Method for detecting various abnormal states of elevator driving host

Abstract

The invention relates to the technical field of elevator operation and maintenance, and provides a method for detecting various abnormal states of an elevator driving host, which comprises the following steps of: acquiring operation data of various types of elevator driving hosts, extracting feature data of the elevator driving hosts from preprocessed data of the operation data of the various types of elevator driving hosts by using a convolutional neural network; training a fault recognition algorithm initial model in an initial state, receiving operation data of various types of elevator driving hosts collected in real time to carry out transfer learning on the fault recognition algorithm initial model, and optimizing the model through a gradient descent algorithm to obtain an optimized fault recognition algorithm model; and the various abnormal state detection results of the elevator driving host are substituted into a preset result display system to be displayed in real time. The problems that one set of equipment is generally only suitable for one type of elevators, the detection time is long, and the requirement for rapid and accurate detection cannot be met are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator operation and maintenance, and in particular to a method for detecting multiple abnormal states of an elevator drive host. Background Art

[0002] As an indispensable means of transportation in modern urban life, the safety and comfort of elevators are of vital importance to passengers. However, if elevators are not properly maintained during long-term use, their overall performance and safety will gradually decline, and may even cause serious personal injury accidents. Therefore, comprehensive monitoring and maintenance of elevators is particularly important.

[0003] Currently, many elevator companies have deployed a variety of sensors and small servers in the elevator car. These devices can collect and analyze data such as vibration, speed and height of the elevator in real time to ensure the normal operation of the elevator. However, these monitoring systems are mainly focused on the elevator car and control system, but rarely pay attention to the elevator drive host. In fact, as the core component of the elevator, the performance and status of the drive host are directly related to the overall operation and safety of the elevator.

[0004] Currently, the detection equipment for elevator drive mainframes on the market is not only scarce in number, but also relatively simple in function, and can often only detect a specific problem such as bearing failure or demagnetization. At the same time, these detection equipment have weak generalization and migration capabilities, and a set of equipment can usually only be applied to one model of elevator, which undoubtedly increases the cost and difficulty of maintenance. In addition, the existing elevator drive mainframe bearing fault detection technology is complex to operate and takes a long time to detect, which cannot meet the needs of fast and accurate detection. Summary of the invention

[0005] The purpose of the present invention is to provide a method for detecting multiple abnormal states of an elevator drive host, so as to solve the problem that a set of equipment is usually only applicable to one type of elevator, the detection time is long, and the demand for fast and accurate detection cannot be met.

[0006] The present invention provides a method for detecting multiple abnormal states of an elevator drive host, including:

[0007] Acquire operation data of various types of elevator drive hosts, wherein the operation data of various types of elevator drive hosts include basic operation data of elevators, elevator performance parameter data, elevator environment and safety data, data on the number of elevator users and frequency of use, and elevator technical parameters, and pre-process the operation data of various types of elevator drive hosts to obtain pre-processed operation data of various types of elevator drive hosts;

[0008] Use convolutional neural networks to extract elevator drive host feature data from the preprocessed data of elevator drive host operation data of various models, and establish elevator operation data labels for the elevator drive host feature data;

[0009] Perform a transfer learning algorithm on one elevator model in the elevator drive host feature data, and transfer the knowledge learned on one elevator model through the transfer learning algorithm to other elevator models;

[0010] Retrieve the data information corresponding to the elevator operation data label, train the initial model of the fault identification algorithm in the initial state, receive the real-time collected operation data of various types of elevator drive hosts to perform transfer learning of the initial model of the fault identification algorithm, and optimize the model through the gradient descent algorithm to obtain the optimized fault identification algorithm model;

[0011] Continuously receive the real-time collected operating data of various models of elevator drive hosts, substitute the continuously received real-time collected operating data of various models of elevator drive hosts into the optimized fault identification algorithm model, obtain the various abnormal state detection results of the elevator drive hosts, substitute the various abnormal state detection results of the elevator drive hosts into the preset result display system for real-time display.

[0012] Further, the operation data of the drive host of multiple models of elevators are obtained, wherein the operation data of the drive host of multiple models of elevators include basic operation data of elevators, performance parameter data of elevators, environment and safety data of elevators, number of elevator users and frequency of use data, and technical parameters of elevators. The operation data of the drive host of multiple models of elevators are preprocessed to obtain preprocessed operation data of the drive host of multiple models of elevators, including:

[0013] The data of various types of elevator drive host operation data are cleaned and preliminarily screened to obtain the first preliminarily processed elevator operation data, and the preliminarily processed elevator operation data are detected for outliers using the interquartile range to obtain the elevator operation abnormality data;

[0014] removing the abnormal elevator operation data in the preliminary processed elevator operation data to obtain second preliminary processed elevator operation data;

[0015] Based on the normal operation mode and abnormal mode of the elevator, the second preliminary processing elevator operation data is identified and noise data is removed to obtain pre-processed data of operation data of elevator drive hosts of various models.

[0016] Furthermore, a convolutional neural network is used to extract elevator drive host feature data from preprocessed data of elevator drive host operation data of various models, and an elevator operation data label is established for the elevator drive host feature data, including:

[0017] Using a convolutional neural network to extract elevator drive host feature data from preprocessed data of elevator drive host operation data of various models, the elevator drive host feature data includes operation status features, elevator performance parameter features, elevator environment and safety features, elevator use frequency and mode features, elevator technical parameter features, elevator fault and abnormality features, and elevator image and video features;

[0018] Corresponding tags are established to extract the characteristic data of the elevator drive host, and the tags include operating status and fault type.

[0019] Further, a transfer learning algorithm is performed on one elevator model in the elevator drive host characteristic data, and the knowledge learned on one elevator model through the transfer learning algorithm is transferred to the remaining elevator models, including;

[0020] Determine the target elevator model that needs to transfer knowledge, use convolutional neural network to extract key features from the feature data of the source elevator model, and train the CNN model on the feature data of the source elevator model;

[0021] Use the trained CNN model as the initial model of the target elevator model, evaluate the model performance after transfer learning on the feature data of the target elevator model, and optimize the model based on the evaluation results;

[0022] The knowledge learned from one elevator model is successfully transferred to other models. At this point, the model of the target elevator model already has the knowledge and feature representation capabilities learned by the source elevator model model.

[0023] Verify the performance of the model after transfer learning in the actual environment, and adjust and optimize it as needed. If the model performance meets expectations, it can be deployed to the actual application system.

[0024] Further, the data information corresponding to the elevator operation data label is retrieved, the initial model of the fault identification algorithm in the initial state is trained, the operation data of the elevator drive host collected in real time of various models are received to perform transfer learning of the initial model of the fault identification algorithm, and the model is optimized by the gradient descent algorithm to obtain the optimized fault identification algorithm model, including;

[0025] Retrieve the data information corresponding to the elevator operation data tag, and select the historical operation data with the fault tag. The historical operation data with the fault tag contains various parameters of the elevator in normal operation and fault state, including speed, acceleration, current, and voltage;

[0026] Clean the selected data to remove outliers, duplicate values, and missing values;

[0027] Label the data and clearly mark the fault status and normal operation status;

[0028] Based on the cleaned data, extract the features related to elevator failure, including the average speed and acceleration mutations in a specific time period;

[0029] Construct the initial model of fault identification algorithm based on support vector machine;

[0030] Use the labeled data set to train the initial model, and adjust the initial model parameters of the fault identification algorithm to achieve the best initial performance of the initial model of the fault identification algorithm;

[0031] Through sensors or other data acquisition equipment, the operating data of various types of elevator drive hosts are collected in real time. According to the characteristics of real-time collection of operating data of various types of elevator drive hosts, the initial model parameters of the fault identification algorithm are fine-tuned to adapt to different types of elevator drive hosts.

[0032] The initial model parameters of the fault identification algorithm are optimized through the gradient descent algorithm. According to the calculated gradient, the parameters of the model are updated using the gradient descent algorithm. The above process is repeated until the model performance reaches the preset standard or the number of iterations reaches the upper limit, and the optimized fault identification algorithm model is obtained.

[0033] Further, continuously receiving the real-time collected operation data of various types of elevator drive hosts, substituting the real-time collected operation data of various types of elevator drive hosts into the optimized fault identification algorithm model, obtaining the detection results of various abnormal states of the elevator drive hosts, substituting the detection results of various abnormal states of the elevator drive hosts into a preset result display system for real-time display, including;

[0034] Continuously receive real-time data of various elevator drive host operation, verify the received data, detect the integrity of the data, and remove data anomalies caused by transmission errors or equipment failures;

[0035] Substitute the continuously received real-time data into the optimized fault identification algorithm model, use the optimized fault identification algorithm model to analyze and predict the input real-time data, and identify the abnormal state of the elevator drive host;

[0036] The prediction results of the model are parsed into specific abnormal status information, including motor overheating, transmission system failure, and load abnormality.

[0037] Further, the operation data of the drive host of multiple models of elevators are obtained, wherein the operation data of the drive host of multiple models of elevators include basic operation data of elevators, performance parameter data of elevators, environment and safety data of elevators, number of elevator users and frequency of use data, and technical parameters of elevators. The operation data of the drive host of multiple models of elevators are preprocessed to obtain preprocessed operation data of the drive host of multiple models of elevators, including:

[0038] The signals in the operating data of various types of elevator drive hosts are decomposed into several IMF components through CEEMDAN.

[0039] The correlation coefficients of several IMF components are screened and the wavelet phase is used for noise reduction. The characteristic frequencies are screened to obtain the reconstructed IMF components and output the noise-reduced signal.

[0040] The beneficial effects of the present invention are as follows: The present invention provides a method for detecting multiple abnormal states of an elevator drive host, and a new device for detecting abnormal states of an elevator drive host. The device can monitor multiple parameters of the drive host in real time, such as temperature, vibration, noise, etc., and analyze and process the data through advanced algorithms to timely discover and warn of potential faults. This detection device is not only comprehensive in function, but also has good generalization and migration capabilities, and can adapt to different models of elevators. At the same time, it is simple to operate and has a short detection time, which greatly improves the efficiency and accuracy of elevator maintenance. We believe that this new device for detecting abnormal states of elevator drive hosts will provide a strong guarantee for the safe operation of elevators.

[0041] Synchronous sampling and multifunctional detection: Through seven-channel synchronous sampling technology, the voltage, current and vibration signals are fully and accurately captured, providing a rich and accurate data basis for subsequent fault detection. At the same time, the device can simultaneously detect multiple abnormal conditions such as bearing faults and demagnetization, greatly improving the detection efficiency and practicality.

[0042] High-precision data acquisition and processing: The use of high-performance data acquisition cards and advanced signal processing technology ensures the accuracy of data acquisition and signal processing. This can not only accurately reflect the actual operating status of the elevator drive host, but also provide a reliable basis for subsequent fault identification and analysis.

[0043] Intelligent fault diagnosis and display: Through the designed multi-channel WDCNN model and domain adversarial neural network structure, intelligent feature extraction and fault diagnosis are realized. This can not only quickly and accurately identify the fault type, but also intuitively display the identification results on the interface, providing great convenience for maintenance personnel.

[0044] Wide applicability: The present invention is not only applicable to elevator drive hosts of specific models, but can also adapt to elevators of different models and working conditions through domain adversarial neural networks. This makes the device widely applicable and has strong migration capabilities, reducing dependence on specific elevator models. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 A flow chart of a method for detecting multiple abnormal states of an elevator drive host is provided for an embodiment of the present invention.

[0047] Figure 2 A schematic diagram of the overall framework of the method for detecting multiple abnormal states of an elevator drive host provided by the present invention.

[0048] Figure 3 This is a schematic diagram of the PXIe8584 board provided by the present invention.

[0049] Figure 4 This is a schematic diagram of the signal denoising process method provided by the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0051] See also Figures 1 to 4 The present invention provides a method for detecting multiple abnormal states of an elevator drive host, including:

[0052] The present invention aims to accurately detect whether the driving host has bearing failure or demagnetization by collecting the voltage, current and vibration signals of the driving host when the elevator is running, and further determine the type and degree of the fault, so as to solve the problems of single fault detection function and insufficient accuracy in the prior art.

[0053] The complete technical solution provided by the present invention

[0054] The present invention proposes a comprehensive and efficient elevator drive host abnormal state detection device, the overall structure of which is as follows: Figure 2As shown, it mainly includes a sensor module, an analog-to-digital conversion module and a host computer module, wherein the sensor module is used for the operation data of various types of elevator drive hosts, and some of the operation data of various types of elevator drive hosts are retrieved through a third-party data interface. The analog-to-digital conversion module is used to convert the electrical signal generated by the elevator operation into a digital signal, and the host computer module is used to process and analyze the corresponding data generated by the elevator.

[0055] The sensor module consists of a voltage sensor, a current sensor and an acceleration sensor, which can realize seven-channel synchronous sampling of the three-phase voltage, three-phase current and vibration signals of the drive host, ensuring accurate capture of various key information of the elevator permanent magnet synchronous motor during operation.

[0056] The analog-to-digital conversion module is responsible for receiving the analog voltage, current and vibration information transmitted by the sensor module and accurately converting it into digital quantities to provide accurate data for subsequent digital signal processing and analysis.

[0057] The host computer module is responsible for recording, processing and analyzing digital information. First, the received signal is denoised to improve the signal-to-noise ratio; then the characteristic information is extracted to provide a key basis for fault identification; finally, the fault type is accurately identified through advanced fault identification algorithms, and the identification results are displayed on the interface, providing maintenance personnel with intuitive and convenient fault location services.

[0058] In terms of data acquisition, the present invention uses a high-performance PXIe8584 board, see Figure 3 Its 8-channel 14-bit 100MS / s sampling rate can ensure accurate capture of high-frequency and high-dynamic range signals. At the same time, it is equipped with up to 2GB of onboard memory to meet the needs of long-term and large-volume data acquisition.

[0059] In terms of signal processing, the present invention adopts advanced denoising strategies such as adaptive noise complete set empirical mode decomposition and wavelet transform to effectively remove environmental noise and ensure the purity of the signal. At the same time, by designing a multi-channel WDCNN model and a domain adversarial neural network structure, intelligent feature extraction and fault diagnosis are realized, which greatly improves the accuracy and speed of fault identification.

[0060] In summary, the present invention achieves efficient and accurate detection of abnormal status of elevator drive host by integrating advanced technology and intelligent algorithms. This not only improves the safety and reliability of elevator operation, but also provides maintenance personnel with convenient and efficient fault location services.

[0061] See also Figure 1 The present invention provides a method for detecting multiple abnormal states of an elevator drive host, including:

[0062] Acquire operation data of various types of elevator drive hosts, wherein the operation data of various types of elevator drive hosts include basic operation data of elevators, elevator performance parameter data, elevator environment and safety data, data on the number of elevator users and frequency of use, and elevator technical parameters, and pre-process the operation data of various types of elevator drive hosts to obtain pre-processed operation data of various types of elevator drive hosts;

[0063] Use convolutional neural networks to extract elevator drive host feature data from the preprocessed data of elevator drive host operation data of various models, and establish elevator operation data labels for the elevator drive host feature data;

[0064] Perform a transfer learning algorithm on one elevator model in the elevator drive host feature data, and transfer the knowledge learned on one elevator model through the transfer learning algorithm to other elevator models;

[0065] Retrieve the data information corresponding to the elevator operation data label, train the initial model of the fault identification algorithm in the initial state, receive the real-time collected operation data of various types of elevator drive hosts to perform transfer learning of the initial model of the fault identification algorithm, and optimize the model through the gradient descent algorithm to obtain the optimized fault identification algorithm model;

[0066] Continuously receive the real-time collected operating data of various models of elevator drive hosts, substitute the continuously received real-time collected operating data of various models of elevator drive hosts into the optimized fault identification algorithm model, obtain the various abnormal state detection results of the elevator drive hosts, substitute the various abnormal state detection results of the elevator drive hosts into the preset result display system for real-time display.

[0067] Specifically, the operation data of the elevator drive host of various models are obtained, wherein the operation data of the elevator drive host of various models include basic operation data of the elevator, performance parameter data of the elevator, environment and safety data of the elevator, number of elevator users and frequency of use data, and technical parameters of the elevator. The operation data of the elevator drive host of various models are pre-processed to obtain pre-processed operation data of the elevator drive host of various models, including:

[0068] The data of various types of elevator drive host operation data are cleaned and preliminarily screened to obtain the first preliminarily processed elevator operation data, and the preliminarily processed elevator operation data are detected for outliers using the interquartile range to obtain the elevator operation abnormality data;

[0069] removing the abnormal elevator operation data in the preliminary processed elevator operation data to obtain second preliminary processed elevator operation data;

[0070] Based on the normal operation mode and abnormal mode of the elevator, the second preliminary processing elevator operation data is identified and noise data is removed to obtain pre-processed data of operation data of elevator drive hosts of various models.

[0071] Specifically, a convolutional neural network is used to extract elevator drive host feature data from the preprocessed data of elevator drive host operation data of various models, and an elevator operation data label is established for the elevator drive host feature data, including;

[0072] Using a convolutional neural network to extract elevator drive host feature data from preprocessed data of elevator drive host operation data of various models, the elevator drive host feature data includes operation status features, elevator performance parameter features, elevator environment and safety features, elevator use frequency and mode features, elevator technical parameter features, elevator fault and abnormality features, and elevator image and video features;

[0073] Corresponding tags are established to extract the characteristic data of the elevator drive host, and the tags include operating status and fault type.

[0074] Specifically, a transfer learning algorithm is used for one elevator model in the feature data of the elevator drive host, and the knowledge learned on one elevator model through the transfer learning algorithm is transferred to the remaining elevator models, including;

[0075] Determine the target elevator model that needs to transfer knowledge, use convolutional neural network to extract key features from the feature data of the source elevator model, and train the CNN model on the feature data of the source elevator model;

[0076] Use the trained CNN model as the initial model of the target elevator model, evaluate the model performance after transfer learning on the feature data of the target elevator model, and optimize the model based on the evaluation results;

[0077] The knowledge learned from one elevator model is successfully transferred to other models. At this point, the model of the target elevator model already has the knowledge and feature representation capabilities learned by the source elevator model model.

[0078] Verify the performance of the model after transfer learning in the actual environment, and adjust and optimize it as needed. If the model performance meets expectations, it can be deployed to the actual application system.

[0079] Specifically, retrieve the data information corresponding to the elevator operation data label, train the initial model of the fault identification algorithm in the initial state, receive the real-time collected operation data of various types of elevator drive host machines to perform transfer learning of the initial model of the fault identification algorithm, optimize the model through the gradient descent algorithm, and obtain the optimized fault identification algorithm model, including;

[0080] Retrieve the data information corresponding to the elevator operation data tag, and select the historical operation data with the fault tag. The historical operation data with the fault tag contains various parameters of the elevator in normal operation and fault state, including speed, acceleration, current, and voltage;

[0081] Clean the selected data to remove outliers, duplicate values, and missing values;

[0082] Label the data and clearly mark the fault status and normal operation status;

[0083] Based on the cleaned data, extract the features related to elevator failure, including the average speed and acceleration mutations in a specific time period;

[0084] Construct the initial model of fault identification algorithm based on support vector machine;

[0085] Use the labeled data set to train the initial model, and adjust the initial model parameters of the fault identification algorithm to achieve the best initial performance of the initial model of the fault identification algorithm;

[0086] Through sensors or other data acquisition equipment, the operating data of various types of elevator drive hosts are collected in real time. According to the characteristics of real-time collection of operating data of various types of elevator drive hosts, the initial model parameters of the fault identification algorithm are fine-tuned to adapt to different types of elevator drive hosts.

[0087] The initial model parameters of the fault identification algorithm are optimized through the gradient descent algorithm. According to the calculated gradient, the parameters of the model are updated using the gradient descent algorithm. The above process is repeated until the model performance reaches the preset standard or the number of iterations reaches the upper limit, and the optimized fault identification algorithm model is obtained.

[0088] Specifically, the operation data of various types of elevator drive hosts collected in real time are continuously received, and the operation data of various types of elevator drive hosts collected in real time are substituted into the optimized fault identification algorithm model to obtain various abnormal state detection results of the elevator drive hosts, and the various abnormal state detection results of the elevator drive hosts are substituted into the preset result display system for real-time display, including;

[0089] Continuously receive real-time data of various elevator drive host operation, verify the received data, detect the integrity of the data, and remove data anomalies caused by transmission errors or equipment failures;

[0090] Substitute the continuously received real-time data into the optimized fault identification algorithm model, use the optimized fault identification algorithm model to analyze and predict the input real-time data, and identify the abnormal state of the elevator drive host;

[0091] The prediction results of the model are parsed into specific abnormal status information, including motor overheating, transmission system failure, and load abnormality.

[0092] Specifically, the operation data of the elevator drive host of various models are obtained, wherein the operation data of the elevator drive host of various models include basic operation data of the elevator, performance parameter data of the elevator, environment and safety data of the elevator, number of elevator users and frequency of use data, and technical parameters of the elevator. The operation data of the elevator drive host of various models are pre-processed to obtain pre-processed operation data of the elevator drive host of various models, including:

[0093] The signals in the operating data of various types of elevator drive hosts are decomposed into several IMF components through CEEMDAN.

[0094] The correlation coefficients of several IMF components are screened and the wavelet phase is used for noise reduction. The characteristic frequencies are screened to obtain the reconstructed IMF components and output the noise-reduced signal.

[0095] It can be seen from the above embodiments that the present invention provides a method for detecting multiple abnormal states of an elevator drive host, obtains operation data of multiple models of elevator drive hosts, pre-processes the operation data of multiple models of elevator drive hosts, obtains pre-processed data of operation data of multiple models of elevator drive hosts, uses a convolutional neural network to extract elevator drive host feature data from the pre-processed data of operation data of multiple models of elevator drive hosts, and establishes elevator operation data labels for the elevator drive host feature data, performs a transfer learning algorithm on an elevator model in the elevator drive host feature data, transfers the knowledge learned on one elevator model through the transfer learning algorithm to the remaining elevator models, retrieves data information corresponding to the elevator operation data label, trains an initial model of the fault recognition algorithm in the initial state, receives the operation data of multiple models of elevator drive hosts collected in real time, performs transfer learning of the initial model of the fault recognition algorithm, optimizes the model through the gradient descent algorithm, obtains an optimized fault recognition algorithm model, continuously receives the operation data of multiple models of elevator drive hosts collected in real time, substitutes the operation data of multiple models of elevator drive hosts collected in real time into the optimized fault recognition algorithm model, obtains the detection results of multiple abnormal states of the elevator drive host, and substitutes the detection results of multiple abnormal states of the elevator drive host into a preset result display system for real-time display. It solves the problem that a set of equipment is usually only applicable to one type of elevator, the detection time is long, and it cannot meet the needs of fast and accurate detection.

[0096] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention or some parts of the embodiments.

[0097] The above-described embodiments of the present invention do not limit the protection scope of the present invention.

Claims

1. A method for detecting various abnormal states of an elevator drive host, characterized in that: include; Acquire operation data of various types of elevator drive hosts, wherein the operation data of various types of elevator drive hosts include basic operation data of elevators, elevator performance parameter data, elevator environment and safety data, data on the number of elevator users and frequency of use, and elevator technical parameters, and pre-process the operation data of various types of elevator drive hosts to obtain pre-processed operation data of various types of elevator drive hosts; Use convolutional neural networks to extract elevator drive host feature data from the preprocessed data of elevator drive host operation data of various models, and establish elevator operation data labels for the elevator drive host feature data; Perform a transfer learning algorithm on one elevator model in the elevator drive host feature data, and transfer the knowledge learned on one elevator model through the transfer learning algorithm to other elevator models; Retrieve the data information corresponding to the elevator operation data label, train the initial model of the fault identification algorithm in the initial state, receive the real-time collected operation data of various types of elevator drive hosts to perform transfer learning of the initial model of the fault identification algorithm, and optimize the model through the gradient descent algorithm to obtain the optimized fault identification algorithm model; Continuously receive the real-time collected operating data of various models of elevator drive hosts, substitute the continuously received real-time collected operating data of various models of elevator drive hosts into the optimized fault identification algorithm model, obtain the various abnormal state detection results of the elevator drive hosts, substitute the various abnormal state detection results of the elevator drive hosts into the preset result display system for real-time display.

2. The method according to claim 1, characterized in that S101: Acquire operation data of multiple types of elevator drive hosts, wherein the operation data of multiple types of elevator drive hosts include basic operation data of elevators, elevator performance parameter data, elevator environment and safety data, data on the number of elevator users and frequency of use, and elevator technical parameters, and pre-process the operation data of multiple types of elevator drive hosts to obtain pre-processed operation data of multiple types of elevator drive hosts, including: The data of various types of elevator drive host operation data are cleaned and preliminarily screened to obtain the first preliminarily processed elevator operation data, and the preliminarily processed elevator operation data are detected for outliers using the interquartile range to obtain the elevator operation abnormality data; removing the abnormal elevator operation data in the preliminary processed elevator operation data to obtain second preliminary processed elevator operation data; Based on the normal operation mode and abnormal mode of the elevator, the second preliminary processing elevator operation data is identified and noise data is removed to obtain pre-processed data of operation data of elevator drive hosts of various models.

3. The method according to claim 1, characterized in that S102: extracting elevator drive host feature data from the preprocessed data of elevator drive host operation data of various models using a convolutional neural network, and establishing elevator operation data labels for the elevator drive host feature data, including: Using a convolutional neural network to extract elevator drive host feature data from preprocessed data of elevator drive host operation data of various models, the elevator drive host feature data includes operation status features, elevator performance parameter features, elevator environment and safety features, elevator use frequency and mode features, elevator technical parameter features, elevator fault and abnormality features, and elevator image and video features; Corresponding tags are established to extract the characteristic data of the elevator drive host, and the tags include operating status and fault type.

4. The method according to claim 1, characterized in that S103: performing a transfer learning algorithm on an elevator model in the elevator drive host feature data, and transferring the knowledge learned on the elevator model through the transfer learning algorithm to the remaining elevator models, including; Determine the target elevator model that needs to transfer knowledge, use convolutional neural network to extract key features from the feature data of the source elevator model, and train the CNN model on the feature data of the source elevator model; Use the trained CNN model as the initial model of the target elevator model, evaluate the model performance after transfer learning on the feature data of the target elevator model, and optimize the model based on the evaluation results; The knowledge learned from one elevator model is successfully transferred to other models. At this point, the model of the target elevator model already has the knowledge and feature representation capabilities learned by the source elevator model model. Verify the performance of the model after transfer learning in the actual environment, and adjust and optimize it as needed. If the model performance meets expectations, it can be deployed to the actual application system.

5. The method according to claim 1, characterized in that S104: Retrieving data information corresponding to the elevator operation data label, training the initial model of the fault identification algorithm in the initial state, receiving the real-time collected operation data of various types of elevator drive hosts to perform transfer learning of the initial model of the fault identification algorithm, optimizing the model through the gradient descent algorithm, and obtaining an optimized fault identification algorithm model, including; Retrieve the data information corresponding to the elevator operation data tag, and select the historical operation data with the fault tag. The historical operation data with the fault tag contains various parameters of the elevator in normal operation and fault state, including speed, acceleration, current, and voltage; Clean the selected data to remove outliers, duplicate values, and missing values; Label the data and clearly mark the fault status and normal operation status; Based on the cleaned data, extract the features related to elevator failure, including the average speed and acceleration mutations in a specific time period; Construct the initial model of fault identification algorithm based on support vector machine; Use the labeled data set to train the initial model, and adjust the initial model parameters of the fault identification algorithm to achieve the best initial performance of the initial model of the fault identification algorithm; Through sensors or other data acquisition equipment, the operating data of various types of elevator drive hosts are collected in real time. According to the characteristics of real-time collection of operating data of various types of elevator drive hosts, the initial model parameters of the fault identification algorithm are fine-tuned to adapt to different types of elevator drive hosts. The initial model parameters of the fault identification algorithm are optimized through the gradient descent algorithm. According to the calculated gradient, the parameters of the model are updated using the gradient descent algorithm. The above process is repeated until the model performance reaches the preset standard or the number of iterations reaches the upper limit, and the optimized fault identification algorithm model is obtained.

6. The method according to claim 1, characterized in that S105: continuously receiving the real-time collected operation data of various types of elevator drive hosts, substituting the real-time collected operation data of various types of elevator drive hosts into the optimized fault identification algorithm model, obtaining the detection results of various abnormal states of the elevator drive hosts, substituting the detection results of various abnormal states of the elevator drive hosts into a preset result display system for real-time display, including; Continuously receive real-time data of various elevator drive host operation, verify the received data, detect the integrity of the data, and remove data anomalies caused by transmission errors or equipment failures; Substitute the continuously received real-time data into the optimized fault identification algorithm model, use the optimized fault identification algorithm model to analyze and predict the input real-time data, and identify the abnormal state of the elevator drive host; The prediction results of the model are parsed into specific abnormal status information, including motor overheating, transmission system failure, and load abnormality.

7. The method according to claim 1, characterized in that Acquire operation data of various types of elevator drive hosts, wherein the operation data of various types of elevator drive hosts include basic operation data of elevators, elevator performance parameter data, elevator environment and safety data, data on the number of elevator users and frequency of use, and elevator technical parameters; pre-process the operation data of various types of elevator drive hosts to obtain pre-processed operation data of various types of elevator drive hosts, including: The signals in the operating data of various types of elevator drive hosts are decomposed into several IMF components through CEEMDAN. The correlation coefficients of several IMF components are screened and the wavelet phase is used for noise reduction. The characteristic frequencies are screened to obtain the reconstructed IMF components and output the noise-reduced signal.

Images