Predictive elevator maintenance method and system

Abstract

The invention discloses an elevator predictive maintenance method and system, and aims to solve the technical problems that in the prior art, elevator maintenance depends on afterward maintenance or regular maintenance, so that response is passive, and cost is high. According to the method, a multi-mode sensor is arranged on an elevator to collect data such as vibration and noise in real time; performing fusion processing and feature extraction on the data; and the health state of the elevator is evaluated through the prediction model, and early warning and maintenance suggestions are generated before potential faults occur. The safety and reliability of elevator operation can be remarkably improved, passive maintenance is changed into active prevention, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of elevator monitoring, and in particular to an elevator predictive maintenance method and system. Background Art

[0002] In recent years, the safe operation and reliability of elevators have garnered widespread attention. Currently, the industry primarily addresses potential elevator failures through two approaches: post-fault diagnosis and scheduled maintenance. On the one hand, after an elevator malfunction occurs, diagnostic equipment and specialized personnel are used to analyze the cause; on the other hand, preventative inspections and maintenance are performed at regular intervals. However, both approaches have significant shortcomings: post-fault diagnosis is a passive response that cannot prevent failures from occurring; and scheduled maintenance, lacking specificity, is often costly and inefficient. Furthermore, manual inspections struggle to detect potential safety hazards in a timely manner.

[0003] With the development of IoT sensing technology and artificial intelligence algorithms, new solutions for elevator status monitoring and early warning have emerged. These utilize sensors installed throughout the elevator to collect real-time operating data, which is then uploaded to the cloud for big data analysis and machine learning processing. This allows for predictive maintenance of elevator faults to a certain extent. However, existing solutions for remote elevator monitoring and fault prediction still have shortcomings: some systems require expensive dedicated hardware or complex deployment, while others rely on expert experience for calibration. Prediction accuracy and real-time performance remain to be improved. Therefore, a more efficient and reliable method and system is urgently needed that can integrate data from multiple types of sensors within the elevator car, intelligently analyze the elevator's operating status, and predict reliability in advance, enabling timely detection of fault signs and optimizing maintenance strategies.

[0004] Therefore, we propose an elevator predictive maintenance method and system to solve the above problems. Summary of the Invention

[0005] The present invention provides an elevator predictive maintenance method and system, which can provide prediction results before a fault occurs, improve the initiative and pertinence of maintenance work, and thus enhance the safety and reliability of elevator operation.

[0006] A first aspect of the present invention provides an elevator predictive maintenance method, which includes: acquiring multimodal sensor data during elevator operation, and transmitting the multimodal sensor data to a data processing unit in real time through a data acquisition device; the data processing unit preprocesses and fuses the received multimodal sensor data, including signal filtering, data alignment and feature extraction, to obtain characteristic parameters that characterize the elevator's operating status; based on a pre-constructed elevator reliability prediction model, the characteristic parameters are calculated and analyzed to obtain the elevator's current reliability assessment results and / or fault prediction information; judging whether the reliability assessment results contain abnormal conditions that exceed a preset threshold; and if so, generating corresponding fault warning information and maintenance suggestion outputs.

[0007] Optionally, in a first implementation method of the first aspect of the present invention, the data processing unit synchronizes the data of different sensors according to timestamps and extracts the elevator operation status characteristics, which include the frequency domain energy distribution of the vibration signal, the sound pressure level of the noise signal, the fluctuation amplitude of the operating load, and the changing trend of the ambient temperature.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the elevator reliability prediction model is constructed using a machine learning algorithm, which is obtained by training by inputting historical multimodal sensor data and corresponding fault occurrence conditions, and can output a probability value of elevator failure occurrence, a health status score or a remaining life estimate. The machine learning algorithm includes a random forest algorithm, a support vector machine algorithm or a deep neural network algorithm.

[0009] Optionally, in a third implementation method of the first aspect of the present invention, it includes: deploying four types of sensors in key fault-sensitive areas of the elevator system; sampling through a multi-channel data acquisition device with a configurable sampling rate, and adding a unified microsecond timestamp to all sampling points based on the hardware clock; constructing a parallel transmission channel, including a main channel and a backup channel, the main channel is connected to the industrial Ethernet through a shielded twisted pair cable, and Manchester encoding is used to eliminate common-mode interference. The backup channel encapsulates data packets through a 4G / 5G module, and adds link layer AES-256 encryption and CRC check; verifies the continuity of timestamps, reorders out-of-order data packets according to the hardware clock, and automatically triggers the acquisition module to incrementally retransmit if the packet loss rate is detected to be greater than 0.1%, and outputs multi-source original sensor data streams.

[0010] Optionally, in a fourth implementation method of the first aspect of the present invention, it includes: performing 20-500Hz mechanical resonance band-pass filtering on the vibration signal, extracting the energy of the characteristic frequency bands of bearings and gears, performing A-weighted filtering and human ear sensitive frequency band energy integration on the noise signal, and quantifying the intensity of abnormal noise; calculating the standard deviation of the load change rate throughout the operation of the load signal, identifying abnormal load addition and reduction patterns, fitting the brake surface temperature rise curve on the temperature signal, and extracting the temperature rise rate per minute; based on the microsecond timestamp of the hardware clock, the four types of features of vibration spectrum energy, noise abnormal sound intensity, load fluctuation value, and temperature rise rate are uniformly mapped to a 10ms time window to generate a synchronous feature matrix; the synchronous feature matrix is ​​compressed through principal component analysis, the principal components with a cumulative contribution rate ≥95% are retained, and a multi-dimensional elevator state feature vector is output.

[0011] Optionally, in a fifth implementation method of the first aspect of the present invention, it includes: decomposing the multi-dimensional elevator state feature vector into independent physical parameters, and comparing them with a preset fault judgment threshold to obtain a fault feature identification set; mapping the number of concurrent fault types in the fault feature identification set to a preset risk level to obtain a risk level index for the entire machine; based on the fault feature identification set and historical maintenance data: for each type of fault identification, calculating the component-level remaining life attenuation coefficient according to the proportion exceeding the threshold, and taking the minimum value of all attenuation coefficients as the remaining effective operating time reference value; matching the preset maintenance strategy matrix according to the risk level index and the remaining effective operating time reference value to generate a quantitative indicator of elevator real-time reliability containing risk level, remaining time and maintenance instructions.

[0012] Optionally, in a sixth implementation method of the first aspect of the present invention, it includes: when the risk level index exceeds the first warning threshold, triggering a yellow warning mark; when the risk level index exceeds the second warning threshold or the remaining effective operating time is lower than a preset safety margin, triggering a red warning mark; obtaining a warning level mark; based on the warning level mark, calling a preset fault location knowledge base to obtain a fault location report; based on the fault location report and the remaining effective operating time, obtaining a maintenance instruction set; encapsulating the warning level mark, the fault location report, and the maintenance instruction set into a fault warning instruction.

[0013] The second aspect of the present invention provides an elevator predictive maintenance system, which includes: a multimodal sensor component installed on the elevator car and its related components, and used to collect various status data of the elevator operation; a data acquisition and transmission module, connected to the sensor component, acquiring the measurement data of each sensor in real time and performing preliminary processing, and then sending the data to a designated data processing device through a communication network; a data processing and prediction module, including at least one processor and a memory, wherein the memory stores an elevator reliability prediction model program, and the processor is configured to execute the program to perform data fusion and reliability prediction analysis on the received multimodal sensor data to generate an elevator operation reliability evaluation result; an alarm and interaction module, which is used to receive the evaluation result and issue a fault warning signal when the result is abnormal, provide a human-computer interaction interface to display elevator status information and prediction reports, and support notification of warning information to maintenance personnel.

[0014] Optionally, in a first implementation method of the second aspect of the present invention, the data processing and prediction module is deployed on a cloud server or a main control computer of an elevator monitoring center, and can centrally analyze data from multiple elevators; the data processing and prediction module uses a distributed database and a parallel computing framework to store and process historical elevator operation data, and regularly updates and trains the elevator reliability prediction model to improve prediction accuracy and system adaptability.

[0015] Optionally, in a second implementation of the second aspect of the present invention, the alarm and interaction module includes a mobile application interface. When the result output by the elevator reliability prediction model exceeds a threshold, the application will generate a fault warning message and push it to the mobile communication device of at least one elevator maintenance personnel so that they can understand the elevator status in time and take maintenance measures.

[0016] Beneficial effects: It can give prediction results before a fault occurs, improve the initiative and pertinence of maintenance work, and thus enhance the safety and reliability of elevator operation. Predict potential fault hazards in advance: Compared with traditional passive fault diagnosis, the present invention can identify abnormal trends of elevators before a fault occurs, issue early warning signals, and reserve sufficient time for maintenance, thus avoiding sudden elevator failures that may cause inconvenience to passengers or safety accidents.

[0017] Predictions can be made before a fault occurs, making maintenance more proactive and targeted, thereby improving elevator safety and reliability. On-demand maintenance reduces costs: Elevator status is assessed using reliability indicators, enabling on-demand maintenance (condition-based repairs), avoiding unnecessary scheduled maintenance and reducing maintenance costs and downtime. Furthermore, timely maintenance when the risk of failure is high prevents minor issues from becoming major failures, reducing the cost of overhauls and component replacement.

[0018] This system can provide predictions before a failure occurs, making maintenance more proactive and targeted, thereby enhancing the safety and reliability of elevator operation. Data-driven decision-making: This system leverages big data analytics and machine learning models to continuously learn about elevator operating characteristics, resulting in accurate and reliable predictions. Managers can use the quantitative reliability assessment results to optimize elevator operation scheduling and maintenance strategies, improving elevator efficiency. Furthermore, through the long-term accumulation of data and models, the system can autonomously improve its prediction algorithms, becoming increasingly intelligent with use and further enhancing safety assurance.

[0019] In summary, the technical solution provided by the present invention overcomes the shortcomings of the existing technology and provides an efficient and reliable elevator reliability prediction model implementation method and its supporting system, which has significant practical value in ensuring the safe operation of elevators. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of an embodiment of a method for predictive maintenance of an elevator according to an embodiment of the present invention; Figure 2 Schematic diagram of an embodiment of an elevator predictive maintenance system in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] Embodiments of the present invention provide an elevator predictive maintenance method and system that can provide prediction results before a fault occurs, improve the proactiveness and targetedness of maintenance work, and thus enhance the safety and reliability of elevator operation. The terms "first," "second," "third," "fourth," and so on (if any) in the specification and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including," "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.

[0022] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the elevator predictive maintenance method in an embodiment of the present invention includes: Step 1: Multimodal data acquisition. First, deploy multiple sensors in the elevator system to capture different types of operational data. Vibration sensors (accelerometers) are installed on the top, inside, and bottom of the elevator car to monitor vibration and impact during startup, operation, and stopping. Noise sensors (microphones) are installed inside the car or in the elevator shaft to pick up acoustic signals generated by the elevator motor, traction machine, and guide rails during operation. Load sensors or strain gauges are installed on the car floor to monitor load and force changes. Current and voltage sensors can also be integrated into the elevator control cabinet to record the electrical parameters of the drive motor, or temperature sensors can be installed to monitor motor and brake temperatures. After sensor installation, the data from these sensors is synchronously sampled and aggregated using a data acquisition module. This module, which can be a standalone embedded acquisition terminal, connects to each sensor, converts the collected analog signals into digital signals, and adds information such as timestamps for data alignment. The data is then transmitted in real time to a back-end data processing device via an industrial fieldbus or wireless communication network. In this embodiment, an embedded gateway is used as the data collection terminal, sending the data to the elevator monitoring center server via Ethernet. Alternatively, a 4G / 5G wireless module can be used to upload the data to a cloud server platform. Necessary anti-interference and encryption measures are implemented during the data communication process to ensure real-time and reliable data transmission.

[0023] Step 2: Data fusion and reliability prediction, data processing device ( Figure 2 After receiving real-time sensor data, the data processing and prediction module in the device first pre-processes and fuses data from different sources. Specifically, the device performs the following operations: Data cleaning and filtering: Apply digital filters (such as bandpass filtering) to signals such as vibration and noise to eliminate environmental noise and interference from irrelevant frequency bands; detect and correct abnormal spikes or missing values ​​that may appear in the sensor to improve data quality.

[0024] Time synchronization and alignment: Data from different sensors are aligned based on timestamps to ensure that the information from each sensor corresponds to the same time base, facilitating fusion analysis. If necessary, high-frequency signals are downsampled or low-frequency signals are interpolated to synchronize the data on the time axis.

[0025] Feature extraction and data fusion: Extract characteristic parameters reflecting the elevator's operating status from the time or frequency domain data of various sensors. Features such as peak acceleration, RMS value, and spectral energy distribution are extracted from vibration signals acquired by accelerometers. Features such as sound intensity level and spectral amplitude in specific frequency bands are extracted from noise signals. Features such as current mean and fluctuation amplitude are extracted from motor current signals. Force signals acquired by load sensors are used to extract load change rates. Features from multiple sensors are then fused according to time windows to form a comprehensive feature vector representing the elevator's status during that time period. Multi-sensor data fusion algorithms (such as weighted fusion, principal component analysis, or deep learning-based fusion networks) can be used to enhance the sensitivity of these features to faults.

[0026] Model Calculation and Reliability Assessment: Using the aforementioned fused features as input, a pre-established elevator reliability prediction model is invoked to perform calculations and obtain the current reliability assessment results. In this embodiment, the reliability prediction model is trained using a large amount of historical data and can output the probability of elevator failure and the health index of key components. The model may utilize a gradient boosted decision tree (GBDT) or deep neural network (DNN) architecture, taking as input a recent sensor feature sequence and outputting a health score between 0 and 1 (1 indicating good condition, close to 0 indicating a serious risk of failure), or directly outputting the predicted remaining uninterrupted service time (RUL). While the elevator is operating, the data processing device periodically runs the model (e.g., every few seconds or after each elevator cycle) to continuously update the elevator reliability index.

[0027] Step 3: Result Analysis and Fault Warning. The data processing device analyzes and determines the model output. If the prediction model indicates that the elevator is in normal condition (the health score is above the safety threshold), the result is stored in the database as a daily monitoring record. The elevator status information ("Operating normally, good reliability") is also updated on the monitoring interface. If the prediction result exceeds the warning threshold (the failure probability exceeds a predetermined percentage, or the health score falls below the threshold), the system triggers the warning mechanism. The processing device generates an alarm record containing detailed information, which may include: the time of the warning, the abnormal characteristics of the sensor involved, the possible fault type or affected component determined by the model (possibly wear of the elevator traction motor bearing), and recommended remedial measures. This alarm information is notified to relevant personnel through the alarm and interaction module. First, an alarm prompt is displayed on the computer in the elevator remote monitoring center, accompanied by an audible or flashing indicator, to attract the attention of the on-duty personnel. Second, the message is pushed to the maintenance personnel's mobile app via the mobile network, allowing them to be immediately informed of the abnormal elevator condition. After receiving the alarm, maintenance personnel can arrange targeted maintenance based on the information provided by the system, such as replacing worn parts or adjusting equipment, to prevent actual failures. After completing an inspection, maintenance personnel can also input actual fault conditions and treatment measures into the system to enrich the equipment fault database. The system also generates regular (monthly) elevator health reports and trend analyses, listing the changes in key elevator indicators, previous warnings, and their treatments, providing data support for property management and maintenance decision-making.

[0028] Another embodiment of the elevator predictive maintenance method in the embodiment of the present invention includes: Step 1: Install various types of sensors on the elevator car and related components to obtain multimodal sensor data during elevator operation, and transmit the multimodal sensor data to a data processing unit in real time through a data acquisition device; Specifically, a multi-physics collaborative sensing network is deployed, with four types of sensors deployed in key fault-sensitive areas of the elevator system: three-axis vibration sensors are installed on the traction motor housing and guide rail brackets to capture the acceleration frequency response characteristics of the mechanical transmission system in real time; broadband noise sensors are installed on the side walls of the elevator shaft, covering the 20Hz-8kHz acoustic characteristics of the motor and bearings; strain gauge load sensors are installed on the load-bearing beams of the car floor to dynamically monitor the load change gradient; and temperature sensors are mounted on the brake friction pads and traction motor winding surfaces to track the thermodynamic state. Adaptive synchronous acquisition of heterogeneous signals, through a multi-channel data acquisition device with configurable sampling rate: vibration signals are sampled at high frequencies of ≥1kHz, noise signals are sampled at audio frequencies of ≥8kHz, and load and temperature signals are sampled at low frequencies of ≥10Hz. A unified microsecond-level timestamp is added to all sampling points based on the hardware clock. Dual-mode redundant anti-interference transmission, building parallel transmission channels: Main channel: connected to industrial Ethernet via shielded twisted pair cable, using Manchester encoding to eliminate common-mode interference; Backup channel: encapsulates data packets through 4G / 5G modules, adding link layer AES-256 encryption and CRC checksum; End-to-end data integrity assurance is performed in the data processing unit: timestamp continuity is verified and out-of-order data packets are reordered according to the hardware clock; if the packet loss rate is detected to be greater than 0.1%, the acquisition module is automatically triggered to retransmit incrementally; and the multi-source raw sensor data stream is output with strict time alignment.

[0029] Step 2: The data processing unit preprocesses and fuses the received multimodal sensor data, including signal filtering, data alignment, and feature extraction, to obtain characteristic parameters that characterize the elevator's operating status; Specifically, the frequency domain physical characteristics are enhanced by applying a 20-500Hz mechanical resonance bandpass filter to the vibration signal to extract the energy of the characteristic frequency bands of bearings and gears. The noise signal is subjected to A-weighted filtering and energy integration in the human ear-sensitive frequency band (500Hz-6kHz) to quantify the intensity of abnormal noise. Extract physical features in the time domain, calculate the standard deviation of the load change rate throughout the entire operation from the load signal, and identify abnormal loading and unloading patterns; fit the brake surface temperature rise curve to the temperature signal and extract the temperature rise rate per minute; Multi-source spatiotemporal lock-step alignment, based on the hardware clock's microsecond timestamp, uniformly maps four types of features: vibration spectrum energy, noise intensity, load fluctuation value, and temperature rise rate to a 10ms time window to generate a synchronous feature matrix. Physical feature dimensionality reduction and fusion are carried out, and the synchronous feature matrix is ​​compressed through principal component analysis (PCA), retaining the principal components with cumulative contribution rate ≥ 95%, and outputting a multi-dimensional elevator state feature vector.

[0030] Step 3: Based on a pre-built elevator reliability prediction model, the characteristic parameters are calculated and analyzed to obtain the current reliability assessment result and / or fault prediction information of the elevator; Specifically, the fault physical feature threshold determination decomposes the multi-dimensional elevator state feature vector into independent physical parameters and compares them with the preset fault determination threshold: when the vibration spectrum energy ratio exceeds the first threshold, the traction machine bearing wear mark is generated; when the sound pressure difference of the noise characteristic frequency band exceeds the second threshold, the motor bearing abnormal noise mark is generated; when the load fluctuation standard deviation exceeds the third threshold, the wire rope fatigue mark is generated; when the brake temperature rise rate exceeds the fourth threshold, the brake slip mark is generated; and thus a set of fault feature marks is obtained; Quantify the risk of multiple fault coupling. Map the number of concurrent fault types in the fault feature identification set to a preset risk level: a single fault type corresponds to the first risk level; the coexistence of two fault types corresponds to the second risk level; the coexistence of three or more fault types corresponds to the third risk level. The risk level index of the entire machine is obtained. Dynamic estimation of remaining effective operating time, based on a collection of fault signatures and historical maintenance data: For each type of fault signature, the component-level remaining life attenuation coefficient is calculated based on the percentage exceeding the threshold; the minimum value of all attenuation coefficients is taken as the baseline value of the remaining effective operating time of the entire machine; Maintenance decisions are adaptively generated, matching the preset maintenance strategy matrix based on the risk level index and the remaining effective operating time benchmark value: when the risk level is level 1 and there is sufficient remaining time, the first maintenance instruction is output; when the risk level is level 3 or the remaining time is insufficient, an emergency maintenance instruction is output; and a real-time quantitative indicator of elevator reliability is generated, including the risk level, remaining time, and maintenance instructions. Step 4: Determine whether the reliability assessment result has any abnormality exceeding a preset threshold; if so, generate corresponding fault warning information and maintenance suggestion output.

[0031] Specifically, the real-time reliability quantitative indicators of the elevator are graded and determined: when the risk level index exceeds the first warning threshold, the yellow warning mark is triggered; when the risk level index exceeds the second warning threshold or the remaining effective operating time is lower than the preset safety margin, the red warning mark is triggered; and the warning level mark is obtained; The fault component location rule base is matched and the preset fault location knowledge base is called according to the warning level identifier: when the yellow warning is issued, a list of single fault components corresponding to the risk characteristics is output; when the red warning is issued, a list of coupled fault components and the failure probability ranking are output; and a fault location report is obtained; Maintenance strategy knowledge base mapping, based on the fault location report and the remaining effective operating time: query the preset maintenance strategy knowledge base to match the historical handling plan for similar faults; when the remaining time falls below the critical value, automatically upgrade the urgency of the maintenance measures and obtain the maintenance instruction set; The early warning instructions are dynamically encapsulated, and the warning level identification, fault location report, and maintenance instruction set are encapsulated into a structured fault early warning instruction, and distributed through the following channels: sending equipment control instructions (such as speed reduction and elevator stop) to the elevator monitoring system; pushing work orders containing location reports and maintenance instructions to the maintenance terminal; and obtaining fault early warning instructions.

[0032] The above describes the elevator predictive maintenance method in the embodiment of the present invention. The following describes the elevator predictive maintenance system in the embodiment of the present invention. Figure 2In one embodiment of the present invention, an elevator predictive maintenance system includes multiple sensors installed on the elevator (represented by sensor assembly 101 in the figure), a data acquisition and transmission module 102, a data processing and prediction module 103, and an alarm and interaction module 104. These components work together to monitor the elevator's operating status and predict faults, as follows: Sensor assembly 101: The sensor assembly in this embodiment consists of several different types of sensors, installed throughout the elevator car interior and related mechanical components. Preferably, a triaxial vibration sensor is installed near the elevator traction motor and guide rails to capture vibration signals from the traction motor and the car's vertical movement. Environmental sensors are installed on the top or side walls of the elevator car to monitor the temperature, humidity, and noise level within the hoistway. Load sensors (such as strain gauges or pressure sensors) are installed on the car floor or suspension to measure changes in car load or cable tension. These sensors are connected via wires to a nearby data collection module or transmit data to a centralized data collection device via a wireless sensor network.

[0033] Data acquisition and transmission module 102: The acquisition module can be deployed on top of the elevator car or in the machine room, serving as the interface between the sensors and the central processing unit. This module includes a multi-channel analog signal conditioning circuit, an A / D converter, a microprocessor, and a communication unit. The analog signal conditioning circuit amplifies, filters, and level-converts the sensor input to prepare it for A / D conversion. The microprocessor periodically reads the sensor values ​​from each channel according to a predetermined program, attaches a unified timestamp to each reading, and then encapsulates the data into data packets. The communication unit then sends the data packets to the data processing module 103. Communication can be achieved via a wired network (such as Ethernet, CAN bus) or a wireless network (such as Wi-Fi, 4G LTE). If wireless transmission is used, to ensure data security and reliability, the system uses the TCP / IP protocol with encryption and authentication mechanisms to prevent data loss or tampering. In this embodiment, the data acquisition module transmits data in real time to a server in the elevator management center via industrial Ethernet.

[0034] Data Processing and Prediction Module 103: This module can be implemented by a server or cloud platform at the elevator monitoring center, or by edge computing devices installed on-site at the elevator. This embodiment uses a cloud server as the primary data processing unit. After receiving data from the acquisition module, the server writes it to a database for storage and invokes a reliability prediction algorithm for analysis and calculation. To improve processing efficiency, the server software architecture can utilize distributed parallel processing technology to handle data aggregation from multiple elevators. During the model analysis phase, the server utilizes a stored database of elevator historical operations to continuously train and optimize the reliability prediction model to adapt to changes in the environment and equipment status. When new data arrives, the latest model is immediately run for evaluation, and the results are stored in a prediction result database. During the model calculation process, if certain patterns in the data are identified that match known fault precursors (for example, an increase in a specific frequency component in the vibration signal corresponds to bearing wear), the system can also provide a preliminary diagnosis of the fault type. The data processing module records the results of each model prediction, including the time, key sensor indicators, and model output values, to form a log of elevator reliability changes over time.

[0035] Alarm and Interaction Module 104: This module consists of an alarm device and user interface software. The alarm device includes an on-site indicator alarm and a remote notification system. When the data processing module determines that an alarm needs to be triggered, the on-site audible and visual alarms are immediately activated (for example, flashing warning lights and audible buzzers in the machine room or monitoring room) to alert on-site personnel. Simultaneously, the server pushes an alarm message to the relevant responsible person's terminal via the network. In this embodiment, a mobile app has been developed as a mobile interface for operations and maintenance personnel. Upon receiving elevator alarm information pushed by the server, the app displays a pop-up notification on the phone, accompanied by an audible alert, allowing maintenance personnel to quickly understand the elevator status. Clicking on the notification opens the detailed page for the corresponding elevator within the app, displaying real-time sensor data curves, failure probability or health index calculated by the prediction model, recommended components for inspection, and historical similar alarm records. The monitoring client software in the management center can intuitively display an overview of the operating status of multiple elevators, using different colors or icons to identify normal and abnormal elevators. Operations and maintenance personnel can also manually query the sensor data and reliability assessment results of any elevator through the interactive interface, or adjust the parameter thresholds of the model (requires permission control). This module ensures smooth information transmission between the system and users, facilitating the execution of maintenance decisions.

[0036] Another embodiment of the elevator predictive maintenance system in the embodiment of the present invention further includes: The user terminal monitoring module 105: the reliability prediction model needs to be trained to achieve good prediction ability at the initial deployment. The training process can be carried out offline with the support of historical data or simulation data. For example, the multi-modal sensor data collected during the long-time operation of the target elevator and the same type of elevator, and various fault records occurring during the operation are collected as training samples. The data is input into a machine learning model (such as a random forest, a support vector machine, or a deep neural network), and the occurrence of a fault or the running time before the fault occurs is used as a label to supervise the training of the model, so that the model learns to capture fault signs from the data pattern. After the training is completed, the model is deployed to the elevator reliability prediction system for online prediction. In addition, the system has self-learning ability: as the elevator runs, a large amount of actual data is continuously accumulated, and the system can periodically retrain or update the parameters of the model using new data. This adaptive mechanism can gradually adapt the model to individual differences and environmental changes of the elevator, improving the prediction accuracy of long-term operation. For example, the model may initially be trained based on general data, without fully considering the special vibration characteristics of a specific elevator. After incrementally learning the measured data of the elevator, the model will be more consistent with the actual situation of the elevator, reducing false positives and false negatives. As can be seen, through model training and continuous optimization, the elevator reliability prediction system of the present application will become more and more intelligent and accurate. In the embodiments of the present application, the elevator reliability prediction platform can be built by using existing mature sensors, data acquisition devices and computing platforms, combined with advanced data analysis algorithms. In practical applications, the present application can be independently deployed on a single elevator to realize local monitoring, or can be extended to a regional elevator Internet of Things monitoring network to gather data of multiple elevators to a cloud platform for unified analysis, forming a group of elevators health management system. The present application has important significance for ensuring the safe operation of elevators and optimizing maintenance resources, and can be widely applied to intelligent elevator monitoring systems of elevator manufacturers, maintenance service providers and elevator supervision departments.

[0037] The present application also provides an elevator predictive maintenance device, which comprises a memory and a processor, the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute the steps of the elevator predictive maintenance method in each embodiment.

[0038] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium, and the computer readable storage medium stores instructions, which, when executed on a computer, make the computer execute the steps of the elevator predictive maintenance method.

[0039] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, and will not be described here again.

[0040] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0041] The above-described and the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predictive maintenance of an elevator, characterized in that: The elevator predictive maintenance method includes: Acquire multimodal sensor data during elevator operation, and transmit the multimodal sensor data to a data processing unit in real time through a data acquisition device; The data processing unit preprocesses and fuses the received multimodal sensor data, including signal filtering, data alignment and feature extraction, to obtain characteristic parameters representing the elevator operation status; Based on a pre-built elevator reliability prediction model, the characteristic parameters are calculated and analyzed to obtain the current reliability assessment result and / or fault prediction information of the elevator; Determine whether the reliability assessment result has any abnormal situation exceeding the preset threshold; if so, generate corresponding fault warning information and maintenance suggestion output.

2. The elevator predictive maintenance method according to claim 1, characterized in that: The data processing unit synchronizes the data of different sensors according to timestamps and extracts the elevator operation status characteristics, which include the frequency domain energy distribution of the vibration signal, the sound pressure level of the noise signal, the fluctuation amplitude of the operating load, and the change trend of the ambient temperature.

3. The elevator predictive maintenance method according to claim 1, characterized in that: The elevator reliability prediction model is constructed using a machine learning algorithm, which is trained by inputting historical multimodal sensor data and corresponding fault occurrence situations, and can output a probability value of elevator fault occurrence, a health status score, or a remaining life estimate. The machine learning algorithm includes a random forest algorithm, a support vector machine algorithm, or a deep neural network algorithm.

4. The elevator predictive maintenance method according to claim 1, characterized in that: include: Deploy four types of sensors in key fault-sensitive areas of the elevator system; Sampling is performed using a multi-channel data acquisition device with configurable sampling rate, and a unified microsecond-level timestamp is added to all sampling points based on the hardware clock; Construct parallel transmission channels, including a primary channel and a backup channel. The primary channel connects to the industrial Ethernet via a shielded twisted pair cable and uses Manchester encoding to eliminate common-mode interference. The backup channel encapsulates data packets via a 4G / 5G module and adds link-layer AES-256 encryption and CRC checksum. Verify the timestamp continuity and reorder the out-of-order data packets according to the hardware clock. If the packet loss rate is greater than 0.1%, the acquisition module is automatically triggered to perform incremental retransmission and output the multi-source original sensor data stream.

5. The elevator predictive maintenance method according to claim 4, characterized in that: include: Apply 20-500Hz mechanical resonance bandpass filtering to the vibration signal to extract the energy of the characteristic frequency bands of bearings and gears. Perform A-weighted filtering and energy integration of the human ear-sensitive frequency band on the noise signal to quantify the intensity of abnormal noise. Calculate the standard deviation of the load change rate throughout the entire operation from the load signal, identify abnormal loading and unloading patterns, fit the brake surface temperature rise curve to the temperature signal, and extract the temperature rise rate per minute; Based on the microsecond timestamp of the hardware clock, the four types of features, namely vibration spectrum energy, noise intensity, load fluctuation value, and temperature rise rate, are uniformly mapped to a 10ms time window to generate a synchronous feature matrix. The synchronous feature matrix is ​​compressed through principal component analysis, the principal components with cumulative contribution rate ≥ 95% are retained, and the multi-dimensional elevator state feature vector is output.

6. The elevator predictive maintenance method according to claim 5, characterized in that: include: Decompose the multi-dimensional elevator state feature vector into independent physical parameters and compare them with the preset fault judgment threshold to obtain a set of fault feature identifications; According to the number of concurrent fault types in the fault feature identification set, it is mapped to the preset risk level to obtain the risk level index of the entire machine; Based on the set of fault signatures and historical maintenance data: For each type of fault signature, the component-level remaining life attenuation coefficient is calculated based on the percentage exceeding the threshold, and the minimum value of all attenuation coefficients is taken as the remaining effective operating time benchmark value; According to the risk level index and the remaining effective operating time benchmark value, the preset maintenance strategy matrix is ​​matched to generate a real-time reliability quantitative index of the elevator including risk level, remaining time and maintenance instructions.

7. The elevator predictive maintenance method according to claim 6, characterized in that: include: When the risk level index exceeds the first warning threshold, the yellow warning mark is triggered; when the risk level index exceeds the second warning threshold or the remaining effective operating time is lower than the preset safety margin, the red warning mark is triggered; obtain the warning level mark; According to the warning level identification, the preset fault location knowledge base is called to obtain the fault location report; Based on the fault location report and the remaining effective running time, a maintenance instruction set is obtained; The warning level identification, fault location report, and maintenance instruction set are encapsulated as a fault warning instruction.

8. An elevator predictive maintenance system, characterized in that: The elevator predictive maintenance system includes: Multimodal sensor components, installed on the elevator car and its related components, are used to collect various status data of the elevator operation; The data acquisition and transmission module is connected to the sensor assembly, acquires the measurement data of each sensor in real time, performs preliminary processing, and then sends the data to the designated data processing device through the communication network; A data processing and prediction module includes at least one processor and a memory, wherein the memory stores an elevator reliability prediction model program, and the processor is configured to execute the program to perform data fusion and reliability prediction analysis on received multimodal sensor data to generate an elevator operation reliability assessment result; The alarm and interaction module is used to receive the evaluation results and issue a fault warning signal when the results are abnormal, provide a human-computer interaction interface to display elevator status information and prediction reports, and support notification of warning information to maintenance personnel.

9. The elevator predictive maintenance system according to claim 8, characterized in that: The data processing and prediction module is deployed on a cloud server or the main control computer of an elevator monitoring center, and can centrally analyze data from multiple elevators. The data processing and prediction module uses a distributed database and a parallel computing framework to store and process historical elevator operation data, and regularly updates and trains the elevator reliability prediction model to improve prediction accuracy and system adaptability.

10. The elevator predictive maintenance system according to claim 8, characterized in that: The alarm and interaction module includes a mobile application interface. When the result output by the elevator reliability prediction model exceeds a threshold, the application will generate a fault warning message and push it to the mobile communication device of at least one elevator maintenance personnel so that they can understand the elevator status in time and take maintenance measures.

Images