Elevator prediction maintenance and fault early warning method and system
By constructing a mechanistic model and a parallel feature extraction network, the health status of elevators is analyzed, and fault warning levels and causal chains are calculated. This solves the problem of lag in traditional elevator predictive maintenance, realizes proactive predictive maintenance, and improves elevator operation and maintenance efficiency and safety.
Patent Information
- Application Number
- CN202511257231.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-04
- Publication Date
- 2025-11-25
AI Technical Summary
Traditional elevator predictive maintenance relies on regular inspections and experience-based judgment, which cannot proactively detect the early signs of malfunctions, resulting in delayed warnings and an inability to accurately pinpoint the root cause, thus failing to achieve proactive predictive maintenance.
By acquiring elevator operation status data and historical logs, a mechanism model is constructed to generate virtual sensor data. Health features are extracted using a parallel feature extraction network, the health evolution trajectory is analyzed, the decline slope of health indicators is calculated, and combined with a fault early warning level mechanism, a health fault causal chain is automatically established to formulate targeted maintenance strategies.
It enables precise analysis of the real-time health status of elevators, dynamic assessment of performance degradation trends, significantly improves the level of intelligent operation and maintenance of elevators, reduces the failure rate and maintenance costs, extends equipment lifespan, and improves safety and reliability.
Smart Images
Figure CN121005328A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to an elevator predictive maintenance and fault early warning method and system, and belongs to the technical field of remote monitoring. BACKGROUND
[0002] Elevator predictive maintenance and fault early warning refers to a modern intelligent operation and maintenance method that uses advanced sensing technology, data analysis algorithms and artificial intelligence models to monitor and analyze the running state of an elevator in real time or quasi-real time, predict its future health status and possible faults, and issue an early warning before the fault occurs and recommend accurate maintenance strategies. Through elevator predictive maintenance and fault early warning, safety hazards caused by sudden faults can be eliminated, and unnecessary regular inspections and emergency maintenance costs can be reduced.
[0003] Traditional elevator predictive maintenance mainly relies on regular inspections and experience-based judgments, that is, technicians perform on-site inspections of elevators at fixed intervals (such as every month), assess the state by relying on subjective methods such as listening, touching and observing, and make rough warnings in combination with historical fault records. This scheme is essentially "planned maintenance" and cannot perceive sudden faults between two inspections, resulting in a lag in elevator predictive maintenance and fault early warning. SUMMARY
[0004] The application provides an elevator predictive maintenance and fault early warning method and system, which aims to improve the efficiency of elevator predictive maintenance and fault early warning.
[0005] To achieve the above-mentioned purpose, the application provides an elevator predictive maintenance and fault early warning method, which comprises: obtaining running state data and historical running logs of an elevator to be maintained to generate fusion data streams of the elevator to be maintained; constructing a mechanism model of the elevator to be maintained, analyzing the running state of the elevator to be maintained based on the fusion data streams, and analyzing virtual sensing data of the elevator to be maintained using the mechanism model according to the running state; extracting health features of the elevator to be maintained using a preset parallel feature extraction network based on the virtual sensing data to analyze the real-time health status of the elevator to be maintained; analyzing the health evolution trajectory of the elevator to be maintained according to the real-time health status to calculate the health degree index decline slope of the elevator to be maintained, and determining the fault early warning level of the elevator to be maintained according to the health degree index decline slope; when the fault early warning level exceeds a preset fault early warning level threshold, establishing a health fault cause-effect chain of the elevator to be maintained to determine the maintenance strategy of the elevator to be maintained.
[0006] Optionally, the constructing the mechanism model of the elevator to be maintained comprises: establishing an elevator data model of the elevator to be maintained, wherein the elevator data model comprises a traction system model, a suspension and guide system model, a door system model, and a control system model; defining a coupling interface of the elevator data model; integrating the elevator data model through the coupling interface to obtain an integrated elevator data model; performing parameter calibration on the integrated elevator data model to obtain a calibrated elevator data model; defining a virtual sensor of the calibrated elevator data model; integrating the virtual sensor into the calibrated elevator data model to obtain the mechanism model of the elevator to be maintained.
[0007] Optionally, the establishing the elevator data model of the elevator to be maintained comprises: establishing a three-phase asynchronous motor dynamic equation and a torque balance equation of the elevator to be maintained; establishing a traction system model of the elevator to be maintained in combination with the three-phase asynchronous motor dynamic equation and the torque balance equation; establishing a mass-spring-damper system of the elevator to be maintained to establish an external excitation force equation of the elevator to be maintained; generating a suspension and guide system model of the elevator to be maintained according to the external excitation force equation; establishing a simulated door opening and closing curve of the elevator to be maintained to determine a door system model of the elevator to be maintained; analyzing a speed curve, a PID controller, and a state machine of the elevator to be maintained; establishing a control system model of the elevator to be maintained through the speed curve, the PID controller, and the state machine.
[0008] establishing the elevator data model of the elevator to be maintained in combination with the traction system model, the suspension and guide system model, the door system model, and the control system model.
[0009] Optionally, the further establishing the external excitation force equation of the elevator to be maintained comprises: defining an elevator mass, a displacement vector, an elevator stiffness, and an elevator damping of the elevator to be maintained according to a mass-spring-damper system of the elevator to be maintained; constructing the external excitation force equation of the elevator to be maintained in combination with the elevator mass, the displacement vector, the elevator stiffness, and the elevator damping, wherein the external excitation force equation:
[0010] wherein, represents an external excitation force vector of the elevator to be maintained, represents an elevator mass of the elevator to be maintained, represents an acceleration of the elevator to be maintained, represents a displacement vector, represents an elevator damping of the elevator to be maintained, represents a velocity of the elevator to be maintained, represents an elevator stiffness of the elevator to be maintained.
[0011] Optionally, the generating the fusion data stream of the elevator to be maintained comprises: cleaning the running state data and the historical running log of the elevator to be maintained respectively to obtain cleaned running state data and cleaned historical running log; labeling a timestamp of the cleaned running state data as a reference main time axis of the cleaned running state data and the cleaned historical running log; identifying discrete events and cumulative statistics of the cleaned historical running log; and converting the cumulative statistics into time point events; mapping the discrete events and the time point events to the reference main time axis to obtain the fusion data stream of the elevator to be maintained.
[0012] Optionally, the analyzing the virtual sensor data of the elevator to be maintained by using the mechanism model comprises: determining a target velocity, a feedback motor speed and a door machine switch instruction of the elevator to be maintained according to a running state of the elevator to be maintained; calculating an electromagnetic torque of the elevator to be maintained by using a control system model corresponding to the elevator to be maintained through the target velocity and the feedback motor speed; calculating a motor speed of the elevator to be maintained based on the electromagnetic torque by using a traction system model corresponding to the elevator to be maintained; converting the motor speed into a car velocity to calculate a car vertical vibration acceleration and a horizontal guiding force of the elevator to be maintained by using a suspension and guiding system model corresponding to the elevator to be maintained; analyzing a door machine motor torque and a door velocity of the elevator to be maintained based on the door machine switch instruction by using a door system model of the elevator to be maintained, taking the electromagnetic torque, the motor speed, the car velocity, the car vertical vibration acceleration, the horizontal guiding force, the door machine motor torque and the door velocity as the virtual sensor data of the elevator to be maintained.
[0013] Optionally, the health feature of the elevator to be maintained is extracted by using a preset parallel feature extraction network based on the virtual sensor data, including: Based on the virtual sensor data, the local mode of the elevator to be maintained is extracted by using a space-time feature extraction branch in the parallel feature extraction network; Based on the virtual sensor data, the long-term evolution state and periodic state of the elevator to be maintained are analyzed by using a time series feature extraction branch in the parallel feature extraction network; The local mode, the long-term evolution state and the periodic state are integrated into a joint feature vector; The element weight of the joint feature vector is defined; The health feature of the elevator to be maintained is generated by the element weight and the joint feature vector.
[0014] Optionally, the health evolution trajectory of the elevator to be maintained is analyzed according to the real-time health state, including: The supervised learning sequence of the elevator to be maintained is generated according to the real-time health state; The supervised learning sequence is encoded into a context vector by using an Encoder layer in a preset Seq2Seq model; The health sequence value of the elevator to be maintained is generated by using a Decoder layer in the Seq2Seq model based on the context vector; The health evolution trajectory of the elevator to be maintained is generated according to the health sequence value.
[0015] Optionally, when the fault warning level exceeds a preset fault warning level threshold, the health fault causal chain of the elevator to be maintained is established, including: The trigger time and time window when the fault warning level exceeds the preset fault warning level threshold are marked; The abnormal event of the time window is extracted as a causal chain node of the elevator to be maintained; The time-elevator state association graph of the abnormal event is established; The causal path and path weight of the elevator to be maintained are determined according to the time-elevator state association graph; The health fault causal chain of the elevator to be maintained is determined in combination with the causal chain node, the time-elevator state association graph, the causal path and the path weight.
[0016] In order to solve the above problems, the present application also provides an elevator predictive maintenance and fault warning system, the system comprising: A data stream fusion module is configured to obtain operation state data and historical operation logs of an elevator to be maintained to generate a fusion data stream of the elevator to be maintained; a virtual sensor data analysis module configured to construct a mechanism model of the elevator to be maintained, analyze an operation state of the elevator to be maintained based on the fused data stream, and analyze virtual sensor data of the elevator to be maintained using the mechanism model according to the operation state; a real-time health analysis module configured to extract health features of the elevator to be maintained using a preset parallel feature extraction network based on the virtual sensor data, and analyze a real-time health state of the elevator to be maintained; a warning level determination module configured to analyze a health evolution trajectory of the elevator to be maintained according to the real-time health state, calculate a health degree index decline slope of the elevator to be maintained, and determine a failure warning level of the elevator to be maintained according to the health degree index decline slope; a maintenance strategy generation module configured to establish a health-failure causal chain of the elevator to be maintained when the failure warning level exceeds a preset failure warning level threshold, and determine a maintenance strategy of the elevator to be maintained.
[0017] The scheme realizes accurate analysis of the real-time health state of the elevator by fusing the elevator operation state data and the historical log, constructing the mechanism model to generate the virtual sensor data, and extracting the health features using the parallel feature extraction network, calculates the health degree index decline slope based on the health evolution trajectory to dynamically evaluate the performance degradation trend of the elevator, effectively identifies the potential failure risk in combination with the failure warning level mechanism, automatically establishes the health-failure causal chain when the warning level exceeds the threshold to accurately locate the failure root cause and evolution path, and provides a scientific basis for formulating the targeted maintenance strategy. The scheme significantly improves the intelligent level of the elevator operation and maintenance, realizes the transition from passive maintenance to active predictive maintenance, greatly reduces the failure rate and maintenance cost, prolongs the service life of the equipment, improves the safety and reliability of the elevator operation, and provides efficient and reliable technical support for smart buildings and city elevator management. Therefore, the present application can improve the efficiency of elevator predictive maintenance and failure warning. BRIEF DESCRIPTION OF DRAWINGS
[0018] Figure 1 A flowchart of an elevator predictive maintenance and failure warning method provided by an embodiment of the present application is shown. Figure 2 A module diagram for implementing the elevator predictive maintenance and failure warning method provided by an embodiment of the present application is shown.
[0019] The object implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION
[0020] It should be understood that the specific embodiments described herein are merely intended to explain the present application and are not intended to limit the present application.
[0021] The elevator predictive maintenance and fault early warning method provided by the embodiment of the application can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc.
[0022] Referring to Figure 1 FIG. 1 is a flowchart of an elevator predictive maintenance and fault early warning method provided by an embodiment of the application. In this embodiment, the elevator predictive maintenance and fault early warning method includes the following steps. S1, obtaining running state data and historical running logs of an elevator to be maintained to generate a fusion data stream of the elevator to be maintained.
[0023] It should be explained that the elevator to be maintained refers to any one or more specific elevators monitored in a system, the running state data refers to time series data reflecting current physical running conditions of the elevator, which is collected in real time or near real time by a sensor network installed on the body of the elevator, including vibration data, current / voltage data, temperature data, noise data and other data, and the historical running logs refer to event, parameter and operation data recorded in time sequence obtained from an elevator control system (such as PLC, mainboard), a building management system (BMS) and operation and maintenance records, including fault history records, running cycle records, maintenance records, control parameter logs, performance index history and other data.
[0024] The fusion data stream of the elevator to be maintained is generated to convert multi-source data into a clean, aligned and feature-rich fusion data stream, which provides a solid data foundation for subsequent deep learning and accurate prediction.
[0025] In detail, the fusion data stream of the elevator to be maintained is generated by the following steps. The running state data and the historical running logs of the elevator to be maintained are cleaned respectively to obtain cleaned running state data and cleaned historical running logs; The timestamps of the cleaned running state data are marked as a reference main time axis of the cleaned running state data and the cleaned historical running logs; Discrete events and cumulative statistics of the cleaned historical running logs are identified; The cumulative statistics are converted into time point events; The discrete events and the time point events are mapped to the reference main time axis to obtain the fusion data stream of the elevator to be maintained.
[0026] The cleaning operation state data refers to high-quality, regular time series data obtained after a series of preprocessing of real-time, streaming data collected by original sensors, the cleaning historical operation log refers to a structured historical event data set obtained after standardization, error correction and integration of historical, discrete records extracted from various business systems, the timestamp refers to a date and time value uniquely identifying a data or event occurrence time, the reference main time axis refers to a continuous time series formed by all timestamps of the cleaning operation state data arranged in strict ascending order, the discrete event refers to an event occurring at a certain specific instant and lasting for a very short time, such as fault occurrence, maintenance operation completion, elevator operation cycle start, safety circuit disconnection and the like, the cumulative statistical quantity refers to a summary value obtained after accumulation, counting or aggregation of a certain activity or state in a certain time period, such as total operation times per day, monthly door fault occurrence times, cumulative traction machine operation hours, the time point event refers to a virtual discrete event created by "snapshotting" a cumulative statistical quantity into an instantaneous value at the end of its statistical period for data fusion, and the fused data stream refers to a final output result.
[0027] Optionally, the operation state data and the historical operation log of the elevator to be maintained are cleaned respectively to obtain cleaning operation state data and cleaning historical operation log, and the cleaning of the operation state data includes linear interpolation processing of missing values, filter smoothing and noise reduction, and statistical methods (such as Z-score and IQR) for removing abnormal values, and the cleaning of the historical operation log includes format standardization, correction of common errors, and event record merging and deduplication.
[0028] S2, a mechanism model of the elevator to be maintained is constructed, the operation state of the elevator to be maintained is analyzed based on the fused data stream, and virtual sensor data of the elevator to be maintained is analyzed by using the mechanism model according to the operation state.
[0029] The mechanism model of the elevator to be maintained constructed by the application can truly map, analyze and predict the state of a physical entity.
[0030] In detail, the construction of the mechanism model of the elevator to be maintained includes: An elevator data model of the elevator to be maintained is established, wherein the elevator data model includes a traction system model, a suspension and guide system model, a door system model and a control system model; A coupling interface of the elevator data model is defined; And through the coupling interface, the elevator data model is integrated to obtain an integrated elevator data model; The integrated elevator data model is calibrated to obtain a calibrated elevator data model; A virtual sensor of the calibrated elevator data model is defined; The virtual sensor is integrated into the calibrated elevator data model to obtain a mechanism model of the elevator to be maintained.
[0031] Wherein, the traction system model refers to a mathematical model based on Newtonian mechanics and motor theory, used to simulate the power transmission core components of the elevator, the suspension and guide system model refers to a mathematical model based on multi-body dynamics and vibration theory, used to simulate the motion posture and stress of the elevator car and counterweight, the door system model refers to a mathematical model based on kinematics and motor control theory, used to simulate the opening and closing process of the elevator door, the control system model refers to an algorithm model based on control theory and logic state, used to simulate the decision-making process of the elevator brain (control cabinet), the coupling interface refers to a logical connection point and data specification that specifies how data is exchanged and interacted between subsystem models, for example, the output of the control system model: torque set value as the input of the traction system model, the integrated elevator data model refers to a single composite model formed by connecting the four subsystem models through the coupling interface, which can simulate the complete operation process of the elevator, the calibrated elevator data model refers to a model highly consistent with the physical entity obtained by optimizing and adjusting unknown or uncertain parameters in the integrated elevator data model using real operation data collected from a specific elevator to be maintained, including calibration of parameters such as friction coefficient, damping coefficient, moment of inertia, which are difficult to obtain through theoretical calculation, the virtual sensor refers to a calculation module embedded in the calibrated elevator data model for calculating potential virtual data, and the mechanism model refers to the final delivery, i.e. the calibrated elevator data model integrated with the virtual sensor.
[0032] Optionally, the parameter calibration of the integrated elevator data model to obtain the calibrated elevator data model is a complete system realized in a simulation environment (such as Simulink), including control closed loop and physical forward path.
[0033] Optionally, the parameter calibration of the integrated elevator data model to obtain the calibrated elevator data model can be calibrated by comparing the model simulation output (such as current curve, speed curve) with the measured data, using an optimization algorithm to automatically adjust the parameters to minimize the error between the two.
[0034] Further, the elevator data model of the elevator to be maintained is established, including: establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained; establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained;
[0035] establishing a dynamic equation and a torque balance equation of a three-phase asynchronous motor of the elevator to be maintained;
[0036] The three-phase asynchronous motor dynamic equation refers to a high-order differential equation set describing the internal electromagnetic energy conversion process of the three-phase asynchronous motor, the torque balance equation refers to an equation established based on the expression of Newton's second law for a rotating system, the traction system model refers to a complete mathematical model for simulating the elevator power source (traction machine) and its directly driven load, which is coupled by the "three-phase asynchronous motor dynamic equation" and the "torque balance equation", the mass-spring-damper system refers to a lumped parameter theory model for abstracting and simplifying a complex vibration system. It represents the inertia, elasticity and energy consumption characteristics of the system by an ideal mass block, a spring and a damper respectively, the external excitation force equation refers to a system motion differential equation derived from the mass-spring-damper model by applying Newton's second law or Lagrange's equation, the suspension and guide system model refers to a dynamic model for simulating the motion state of the elevator car and the counterweight on the suspension and guide device thereof based on the "mass-spring-damper system" theory and the specific mathematical expression of the "external excitation force equation", the simulated door opening and closing curve refers to a reference instruction curve defining the ideal position and ideal speed corresponding relationship of the elevator door panel in the opening and closing process, the door system model refers to a mathematical model simulating the elevator door machine and its controller, and the core task thereof is to track the "simulated door opening and closing curve", the speed curve refers to a reference instruction curve generated by the elevator control system, which defines the ideal speed and ideal position (or time) corresponding relationship of the car in the running process, the PID controller refers to a controller generating a control signal by calculating the error between the "expected value" (speed curve) and the "actual value" (feedback speed), and the state machine refers to various states (such as "stopping", "starting", "uniform speed", "deceleration", "opening door", "closing door") that the elevator can be in during operation.
[0037] Further, the further, the external excitation force equation of the elevator to be maintained is established, comprising: According to the mass-spring-damper system of the elevator to be maintained, the elevator mass, displacement vector, elevator stiffness and elevator damping of the elevator to be maintained are defined. Combined with the elevator mass, displacement vector, elevator stiffness and elevator damping, the external excitation force equation of the elevator to be maintained is constructed.
[0038] Further, as another embodiment of the present application, the external excitation force equation:
[0039] Wherein, represents the external excitation force vector of the elevator to be maintained, represents the elevator mass of the elevator to be maintained, an acceleration of the elevator to be maintained, a displacement vector, an elevator stiffness of the elevator to be maintained, a speed of the elevator to be maintained, an elevator damping of the elevator to be maintained.
[0040] wherein the elevator mass refers to inertia characteristics of components of the elevator system, the displacement vector refers to displacement states of degrees of freedom of the system, the elevator stiffness refers to the ability of the system to resist deformation, the elevator damping refers to energy dissipation in the system, the external excitation force vector refers to external forces acting on the system, the acceleration refers to a second order derivative of the displacement vector with respect to time, and the speed refers to a first order derivative of the displacement vector with respect to time.
[0041] The application is based on the fusion data stream, analyzes the running state of the elevator to be maintained, and according to the running state, uses the mechanism model to analyze the virtual sensor data of the elevator to be maintained, which lays a solid and reliable foundation for the next step of "health feature extraction" and "health state analysis". The running state includes a stationary state, an acceleration state, a uniform speed state, a deceleration state, a full load uplink state, etc.
[0042] In detail, the use of the mechanism model to analyze the virtual sensor data of the elevator to be maintained includes: According to the running state of the elevator to be maintained, the target speed, the feedback motor speed and the door machine opening and closing instruction of the elevator to be maintained are determined; By the target speed and the feedback motor speed, the electromagnetic torque of the elevator to be maintained is calculated by using the corresponding control system model of the elevator to be maintained; Based on the electromagnetic torque, the motor speed of the elevator to be maintained is calculated by using the corresponding traction system model of the elevator to be maintained; The motor speed is converted into car speed to calculate the car vertical vibration acceleration and horizontal guide force of the elevator to be maintained by using the corresponding suspension and guide system model of the elevator to be maintained; Based on the door machine opening and closing instruction, the door machine motor torque and the door speed of the elevator to be maintained are analyzed by using the door system model of the elevator to be maintained, The electromagnetic torque, the motor speed, the car speed, the car vertical vibration acceleration, the horizontal guide force, the door machine motor torque and the door speed are used as the virtual sensor data of the elevator to be maintained.
[0043] Wherein, the reference speed refers to the ideal car running speed set by the elevator control system, the feedback motor speed refers to the actual motor speed collected and fed back by the rotary encoder on the motor shaft, the door machine switch instruction refers to the logical instruction from the elevator main controller or the door controller, such as "open door", "close door", "stop", the electromagnetic torque refers to the output torque generated by the motor through electromagnetic action, which is the power source for driving the elevator to run, the motor speed refers to the motor speed calculated based on the traction system model, which is used for comparison with the feedback speed to evaluate the motor response performance and load disturbance, the car speed refers to the actual car running speed converted from the motor speed through the traction sheave radius and the speed reduction ratio, the car vertical vibration acceleration refers to the vibration acceleration of the car in the vertical direction calculated by the suspension and guide system model, the horizontal guide force refers to the horizontal force exerted by the guide shoe on the guide rail during the running of the car, the door machine motor torque refers to the torque output by the door machine motor when executing the opening and closing door action, the door speed refers to the instantaneous movement speed of the door panel during the opening and closing door process, and the virtual sensor data refers to the internal state or key parameter of the system calculated or reconstructed without direct physical sensors.
[0044] S3, based on the virtual sensor data, extracting the health features of the elevator to be maintained by using a preset parallel feature extraction network to analyze the real-time health status of the elevator to be maintained.
[0045] Based on the virtual sensor data, the health features of the elevator to be maintained are extracted by using a preset parallel feature extraction network, which can improve the accuracy of the later health analysis.
[0046] In detail, based on the virtual sensor data, the health features of the elevator to be maintained are extracted by using a preset parallel feature extraction network, which includes: Based on the virtual sensor data, the local mode of the elevator to be maintained is extracted by using the space-time feature extraction branch in the parallel feature extraction network; Based on the virtual sensor data, the long-term evolution state and the periodic state of the elevator to be maintained are analyzed by using the time series feature extraction branch in the parallel feature extraction network; The local mode, the long-term evolution state and the periodic state are integrated into a joint feature vector; The element weight of the joint feature vector is defined; The health features of the elevator to be maintained are generated by the element weight and the joint feature vector.
[0047] The parallel feature extraction network refers to a deep learning network structure, which includes multiple parallel branches, each branch is responsible for extracting different types of features from input data, the spatio-temporal feature extraction branch refers to extracting local patterns in time and space dimensions from virtual sensor data, the local pattern refers to the behavior characteristics of the elevator in a specific time period or a specific running stage, the time series feature extraction branch refers to extracting evolution rules and periodic behaviors in a long time scale from virtual sensor data, the long-term evolution state refers to the change trend of the running state of the elevator in a long time dimension, such as performance degradation, component wear, control accuracy decline, etc., the periodic state refers to the behavior mode with obvious time periodicity in the running of the elevator, such as daily peak running frequency, repeated control error after every N cycles, etc., the joint feature vector refers to a high-dimensional vector formed by splicing or weighted fusion of features (local patterns, long-term evolution state, periodic state) from different branches, the element weight refers to the weight value corresponding to each feature component in the joint feature vector, which is used to represent the importance of the feature to the health state of the elevator, and the health feature refers to the final representation generated by weighted processing of the joint feature vector.
[0048] Optionally, the local pattern of the elevator to be maintained is extracted by using the spatio-temporal feature extraction branch in the parallel feature extraction network to perform sliding window calculation on the input data using one-dimensional or two-dimensional convolution kernels, each convolution kernel acts as a feature detector and learns to extract a specific local pattern, for example, one convolution kernel may learn to activate high-frequency impact, and another convolution kernel may learn to activate resonance in a specific frequency band.
[0049] The analysis of the real-time health state of the elevator to be maintained not only provides quantitative evaluation of the current state of the elevator, but also supports root cause positioning and trend prediction. The real-time health state refers to the dynamic evaluation result of the current running quality, functional integrity, potential risk and degradation trend of the elevator obtained by comprehensive analysis of the elevator running data, virtual sensor data and health features at a specific time point.
[0050] S4, according to the real-time health state, analyzing the health evolution trajectory of the elevator to be maintained to calculate the health degree index decline slope of the elevator to be maintained, and determining the fault warning level of the elevator to be maintained according to the health degree index decline slope.
[0051] According to the real-time health state, the analysis of the health evolution trajectory of the elevator to be maintained changes passive current state monitoring to active future trend prediction, so as to realize real predictive maintenance instead of post-maintenance or regular maintenance.
[0052] In detail, the analyzing the health evolution trajectory of the elevator to be maintained according to the real-time health state comprises: generating a supervised learning sequence of the elevator to be maintained according to the real-time health state; encoding the supervised learning sequence into a context vector by using an Encoder layer in a preset Seq2Seq model; generating a health sequence value of the elevator to be maintained by using a Decoder layer in the Seq2Seq model based on the context vector; generating the health evolution trajectory of the elevator to be maintained according to the health sequence value.
[0053] The supervised learning sequence refers to an input-output data pair with explicit labels artificially constructed for training a time series prediction model, the Seq2Seq model refers to a deep learning architecture specially used for processing a sequence as input and a sequence as output, the Encoder layer refers to the first half of the Seq2Seq model, which is an encoder. Its task is to compress a variable-length input sequence into a fixed-length context vector containing all important information, the context vector refers to a fixed-dimension vector generated after the Encoder layer processes the entire input sequence. It is a highly condensed and abstract representation of the input sequence information, the Decoder layer refers to the second half of the Seq2Seq model, which is a decoder. Its task is to generate (predict) each element of the output sequence step by step with the context vector as the initial condition, the health sequence value refers to the original prediction value of the future time step directly output by the Decoder layer, and the health evolution trajectory refers to a complete health state curve extending from the past to the future formed by connecting the historical health sequence and the future health sequence value predicted by the Decoder on the time axis.
[0054] The application calculates the health degree index decline slope of the elevator to be maintained, and determines the fault warning level of the elevator to be maintained according to the health degree index decline slope. The health degree index decline slope refers to the rate of change of the elevator health degree index (such as health score) with time, reflecting the speed of deterioration of the elevator health state, and the fault warning level refers to a risk classification set based on the health degree decline slope and the current health degree, which is used to guide the maintenance decision, such as normal, attention, warning, alarm and fault.
[0055] S5, when the fault warning level exceeds a preset fault warning level threshold, a health-fault causal chain of the elevator to be maintained is established to determine a maintenance strategy of the elevator to be maintained.
[0056] The application establishes a health fault causal chain of the to-be-maintained elevator when the fault warning level exceeds the preset fault warning level threshold, thereby providing a visual basis for later elevator maintenance, and improving the maintenance effect of the to-be-maintained elevator.
[0057] In detail, the health fault causal chain of the to-be-maintained elevator is established when the fault warning level exceeds the preset fault warning level threshold, including: marking a trigger time and a time window when the fault warning level exceeds the preset fault warning level threshold; extracting abnormal events in the time window as causal chain nodes of the to-be-maintained elevator; establishing a time-elevator state association graph of the abnormal events; determining a causal path and a path weight of the to-be-maintained elevator according to the time-elevator state association graph; determining a health fault causal chain of the to-be-maintained elevator in combination with the causal chain nodes, the time-elevator state association graph, the causal path, and the path weight.
[0058] The fault warning level threshold refers to one or more preset numerical limits for judging whether the current health risk of the elevator reaches a degree requiring action, the trigger time refers to a time point when the fault warning level exceeds the preset threshold for the first time, marking the starting point of the causal chain analysis, the time window refers to a time range backtracking from the trigger time as the end point, used for collecting and analyzing historical data related to the current fault, the abnormal event refers to a significant change or abnormal pattern deviating from the normal range in the elevator operating state or virtual sensor data within the time window, the causal chain node refers to a basic element in the health fault causal chain, representing a certain state, event or component of the elevator, the time-elevator state association graph refers to a chart taking time as the horizontal axis and elevator state as the vertical axis, used for visualizing the change trend of the elevator state over time, the causal path refers to a logical evolution path from the initial abnormal event to the final fault result in the causal chain, composed of multiple nodes and edges, the path weight refers to the strength or probability of each edge (i.e. the relationship between nodes) in the causal path, representing the influence degree of one state on another state, and the health fault causal chain refers to a structured and traceable causal relationship model for describing the complete evolution path from the initial abnormal state to the final fault result.
[0059] Optionally, the abnormal events in the time window can be events such as "excessive guide rail vibration", "intensified guide rail wear", and "decreased running smoothness".
[0060] Finally, the maintenance strategy of the elevator to be maintained can realize the transition from passive repair to active prevention, greatly improving the safety and economy of the elevator. In detail, the maintenance strategy is based on the current health status of the elevator, the failure risk, and the result of the cause-effect chain analysis to formulate a systematic maintenance plan, including maintenance content, priority, time arrangement, and resource allocation, etc. The maintenance target is to eliminate or alleviate the failure risk and restore the health status of the elevator; the maintenance content is the specific components to be repaired or replaced (such as guide rails, door machines, motors, etc.); the maintenance method is preventive maintenance, predictive maintenance, repair maintenance, maintenance priority is sorted according to the risk level and the importance of the cause-effect chain; the maintenance time is emergency maintenance, planned maintenance, and periodic maintenance; and the resource allocation is personnel, spare parts, tools, and time window, etc.
[0061] The scheme realizes accurate analysis of the real-time health status of the elevator by fusing elevator operation state data and historical logs, constructing a mechanism model to generate virtual sensor data, and extracting health features using a parallel feature extraction network. Based on the health evolution trajectory, the health degree index decline slope is calculated to dynamically evaluate the degradation trend of the elevator performance. Combined with the failure warning level mechanism, potential failure risks are effectively identified in advance. When the warning level exceeds the threshold, the system automatically establishes a health failure cause-effect chain to accurately locate the failure root cause and evolution path, providing a scientific basis for formulating targeted maintenance strategies. The scheme significantly improves the intelligent level of elevator operation and maintenance, realizes the transition from passive repair to active predictive maintenance, greatly reduces the failure rate and maintenance cost, prolongs the service life of the equipment, and improves the safety and reliability of elevator operation. It provides efficient and reliable technical support for smart buildings and city elevator management. Therefore, the present application can improve the efficiency of elevator predictive maintenance and failure warning.
[0062] As Figure 2 shown, it is a functional module diagram of an elevator predictive maintenance and failure warning system according to the present application.
[0063] The elevator predictive maintenance and failure warning system 200 according to the present application can be installed in an electronic device. According to the functions implemented, the elevator predictive maintenance and failure warning system can include a data stream fusion module 201, a virtual sensor data analysis module 202, a real-time health analysis module 203, a warning level determination module 204, and a maintenance strategy generation module 205. The modules according to the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, stored in the memory of the electronic device.
[0064] In the embodiments of the present application, the functions of each module / unit are as follows: The data stream fusion module 201 is configured to acquire operation state data and historical operation logs of the elevator to be maintained to generate a fusion data stream of the elevator to be maintained. The virtual sensor data analysis module 202 is configured to construct a mechanism model of the elevator to be maintained, analyze an operation state of the elevator to be maintained based on the fusion data stream, and analyze virtual sensor data of the elevator to be maintained by using the mechanism model according to the operation state. The real-time health analysis module 203 is configured to extract a health feature of the elevator to be maintained by using a preset parallel feature extraction network based on the virtual sensor data to analyze a real-time health state of the elevator to be maintained. The early warning level determination module 204 is configured to analyze a health evolution trajectory of the elevator to be maintained according to the real-time health state, calculate a health degree index decline slope of the elevator to be maintained, and determine a fault early warning level of the elevator to be maintained according to the health degree index decline slope. The maintenance strategy generation module 205 is configured to establish a health fault cause-effect chain of the elevator to be maintained when the fault early warning level exceeds a preset fault early warning level threshold to determine a maintenance strategy of the elevator to be maintained.
[0065] In detail, the modules in the elevator predictive maintenance and fault early warning system 200 in the embodiments of the present application adopt the same technical means as the elevator predictive maintenance and fault early warning method in the above Figure 1 , and can produce the same technical effects, which will not be described here.
[0066] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.
[0067] Finally, it should be noted that in the above embodiments, each embodiment can be combined or independent, and deleting any one of them does not affect the technical implementation of the other embodiments. The above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.
Claims
1. A method for elevator predictive maintenance and failure early warning, characterized by, The method comprises: obtaining the running state data and historical running log of the elevator to be maintained to generate a fusion data stream of the elevator to be maintained; constructing a mechanism model of the elevator to be maintained, analyzing the running state of the elevator to be maintained based on the fusion data stream, and analyzing the virtual sensor data of the elevator to be maintained according to the running state by using the mechanism model; based on the virtual sensor data, extracting the health features of the elevator to be maintained by using a preset parallel feature extraction network to analyze the real-time health state of the elevator to be maintained; according to the real-time health state, analyzing the health evolution trajectory of the elevator to be maintained to calculate the health degree index decline slope of the elevator to be maintained, and determining the fault warning level of the elevator to be maintained according to the health degree index decline slope; when the fault warning level exceeds a preset fault warning level threshold, establishing a health fault causal chain of the elevator to be maintained to determine the maintenance strategy of the elevator to be maintained.
2. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, The construction of the mechanism model of the elevator to be maintained comprises: establishing an elevator data model of the elevator to be maintained, wherein the elevator data model comprises a traction system model, a suspension and guide system model, a door system model and a control system model; defining the coupling interface of the elevator data model; and integrating the elevator data model through the coupling interface to obtain an integrated elevator data model; calibrating the integrated elevator data model to obtain a calibrated elevator data model; defining a virtual sensor of the calibrated elevator data model; integrating the virtual sensor into the calibrated elevator data model to obtain the mechanism model of the elevator to be maintained.
3. The elevator predictive maintenance and breakdown early warning method of claim 2, wherein, The establishment of the elevator data model of the elevator to be maintained comprises: establishing a three-phase asynchronous motor dynamic equation and a torque balance equation of the elevator to be maintained; combining the three-phase asynchronous motor dynamic equation and the torque balance equation to establish a traction system model of the elevator to be maintained; establishing a mass-spring-damper system of the elevator to be maintained to establish an external excitation force equation of the elevator to be maintained; and generating a suspension and guide system model of the elevator to be maintained according to the external excitation force equation; establishing a simulated door opening and closing curve of the elevator to be maintained to determine a door system model of the elevator to be maintained; analyzing the speed curve, PID controller and state machine of the elevator to be maintained; establishing a control system model of the elevator to be maintained through the speed curve, PID controller and state machine. Combining the traction system model, the suspension and guide system model, the door system model and the control system model, the elevator data model of the elevator to be maintained is established.
4. The elevator predictive maintenance and breakdown early warning method of claim 3, wherein, The establishment of the external excitation force equation of the elevator to be maintained comprises: defining the elevator mass, displacement vector, elevator stiffness and elevator damping of the elevator to be maintained according to the mass-spring-damper system of the elevator to be maintained; combining the elevator mass, displacement vector, elevator stiffness and elevator damping to construct the external excitation force equation of the elevator to be maintained, wherein the external excitation force equation: wherein denotes an external excitation force vector of the elevator to be maintained, denotes an elevator mass of the elevator to be maintained, denotes an acceleration of the elevator to be maintained, denotes a displacement vector, denotes an elevator damping of the elevator to be maintained, denotes a velocity of the elevator to be maintained, denotes an elevator stiffness of the elevator to be maintained.
5. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, The generation of the fusion data stream of the elevator to be maintained comprises: cleaning operation state data and historical operation logs of the elevator to be maintained respectively to obtain cleaned operation state data and cleaned historical operation logs; marking timestamps of the cleaned operation state data as a reference main time axis of the cleaned operation state data and the cleaned historical operation logs; identifying discrete events and cumulative statistics of the cleaned historical operation logs; and converting the cumulative statistics into time point events; mapping the discrete events and the time point events to the reference main time axis to obtain fusion data flow of the elevator to be maintained.
6. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, the analysis of the virtual sensor data of the elevator to be maintained by using the mechanism model comprises: determining target speed, feedback motor speed and door machine switch instruction of the elevator to be maintained according to the operation state of the elevator to be maintained; calculating electromagnetic torque of the elevator to be maintained by using the control system model corresponding to the elevator to be maintained through the target speed and the feedback motor speed; calculating motor speed of the elevator to be maintained based on the electromagnetic torque by using the traction system model corresponding to the elevator to be maintained; converting the motor speed into car speed to calculate car vertical vibration acceleration and horizontal guiding force of the elevator to be maintained by using the suspension and guiding system model corresponding to the elevator to be maintained; analyzing door machine motor torque and door speed of the elevator to be maintained based on the door machine switch instruction by using the door system model of the elevator to be maintained, electromagnetic torque, motor speed, car speed, car vertical vibration acceleration, horizontal guiding force, door machine motor torque and door speed as virtual sensor data of the elevator to be maintained.
7. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, the extraction of health features of the elevator to be maintained by using a preset parallel feature extraction network based on the virtual sensor data comprises: extracting local patterns of the elevator to be maintained by using a space-time feature extraction branch in the parallel feature extraction network based on the virtual sensor data; analyzing long-term evolution state and periodic state of the elevator to be maintained by using a time series feature extraction branch in the parallel feature extraction network based on the virtual sensor data; integrating the local patterns, the long-term evolution state and the periodic state into a joint feature vector; defining element weight of the joint feature vector; generating health features of the elevator to be maintained through the element weight and the joint feature vector.
8. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, the analysis of health evolution trajectory of the elevator to be maintained according to the real-time health state comprises: generating a supervised learning sequence of the elevator to be maintained according to the real-time health state; encoding the supervised learning sequence into a context vector by using an Encoder layer in a preset Seq2Seq model; generating health sequence values of the elevator to be maintained by using a Decoder layer in the Seq2Seq model based on the context vector; generating health evolution trajectory of the elevator to be maintained according to the health sequence values.
9. The elevator predictive maintenance and breakdown early warning method of claim 1, wherein, when the fault warning level exceeds a preset fault warning level threshold, the establishment of health fault causal chain of the elevator to be maintained comprises: mark a trigger time and a time window of the failure warning level exceeding a preset failure warning level threshold value; extract an abnormal event of the time window as a causal chain node of the elevator to be maintained; establish a time-elevator state association graph of the abnormal event; determine a causal path and a path weight of the elevator to be maintained according to the time-elevator state association graph; determine a health failure causal chain of the elevator to be maintained in combination with the causal chain node, the time-elevator state association graph, the causal path, and the path weight.
10. An elevator predictive maintenance and failure warning system, characterized by, The system comprises: a data stream fusion module configured to acquire running state data and historical running logs of an elevator to be maintained to generate a fusion data stream of the elevator to be maintained; a virtual sensor data analysis module configured to construct a mechanism model of the elevator to be maintained, analyze a running state of the elevator to be maintained based on the fusion data stream, and analyze virtual sensor data of the elevator to be maintained by using the mechanism model according to the running state; a real-time health analysis module configured to extract health features of the elevator to be maintained by using a preset parallel feature extraction network based on the virtual sensor data to analyze a real-time health state of the elevator to be maintained; a warning level determination module configured to analyze a health evolution trajectory of the elevator to be maintained according to the real-time health state to calculate a health degree index decline slope of the elevator to be maintained, and determine a failure warning level of the elevator to be maintained according to the health degree index decline slope; a maintenance strategy generation module configured to establish a health failure causal chain of the elevator to be maintained when the failure warning level exceeds a preset failure warning level threshold value to determine a maintenance strategy of the elevator to be maintained.