streams to ensure that video data can be stably received and processed. Determine the sampling interval of image frames, such as extracting 1 frame per second or 1 frame every 5 seconds. The specific interval can be adjusted according to actual needs and system performance. The sampling strategy should balance the number of images and storage requirements. Image frames should be extracted from the video stream according to the set sampling interval. Each extracted image should be saved in a standard format (such as JPEG or PNG) for subsequent processing. The extracted image frames should be stored in a specified directory and named by timestamp or serial number to form a continuous image frame sequence. These frame sequences will serve as the basis for subsequent analysis. Establish an image frame management mechanism to ensure the data integrity and accessibility of the frame sequence, and use a database or file system for management.

[0031] Step S3: performing a three-dimensional topological structure layout analysis on the monitoring image frame sequence to obtain the three-dimensional topological structure layout data of the elevator; and constructing a three-dimensional topological geometric model of the elevator based on the three-dimensional topological structure layout data of the elevator; In this embodiment, the monitoring image frame sequence is preprocessed, including denoising, contrast enhancement and brightness adjustment, to improve the image quality, which is achieved through image processing technology (such as filtering, histogram equalization), and a feature point detection algorithm (such as SIFT, SURF, ORB) is used in each image frame to extract key feature points. These feature points will be used for subsequent matching and construction of a three-dimensional structure, and feature point matching is performed between different image frames. A matching algorithm (such as BFMatcher or FLANN) is used to identify the same feature points. This step uses the RANSAC algorithm to eliminate mismatches, and three-dimensional reconstruction is performed using the matched feature points. The three-dimensional coordinates of each feature point are calculated using a triangulation method or a structured light method. Combined with the internal and external parameters of the camera, the three-dimensional point cloud data of the elevator is generated. Based on the three-dimensional point cloud data, the topological structure of the elevator is analyzed to determine the connection relationship between the feature points, and the topological structure is performed using a graph theory method (such as Delaunay triangulation). The definition of , organizes the obtained elevator 3D topological structure layout data to form a structured data set, records the spatial position of each feature point and their mutual relationship, selects appropriate 3D modeling tools or software (such as Blender, Maya, Open3D) to construct the geometric model, creates the 3D geometric model of the elevator according to the organized 3D topological structure layout data, and uses the point cloud data to generate a mesh model, which usually includes the following steps: converting the point cloud data into a triangular mesh, smoothing the mesh, optimizing the visualization effect of the model, adding the geometric features of the elevator to the model, such as the car, door, motor and other components, to ensure the integrity and accuracy of the model, optimizing the generated 3D model, ensuring that the number of polygons of the model is moderate, so as to facilitate subsequent rendering and analysis, processing the model through subdivision, simplification and other methods, and saving the constructed elevator 3D topological geometry model in an appropriate format (such as OBJ, STL, FBX) for subsequent use and analysis.

[0032] Step S4: Analyze the wear state of components on the three-dimensional topological geometric model of the elevator, and analyze potential structural defects, and extract potential topological structural defect data; In this embodiment, each key component (such as the car, guide rail, motor, etc.) is identified in the three-dimensional model, and a unique identifier is assigned to each component for subsequent analysis. Characteristic indicators related to component wear, such as wear depth, surface roughness, crack length, etc., are defined. These characteristics are evaluated by standardized detection methods (such as visual detection, laser ranging). The operation data of the elevator, such as load history, operation frequency, maintenance records, etc., are collected, and analyzed in combination with the above wear characteristics. A wear state model is established using statistical analysis or machine learning methods (such as regression analysis). The established model is applied to evaluate the wear state of each component, calculate the degree of wear, and give a corresponding wear grade (such as normal, slight wear, moderate wear, severe wear). The wear state analysis results are sorted out to form a Report, record the wear status, evaluation indicators and recommended maintenance measures of each component, determine the identification criteria for potential structural defects, including the degree of wear, fatigue life of the material used, historical failure data, etc., analyze the relationship between the wear status and potential defects, identify the structural defects caused by severely worn components, use association rule mining or decision tree analysis to achieve this, build a probability model (such as Bayesian network), evaluate the relationship between the wear of each component and potential structural defects, calculate the probability of defects, extract potential topological structural defect data, record the defect type, location and its potential impact, identify high-risk components, and generate a defect risk map, organize the analysis results, form a report on potential structural defects, and recommend corresponding maintenance and repair measures to reduce the risk of failures.

[0033] Step S5: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating an elevator fault diffusion trajectory; In this embodiment, the temperature and humidity time series state characteristic data of the elevator are collected to ensure the integrity and accuracy of the data. These data are obtained through real-time monitoring by sensors and organized into a time series form. A suitable defect evolution model (such as a finite element analysis model, an agent-based model, or a physics-based model) is selected to simulate the evolution of the defect state. These models can reflect the impact of temperature and humidity changes on the defect state. The parameters in the model are set according to the actual situation, including the range of temperature and humidity changes, the defect development rate, the physical properties of the material, etc. The evolution model is run, and the temperature and humidity time series data are input to simulate the state changes of the defects at different time points. This process is realized by numerical calculation methods (such as finite difference and Monte Carlo simulation). The simulation results are analyzed to extract key state change data in the defect evolution process, such as defect expansion and shape change, to form a defect state evolution model. The data set is used to select a suitable fault diffusion model (such as a propagation model or a network model) to reflect the process of defects spreading from one component to other components. A graph-based method is used to model the relationship between components. The evolution data obtained from the defect state evolution simulation is input into the fault diffusion model as the starting condition of the simulation. The fault diffusion model is run to simulate the propagation process of the fault in the elevator system. Influencing factors such as component connectivity and fault diffusion rate are considered to generate fault states at different time points. The key nodes and trajectory data in the fault diffusion process are recorded to form a fault diffusion trajectory diagram. Visualization tools (such as Matplotlib and Plotly) are used to visualize the fault diffusion trajectory for easy analysis and understanding. The results of the fault diffusion trajectory generation are sorted out to form a report that records the path, impact range and consequences of the fault diffusion, providing support for fault warning and maintenance decision-making.

[0034] Step S6: Performing an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; generating an adaptive fault warning signal based on the adaptive fault decision strategy.

[0035] In this embodiment, fault diffusion trajectory data, including fault type, diffusion speed, impact range, etc., are collected and sorted to provide basic data for decision generation. Appropriate decision models (such as decision trees, random forests, or fuzzy logic systems) are selected to generate decision strategies based on the fault diffusion trajectory data. Decision rules are defined based on historical fault cases and professional knowledge. For example, when a fault spreads to a specific component, maintenance is triggered; based on the severity level of the fault, whether to shut down immediately is decided. The collected fault data is used to train the decision model, optimize the decision strategy, and evaluate the accuracy and reliability of the model through cross-validation. Based on the trained model and defined rules, adaptive fault decision strategies are generated. These strategies should be able to be adjusted according to real-time data changes to adapt to different fault scenarios and determine the fault. The standards and triggering conditions for fault warning, such as the spread of faults to a specific range and the severity of fault types, are established. A real-time monitoring system is established to continuously monitor the operating status of the elevator and the trajectory of fault spread. A data stream processing framework (such as Apache Kafka and Apache Flink) is used to process real-time data. According to the adaptive fault decision strategy and the set warning standards, the warning generation logic is designed. When the monitored status meets the warning conditions, a fault warning signal is immediately generated. The generated fault warning signal is sent to relevant personnel or maintenance teams through appropriate channels (such as SMS, email, system notification) to ensure timely response. A feedback mechanism is established to collect data after fault handling, evaluate the effectiveness of the warning signal, and continuously adjust and optimize the fault decision strategy and warning logic based on the feedback results.

[0036] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: obtaining real-time multi-dimensional operating status parameters of the elevator based on sensors; Step S12: performing time synchronization processing on the multi-dimensional operating status parameters to obtain the operating status parameters with consistent timestamps; Step S13: Perform temperature and humidity characteristic analysis on the time-stamp consistent operating status parameters to extract real-time temperature and humidity characteristic data; Step S14: Calculate the change amplitude of multiple time points on the real-time temperature and humidity characteristic data to generate temperature and humidity change amplitude data; Step S15: mining the time series variation characteristics of the temperature and humidity variation amplitude data to obtain the time series state characteristics of the elevator temperature and humidity.

[0037] In this embodiment, a variety of sensors are installed in the elevator, including temperature sensors, humidity sensors, accelerometers, speed sensors, etc., to obtain multi-dimensional state parameters of the elevator operation. Use a data acquisition module (such as Arduino, RaspberryPi or a dedicated data acquisition device) to read the sensor data regularly and store it as a data record. Each record should contain parameters such as timestamp, temperature, humidity, speed, etc. The acquired multi-dimensional operating status parameters are organized into a structured format (such as JSON, CSV) to ensure that the data is easy to process later. Ensure that the timestamps of all sensor data are consistent. Use a clock synchronization protocol (such as NTP) or use a unified time base when collecting data. Merge the data from different sensors to ensure that each data record has the same timestamp. If there is missing data, interpolation can be used to fill it.

[0038] importpandas aspd #Assume the sensor data is stored in a CSV file temperature_data=pd.read_csv('temperature_data.csv') humidity_data=pd.read_csv('humidity_data.csv') #Merge data combined_data=pd.merge_asof(temperature_data, humidity_data, on='timestamp') Extract temperature and humidity features from the time-synchronized data. Calculate statistical features such as the mean, standard deviation, maximum, and minimum of temperature and humidity.

[0039] temperature_features=combined_data['temperature'].describe() humidity_features=combined_data['humidity'].describe() Calculate the magnitude of temperature and humidity changes between different time points, using both absolute and relative changes.

[0040] combined_data['temperature_change']=combined_data['temperature'].diff().fillna(0) combined_data['humidity_change']=combined_data['humidity'].diff().fillna(0) The calculated temperature and humidity change amplitudes are organized into a new data frame, including timestamp, temperature change amplitude, and humidity change amplitude. The temperature and humidity change amplitude data is mined for time series features. Time series analysis methods such as autocorrelation analysis and Fourier transform are used to identify periodicity and trends. Time series change features are organized into a report, including the statistical characteristics of the change amplitude and the periodicity analysis results. Use visualization tools such as Matplotlib to display the time series graph of temperature and humidity changes.

[0041] importmatplotlib.pyplotasplt plt.figure(figsize=(12, 6)) plt.plot(combined_data['timestamp'], combined_data['temperature_change'], label='TemperatureChange') plt.plot(combined_data['timestamp'], combined_data['humidity_change'], label='HumidityChange') plt.xlabel('Time') plt.ylabel('ChangeAmplitude') plt.title('TemperatureandHumidityChangeOverTime') plt.legend() plt.grid() plt.show() In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: Obtain multi-directional elevator monitoring video stream; Step S22: performing video interval sampling on the multi-directional elevator monitoring video stream to obtain multiple single monitoring image frames; Step S23: reconstructing the details of the multiple single surveillance image frames frame by frame to obtain multiple detail reconstructed image frames; Step S24: performing image-by-image temporal coding processing on the plurality of detail reconstructed image frames to obtain a code for each image frame; Step S25: performing pixel size standardization processing on a plurality of detail reconstructed image frames to obtain a plurality of standardized detail reconstructed image frames; Step S26: sequentially encode and sequence a plurality of standardized detail reconstructed image frames according to the encoding of each image frame to construct a monitoring image frame sequence.

[0042] In this embodiment, multiple surveillance cameras are installed inside and outside the elevator to ensure that all important viewing angles (such as elevator doors, inside the elevator car, control panels, etc.) are covered, camera settings are configured, including resolution, frame rate, and video encoding format (such as H.264, MJPEG), to ensure the quality and stability of the video stream, the video stream is transmitted to a central processing unit or a storage device through a network protocol (such as RTSP, RTMP) to ensure real-time monitoring and storage, a sampling interval (such as sampling once per second or every 5 frames) is determined to extract image frames from the video stream, a suitable video processing tool or library (such as OpenCV, FFmpeg) is selected to achieve video frame extraction, the video stream is processed using the selected tool, image frames are extracted according to the set sampling interval, multiple single surveillance image frames are generated, the extracted image frames are sequentially saved as image files (such as JPEG, PNG), to ensure accessibility for subsequent processing, a suitable image reconstruction algorithm (such as super-resolution reconstruction, image denoising, etc.) is selected to improve the details and clarity of the image, the reconstruction algorithm is applied to each extracted image frame, and frame by frame processing is performed to enhance the image details, and each extracted image frame is converted into a single image file. The processed detail reconstructed image frames are saved as new image files for subsequent analysis. A suitable image encoding algorithm (such as JPEG, H.264, VP9) is selected to realize the encoding of the image frames. The selected encoding algorithm is applied to each detail reconstructed image frame. The encoding process is performed image by image. The encoding data of each encoded image frame is saved for subsequent use and analysis. A standardized image size (such as 1920x1080, 1280x720) is determined to ensure image consistency. Image processing tools or libraries (such as OpenCV, PIL) are used to resize the images. The resizing process is applied to each detail reconstructed image frame to generate standardized image frames. The standardized image frames are saved as new image files to ensure uniform format and quality. A suitable serialization format (such as MP4, AVI, GIF) is determined to store the image frame sequence. The standardized image frames are sequentially encoded into a video file using a video processing library (such as OpenCV, FFmpeg). A surveillance image frame sequence is constructed. The generated surveillance image frame sequence is saved as a video file to ensure its playability and storage accessibility.

[0043] Specific implementation steps code example: importcv2 video_capture=cv2.VideoCapture('video_stream.mp4') frame_rate=5#sample one frame every 5 seconds frames=[] while1: ret,frame=video_capture.read() ifnotret: break current_time = video_capture.get(cv2.CAP_PROP_POS_MSEC) / 1000#Current time (seconds) ifcurrent_time%frame_rate==0: frames.append(frame) video_capture.release() defdetail_reconstruction(frame): #Use super-resolution or denoising algorithms returnenhanced_frame#Return the reconstructed image detailed_frames=[detail_reconstruction(frame)forframeinframes] In this embodiment, the specific steps of step S23 are: Extracting a first image frame based on a plurality of single surveillance image frames; Performing brightness enhancement processing on the first image frame to obtain a brightness enhanced image frame; Perform forward multi-layer convolution calculation on the brightness enhanced image frame to extract the low-level texture features of the image frame; Nonlinearly compress low-level texture features of the image frame to obtain mid-level features of the image frame; Perform deep local feature extraction on the mid-level features of the image frame to obtain spatial local convolution features; Perform temporal convolution arrangement on the low-level texture features of the image frame, the mid-level features of the image frame and the spatial local convolution features to obtain a multi-channel convolution feature sequence; Performing convolution detail reconstruction on the first image frame according to the multi-channel convolution feature sequence to obtain a first detail reconstructed image frame; The convolution detail reconstruction process is repeated on the multiple single monitoring image frames until all the single monitoring image frames are processed, thereby obtaining multiple detail reconstructed image frames.

[0044] In this embodiment, the first frame is extracted from multiple single surveillance image frames. The frames are stored in a list and the first element is directly accessed. A suitable brightness enhancement algorithm is selected, such as histogram equalization or gamma correction. The first frame is brightness enhanced using OpenCV or a similar library.

[0045] importcv2 enhanced_frame = cv2.convertScaleAbs(first_frame, alpha=1.5, beta=0) #brightness enhancement Design a multi-layer convolutional neural network (CNN), including multiple convolutional layers and activation layers, to extract low-level texture features. Use a deep learning framework such as TensorFlow or PyTorch to build a convolutional model and perform a forward pass on the brightness-enhanced image to extract features.

[0046] importtorch importtorch.nnasnn classSimpleCNN(nn.Module): def __init__(self): super(SimpleCNN, self).__init__() self.conv1=nn.Conv2d(3, 16, kernel_size=3, padding=1) self.relu = nn.ReLU() self.conv2=nn.Conv2d(16, 32, kernel_size=3, padding=1) def forward(self, x): x = self.relu(self.conv1(x)) x = self.relu(self.conv2(x)) returnx model=SimpleCNN() input_tensor = torch.from_numpy(enhanced_frame).permute(2, 0, 1).unsqueeze(0).float() #Convert to tensor low_level_features=model(input_tensor)#Extract low-level texture features Choose nonlinear compression techniques, such as autoencoders or pooling operations, to reduce the low-level feature dimensions. Apply pooling layers or autoencoders to compress low-level texture features to obtain mid-level features.

[0047] middle_level_features = nn.MaxPool2d(2)(low_level_features) #Use maximum pooling for compression A local convolutional layer is designed to extract spatial local features. Further convolutional layers are added to the network to extract deep local features.

[0048] conv3=nn.Conv2d(32, 64, kernel_size=3, padding=1) deep_local_features=conv3(middle_level_features)#Extract deep local features The low-level texture features, mid-level features, and deep-level features are temporally convolved to form a multi-channel feature sequence. The features are stacked together using appropriate tensor operations.

[0049] multi_channel_features = torch.cat((low_level_features, middle_level_features, deep_local_features), dim=1) # concatenate in the channel dimension Design a deconvolution or transposed convolution network to reconstruct the details of the first frame image. Use a deep learning framework for back propagation to generate a detailed reconstructed image of the first frame.

[0050] classReconstructionNetwork(nn.Module): def __init__(self): super(ReconstructionNetwork, self).__init__() self.deconv1=nn.ConvTranspose2d(64, 32, kernel_size=3, padding=1) self.deconv2=nn.ConvTranspose2d(32, 3, kernel_size=3, padding=1) def forward(self, x): x = self.relu(self.deconv1(x)) return self.deconv2(x) reconstruction_model=ReconstructionNetwork() reconstructed_frame=reconstruction_model(multi_channel_features)#Details reconstruction The above processing steps are encapsulated in a loop and repeated for all single surveillance image frames.

[0051] detailed_reconstructed_frames=[] forframeinframes: #Repeat steps 1 to 7 #Extraction, enhancement, convolution, feature extraction, reconstruction... detailed_reconstructed_frames.append(reconstructed_frame) In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Step S31: identifying structural feature points of the monitoring image frame sequence and marking the three-dimensional structural feature points; Step S32: performing multi-angle feature matching on the three-dimensional structural feature points, thereby obtaining spatial matching data of the structural feature points; Step S33: Calculating the spatial position of the feature points based on the spatial matching data of the structural feature points to generate the spatial position coordinates of the structural feature points; Step S34: performing a three-dimensional topological structure layout analysis on the spatial position coordinates of the structural feature points to obtain three-dimensional topological structure layout data of the elevator; Step S35: performing key connection detection on the three-dimensional structural feature points to obtain structural connection points; Step S36: Based on the structural connection points, three-dimensional topological connection modeling is performed on the elevator three-dimensional topological structure layout data to construct a three-dimensional topological geometric model of the elevator.

[0052] In this embodiment, descriptors are generated for the identified feature points. These descriptors are numerical representations of the feature points and can be used to match the same feature points in different image frames. Feature matching is performed between different image frames to identify the same feature points. This is usually achieved by calculating the distance (similarity) between descriptors, selecting the most matching feature point pair, filtering unreliable matching results by setting a threshold, and retaining only high-quality matching pairs to improve the accuracy of subsequent calculations. The three-dimensional spatial coordinates of each feature point are calculated using the matched feature points and the internal and external parameters of the camera, using triangulation or other three-dimensional reconstruction techniques. The generated three-dimensional coordinates provide basic data for subsequent topological analysis, ensuring the precise position of each feature point in space. According to the three-dimensional coordinates of the feature points, the topological structure of the elevator is defined, including nodes (feature points) and edges (connections between feature points). Use appropriate analysis methods (such as Delaunay triangulation) to perform layout analysis on feature points, identify the relationship and connection method between each feature point, and form the three-dimensional topological structure of the elevator. Use detection methods based on distance, angle or other standards to identify the key connections between feature points. These connections usually reflect the structural characteristics of the elevator. Record the detected key connection points to provide a basis for subsequent modeling and analysis, ensure the accuracy and completeness of the connection points, use appropriate three-dimensional modeling tools or software, prepare for stereo topological connection modeling, and construct a three-dimensional geometric model of the elevator based on the structural connection point information and three-dimensional topological layout data. The model should accurately reflect the spatial structure and characteristics of the elevator. Optimize the generated three-dimensional model to ensure the smoothness and visual effect of the model, and save it in a suitable format for subsequent use or display.

[0053] In this embodiment, step S4 includes the following steps: Step S41: performing surface optical flow fluctuation analysis on the three-dimensional topological geometric model of the elevator to obtain the elevator surface structure fluctuation data; Step S42: positioning elevator components based on the elevator surface structure fluctuation data to obtain component nodes at defective locations; Step S43: Analyze the component wear status of the component node at the defective position to generate component wear status data; Step S44: Perform potential structural defect analysis on the component wear state data to extract potential topological structural defect data.

[0054] In this embodiment, a suitable optical flow calculation algorithm (such as the Lucas-Kanade method or the Horn-Schunck method) is selected to analyze the movement and changes of the model surface, and the three-dimensional topological model of the elevator is converted into a format suitable for optical flow analysis to ensure that the surface details and texture of the model are preserved. The surface of the model is subjected to optical flow analysis to extract the motion characteristics of the surface layer and identify the surface fluctuation changes. This is achieved by analyzing the motion vectors of the surface points. By analyzing the fluctuation data of the surface structure, the defective area is identified, which includes finding abnormal fluctuation patterns or fluctuations. Machine learning or pattern recognition algorithms are used in combination with the design drawings of the elevator to determine the position of each component in the three-dimensional model, determine the component nodes where the defects are located, and record the spatial coordinates of these nodes for later use. Provide a basis for subsequent wear status analysis, determine the standards and indicators for wear status evaluation, including wear degree, wear type (such as surface scratches, cracks, etc.), collect relevant data on defect locations (such as historical operating time, load conditions), and analyze them in combination with fluctuation data to evaluate the wear status of components, generate component wear status data based on the analysis results, record the wear level and potential risks of each component, select appropriate defect analysis methods (such as risk assessment model, fault tree analysis) to identify potential structural defects, use component wear status data to analyze the factors that lead to structural defects, identify high-risk components and their connection relationships, extract potential topological structure defect data, and form a detailed defect report, including defect type, location, severity level and other information, to provide guidance for subsequent maintenance and repair.

[0055] In this embodiment, step S5 includes the following steps: Step S51: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics to generate topological structure defect evolution data; Step S52: performing deep defect evolution learning on topological structure defect evolution data to obtain topological defect evolution rules; Step S53: predicting the elevator failure probability based on the topological defect evolution law, thereby obtaining the elevator failure probability; Step S54: Evolving the fault diffusion trajectory based on the elevator failure probability, thereby generating the elevator fault diffusion trajectory.

[0056] In this embodiment, the temperature and humidity time series state data of the elevator are collected, and the time consistency and integrity of the data are ensured. These data are obtained by real-time monitoring by sensors. Select a suitable defect evolution simulation model (such as a finite element model, a state transition model) to simulate the evolution process of defects under different temperature and humidity conditions. Input the temperature and humidity data into the selected evolution model to simulate the defect state. Through simulation calculation, the evolution data of topological structure defects at different time points are generated. The generated topological structure defect evolution data is sorted into a format suitable for model training to ensure that the characteristics and labels of the data are clear. Select a suitable deep learning model (such as LSTM, convolutional neural network) to learn the time series characteristics of defect evolution. Use the prepared evolution data to train the deep learning model and adjust the model parameters to optimize the learning effect. The accuracy of the model is evaluated by cross-validation technology during the training process. Extract the evolution law of topological defects from the trained model, and identify the important factors affecting the evolution of defects and their interrelationships. Based on the evolution law of topological defects, a fault probability prediction model is constructed. It is a statistical model (such as logistic regression) or a machine learning model (such as random forest, support vector machine). The defect evolution law and related data are input into the fault probability model to predict the fault probability. Evaluate the accuracy and reliability of the prediction results and verify the effectiveness of the model by comparing with historical fault data. Select a suitable fault diffusion model (such as a propagation model, a network model) to simulate the fault diffusion process in the elevator system. According to the predicted fault probability, simulate the fault diffusion trajectory in the elevator system. This requires considering the connection relationship between the various components and the path of fault propagation. Visualize the generated fault diffusion trajectory to show the impact of the fault on the elevator system and the propagation path, so as to facilitate subsequent maintenance and risk assessment.

[0057] Specific implementation steps code example: defsimulate_defect_evolution(temperature_data, humidity_data, initial_defect_data): evolution_data=[] #Simulation logic fortinrange(len(temperature_data)): #Defect changes according to temperature and humidity current_state=initial_defect_data#Update defect status evolution_data.append(current_state) returnevolution_data defect_evolution_data=simulate_defect_evolution(temperature_data, humidity_data, potential_defect_data) The generated defect evolution data is used to train the model to extract the patterns and laws of defect evolution.

[0058] importtensorflowastf model = tf.keras.Sequential([ tf.keras.layers.LSTM(64, return_sequences=1, input_shape=(timesteps, features)), tf.keras.layers.Dense(1) ]) model.compile(optimizer='adam', loss='mean_squared_error') model.fit(defect_evolution_data, epochs=100)#training model Based on the learned evolution rules, the failure probability of the elevator under specific conditions is predicted.

[0059] fromsklearn.ensembleimportRandomForestClassifier rf_model=RandomForestClassifier() rf_model.fit(X_train, y_train) #X_train is the feature, y_train is the label fault_probabilities = rf_model.predict_proba(X_test)[:, 1] # predict fault probability Based on the predicted failure probability, an elevator fault diffusion trajectory is generated to record how the failure affects different components of the elevator.

[0060] defsimulate_fault_spread(fault_probabilities): spread_trajectory=[] forprobabilityinfault_probabilities: #Logic: Generate diffusion trajectory based on probability spread = np.random.choice(['normal', 'fault'], p=[1-probability, probability]) spread_trajectory.append(spread) returnspread_trajectory fault_spread_trajectory=simulate_fault_spread(fault_probabilities) In this embodiment, step S6 includes the following steps: Step S61: performing fault type identification on the elevator fault diffusion trajectory to generate the elevator fault type; Step S62: Perform safety risk quantitative analysis on the elevator fault diffusion trajectory to obtain the elevator fault safety risk level; Step S63: performing an adaptive fault decision based on the elevator fault type and the elevator fault safety risk level to generate an adaptive fault decision strategy; Step S64: Perform fault warning based on the adaptive fault decision strategy and generate an adaptive fault warning signal.

[0061] In this embodiment, key features are extracted from the fault diffusion trajectory, such as the time, location, and impact range of the fault, so as to facilitate subsequent analysis, and a suitable classification algorithm (such as decision tree, random forest, or support vector machine) is selected to model the fault type. The selected model is trained and verified using historical fault data to ensure that the model can accurately identify different types of faults. After the training is completed, the model is used to identify the type of the current fault diffusion trajectory, generate the elevator fault type, determine the standards and indicators for safety risk assessment, such as the frequency of faults, the severity of the impact, the potential consequences, etc., select a suitable risk analysis method (such as failure mode and effects analysis FMEA, quantitative risk assessment), conduct a detailed analysis of the fault diffusion trajectory, quantify the safety risk according to the set standards, give the risk level of each fault type, and Based on the fault type and safety risk level, an adaptive decision model is constructed. Based on the rule engine, decision tree or machine learning algorithm, the decision logic is designed. The response strategies and action plans under different fault types and risk levels are clarified. The decision model is tested and optimized to ensure its effectiveness and accuracy in different scenarios. An adaptive fault decision strategy is generated according to the model output to provide guidance for subsequent fault handling. The standards for fault warning are determined, including the triggering conditions, warning levels and corresponding response measures of the warning. A fault warning system is designed. The monitoring data and decision-making strategy are combined to realize the automated warning function. When the fault type and risk level meet the warning conditions, the system automatically generates a warning signal and sends it to relevant personnel or systems. An early warning feedback mechanism is established to facilitate timely adjustment of decision strategies and response measures to ensure the flexibility and effectiveness of the system.

[0062] In this embodiment, a deep learning-based intelligent elevator safety fault warning system is provided, which is used to execute the deep learning-based intelligent elevator safety fault warning method as described above, including: The time series status module is used to obtain the real-time multi-dimensional operating status parameters of the elevator; the time series change characteristics of the multi-dimensional operating status parameters are mined to obtain the time series status characteristics of the elevator temperature and humidity; An image processing module is used to obtain a multi-directional elevator monitoring video stream; perform video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; A topology structure module is used to perform a three-dimensional topology structure layout analysis on a monitoring image frame sequence to obtain the three-dimensional topology structure layout data of the elevator; and to construct a three-dimensional topology geometric model of the elevator based on the three-dimensional topology structure layout data of the elevator; The structural defect module is used to analyze the wear status of components of the elevator's three-dimensional topological geometric model, analyze potential structural defects, and extract potential topological structural defect data; The fault prediction module is used to simulate the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating the elevator fault diffusion trajectory; The fault warning module is used to make an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; and generate an adaptive fault warning signal based on the adaptive fault decision strategy.

[0063] The present invention can realize real-time monitoring of the elevator status by acquiring the real-time multi-dimensional operating status parameters of the elevator. By mining the time series change characteristics of the multi-dimensional operating status parameters, the time series status characteristics of the elevator temperature and humidity are obtained, which is helpful to identify abnormal states and trends, obtain multi-directional elevator monitoring video streams, provide a full range of monitoring vision, and construct monitoring image frame sequences through video interval sampling, which is helpful to analyze the elevator operating status and detect abnormal situations. The monitoring image frame sequence is subjected to three-dimensional topological structure layout analysis to obtain the elevator three-dimensional topological structure layout data, which helps to understand the elevator structure layout and the positions of key components. The elevator three-dimensional topological geometric model is constructed based on the three-dimensional topological structure layout data, which provides a basis for subsequent analysis. Data, analyze the component wear status of the elevator's three-dimensional topological geometric model, extract potential structural defect data, which helps to identify existing structural problems, conduct potential structural defect analysis, help identify potential structural problems and prevent failures, simulate the defect state evolution of potential topological structural defect data according to the elevator's temperature and humidity time series state characteristics, generate the elevator fault diffusion trajectory, which helps to predict the path of the failure, make adaptive fault decisions on the elevator fault diffusion trajectory, generate adaptive fault decision strategies, help formulate targeted fault handling plans, generate adaptive fault warning signals based on the adaptive fault decision strategies, issue alarms in time and take preventive measures to improve the safety and reliability of elevators.

[0064] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0065] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A smart elevator safety fault warning method based on deep learning, characterized in that: The following steps are involved: Step S1: Obtaining real-time multi-dimensional operating status parameters of the elevator; Mining the time series variation characteristics of the multi-dimensional operating state parameters to obtain the time series state characteristics of elevator temperature and humidity; Step S2: obtaining a multi-directional elevator monitoring video stream; performing video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; Step S3: performing a three-dimensional topological structure layout analysis on the monitoring image frame sequence to obtain the three-dimensional topological structure layout data of the elevator; Constructing a 3D topological geometric model of the elevator based on the 3D topological structure layout data of the elevator; Step S4: Analyze the wear state of components on the three-dimensional topological geometric model of the elevator, and analyze potential structural defects, and extract potential topological structural defect data; Step S5: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating an elevator fault diffusion trajectory; Step S6: performing adaptive fault decision-making on the elevator fault diffusion trajectory to generate an adaptive fault decision-making strategy; Generate adaptive fault warning signals based on adaptive fault decision strategies.

2. The intelligent elevator safety fault early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: obtaining real-time multi-dimensional operating status parameters of the elevator based on sensors; Step S12: performing time synchronization processing on the multi-dimensional operating status parameters to obtain the operating status parameters with consistent timestamps; Step S13: Perform temperature and humidity characteristic analysis on the time-stamp consistent operating status parameters to extract real-time temperature and humidity characteristic data; Step S14: Calculate the change amplitude of multiple time points on the real-time temperature and humidity characteristic data to generate temperature and humidity change amplitude data; Step S15: mining the time series variation characteristics of the temperature and humidity variation amplitude data to obtain the time series state characteristics of the elevator temperature and humidity.

3. The intelligent elevator safety failure early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S2 are: Step S21: Obtain multi-directional elevator monitoring video stream; Step S22: performing video interval sampling on the multi-directional elevator monitoring video stream to obtain multiple single monitoring image frames; Step S23: reconstructing the details of the multiple single surveillance image frames frame by frame to obtain multiple detail reconstructed image frames; Step S24: performing image-by-image temporal coding processing on the plurality of detail reconstructed image frames to obtain a code for each image frame; Step S25: performing pixel size standardization processing on a plurality of detail reconstructed image frames to obtain a plurality of standardized detail reconstructed image frames; Step S26: sequentially encode and sequence a plurality of standardized detail reconstructed image frames according to the encoding of each image frame to construct a monitoring image frame sequence.

4. The intelligent elevator safety failure early warning method based on deep learning according to claim 3 is characterized in that: The specific steps of step S23 are: Extracting a first image frame based on a plurality of single surveillance image frames; Performing brightness enhancement processing on the first image frame to obtain a brightness enhanced image frame; Perform forward multi-layer convolution calculation on the brightness enhanced image frame to extract the low-level texture features of the image frame; Nonlinearly compress low-level texture features of the image frame to obtain mid-level features of the image frame; Perform deep local feature extraction on the mid-level features of the image frame to obtain spatial local convolution features; Perform temporal convolution arrangement on the low-level texture features of the image frame, the mid-level features of the image frame and the spatial local convolution features to obtain a multi-channel convolution feature sequence; Performing convolution detail reconstruction on the first image frame according to the multi-channel convolution feature sequence to obtain a first detail reconstructed image frame; The convolution detail reconstruction process is repeated on the multiple single monitoring image frames until all the single monitoring image frames are processed, thereby obtaining multiple detail reconstructed image frames.

5. The intelligent elevator safety fault early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S3 are: Step S31: identifying structural feature points of the monitoring image frame sequence and marking the three-dimensional structural feature points; Step S32: performing multi-angle feature matching on the three-dimensional structural feature points, thereby obtaining spatial matching data of the structural feature points; Step S33: Calculating the spatial position of the feature points based on the spatial matching data of the structural feature points to generate the spatial position coordinates of the structural feature points; Step S34: performing a three-dimensional topological structure layout analysis on the spatial position coordinates of the structural feature points to obtain three-dimensional topological structure layout data of the elevator; Step S35: performing key connection detection on the three-dimensional structural feature points to obtain structural connection points; Step S36: Based on the structural connection points, three-dimensional topological connection modeling is performed on the elevator three-dimensional topological structure layout data to construct a three-dimensional topological geometric model of the elevator.

6. The intelligent elevator safety fault early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: performing surface optical flow fluctuation analysis on the three-dimensional topological geometric model of the elevator to obtain the elevator surface structure fluctuation data; Step S42: positioning elevator components based on the elevator surface structure fluctuation data to obtain component nodes at defective locations; Step S43: Analyze the component wear status of the component node at the defective position to generate component wear status data; Step S44: Perform potential structural defect analysis on the component wear state data to extract potential topological structural defect data.

7. The intelligent elevator safety failure early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: simulating the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics to generate topological structure defect evolution data; Step S52: performing deep defect evolution learning on topological structure defect evolution data to obtain topological defect evolution rules; Step S53: predicting the elevator failure probability based on the topological defect evolution law, thereby obtaining the elevator failure probability; Step S54: Evolving the fault diffusion trajectory based on the elevator failure probability, thereby generating the elevator fault diffusion trajectory.

8. The intelligent elevator safety failure early warning method based on deep learning according to claim 1 is characterized in that: The specific steps of step S6 are: Step S61: performing fault type identification on the elevator fault diffusion trajectory to generate the elevator fault type; Step S62: Perform safety risk quantitative analysis on the elevator fault diffusion trajectory to obtain the elevator fault safety risk level; Step S63: performing an adaptive fault decision based on the elevator fault type and the elevator fault safety risk level to generate an adaptive fault decision strategy; Step S64: Perform fault warning based on the adaptive fault decision strategy and generate an adaptive fault warning signal.

9. An intelligent elevator safety fault warning system based on deep learning, characterized in that: The method for executing the intelligent elevator safety fault early warning method based on deep learning as claimed in claim 1 comprises: The time series status module is used to obtain the real-time multi-dimensional operation status parameters of the elevator; the time series change characteristics of the multi-dimensional operation status parameters are mined to obtain the time series status characteristics of the elevator temperature and humidity; An image processing module is used to obtain a multi-directional elevator monitoring video stream; perform video interval sampling on the multi-directional elevator monitoring video stream to construct a monitoring image frame sequence; A topology structure module is used to perform a three-dimensional topology structure layout analysis on a monitoring image frame sequence to obtain the three-dimensional topology structure layout data of the elevator; and to construct a three-dimensional topology geometric model of the elevator based on the three-dimensional topology structure layout data of the elevator; The structural defect module is used to analyze the wear status of components of the elevator's three-dimensional topological geometric model, analyze potential structural defects, and extract potential topological structural defect data; The fault prediction module is used to simulate the defect state evolution of potential topological structure defect data according to the elevator temperature and humidity time series state characteristics, thereby generating the elevator fault diffusion trajectory; The fault warning module is used to make an adaptive fault decision on the elevator fault diffusion trajectory to generate an adaptive fault decision strategy; and generate an adaptive fault warning signal based on the adaptive fault decision strategy.

Images