Escalator fault early warning method and system based on transfer learning

Abstract

The application discloses an escalator fault early warning method and system based on transfer learning, and the method comprises the following steps: collecting target escalator analysis characteristic data, wherein the analysis characteristic data comprises vibration data, temperature data and noise data; inputting the analysis characteristic data into a fault early warning identification model for analysis and identification, and determining a fault trend grade; when the fault trend grade exceeds a preset early warning threshold, outputting fault early warning data for fault risk early warning. The application has high stability, high accuracy and low cost, and can be widely applied to the technical field of escalator management systems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of escalator management system, and in particular to an escalator fault early warning method and system based on transfer learning. BACKGROUND

[0002] At present, the urban rail transit industry is developing rapidly, and with the continuous expansion of the scale of the public transportation system, higher performance requirements are put forward for special equipment used in the line network, especially for escalators. Thanks to the continuous operation and strong conveying capacity of escalators, escalators have become an important passenger transport equipment for people to enter and exit stations, transfer and transport goods. Because of the large passenger flow, the escalator is required to operate at high load for a long time, so it is inevitable to have wear and tear of different mechanical structures of the escalator, especially the transmission parts. Through real-time monitoring and risk assessment of the working state of the key mechanical structure of the escalator, timely fault early warning and predictive maintenance can be realized, which is of great significance for escalator maintenance and passenger safety accidents.

[0003] Existing research mainly installs vibration sensors on key components of the escalator to obtain escalator vibration information, extracts important features therefrom, and establishes a fault identification model for the escalator by combining a machine learning algorithm to realize fault warning. However, many times the escalator fails instantly, resulting in irreparable passenger injury accidents or serious mechanical wear and tear. At this time, it is too late to issue a warning and take remedial action. Therefore, some research identifies the escalator failure trend by analyzing the vibration information of the escalator and classifying the working condition of the escalator components before failure based on the extracted key feature information. When it is identified that the working condition is closest to the failure point, a warning is issued to realize the escalator failure trend early warning. However, this requires a large number of escalator samples and time for experiments, which is difficult for escalator manufacturers to achieve at the present stage. In addition, the above only uses the vibration information of the escalator for research and does not consider the sound and temperature characteristics of the escalator. However, noise and temperature are important indicators in traditional escalator fault identification. SUMMARY

[0004] Therefore, the present application provides an escalator fault early warning method and system based on transfer learning, which has high stability, high accuracy and low cost.

[0005] One aspect of the present application provides an escalator fault early warning method based on transfer learning, comprising:

[0006] Collecting target escalator analysis feature data, the analysis feature data including vibration data, temperature data and noise data;

[0007] inputting the feature data to be analyzed into a fault early warning identification model for analysis and identification to determine a fault trend level;

[0008] When the fault trend level exceeds a preset early warning threshold, outputting fault early warning data for fault risk early warning;

[0009] The fault early warning identification model is obtained through a migration learning technique.

[0010] Optionally, the feature data to be analyzed of the target escalator includes at least one of the following:

[0011] vibration data of the motor is collected through a vibration detection sensor;

[0012] vibration data of the speed reducer is collected through a vibration detection sensor;

[0013] vibration data of the anchor bolt is collected through a vibration detection sensor;

[0014] vibration data of the main drive wheel is collected through a vibration detection sensor;

[0015] vibration data of the tensioning sprocket is collected through a vibration detection sensor;

[0016] temperature data of the reducer oil is collected through a temperature sensor;

[0017] temperature data of the handrail belt is collected through a temperature sensor;

[0018] temperature data of the environment is collected through a temperature sensor;

[0019] noise data of the motor is collected through a noise detection sensor;

[0020] noise data of the main drive wheel is collected through a noise detection sensor;

[0021] noise data of the tensioning wheel is collected through a noise detection sensor;

[0022] noise data of the steps is collected through a noise detection sensor.

[0023] Optionally, the training process of the fault early warning identification model includes the following steps:

[0024] obtaining a fault trend grading model of similar equipment of the escalator equipment, wherein the similar equipment includes but is not limited to a turbine system, a straight escalator system and a conveyor belt system;

[0025] obtaining source domain training data according to the already trained fault trend grading model of the similar equipment, and obtaining target domain training data according to the escalator sample;

[0026] constructing a training set according to the source domain training data and the target domain training data, and randomly selecting a subset from the target domain training data as a validation set;

[0027] According to the training set and the validation set, the similar equipment fault trend classification model is migrated from the data level and the feature level, and a fault early warning identification model of escalator equipment is constructed.

[0028] Optionally, in the step of migrating the similar equipment fault trend classification model from the data level and the feature level according to the training set and the validation set, and constructing a fault early warning identification model of escalator equipment, migrating the similar equipment fault trend classification model from the data level comprises:

[0029] Taking the validation set as an optimization target, iteratively training according to the training set;

[0030] Adjusting the learning weight of each training data in the training set in each iteration training process to realize the migration of the source domain to the target domain fault trend classification model.

[0031] Optionally, in the step of migrating the similar equipment fault trend classification model from the data level and the feature level according to the training set and the validation set, and constructing a fault early warning identification model of escalator equipment, migrating the similar equipment fault trend classification model from the feature level comprises:

[0032] Finding the common feature representation between the source domain and the target domain;

[0033] Mapping the source domain training data and the target domain training data from the original feature space to a new feature space, wherein the data distribution of the source domain training data and the target domain training data in the new feature space is the same;

[0034] Migrating the similar equipment fault trend classification model based on the new feature space.

[0035] Optionally, the method further comprises:

[0036] Collecting the to-be-analyzed feature data of the target escalator through an Internet of Things module;

[0037] Uploading the to-be-analyzed feature data to a cloud server through the Internet of Things module; wherein the cloud server trains the fault early warning identification model and deploys the fault early warning identification model;

[0038] Obtaining the fault early warning identification model and the to-be-analyzed feature data from the cloud server through an edge computing module;

[0039] The edge computing module identifies a fault trend level of the to-be-analyzed feature data through the fault early warning identification model.

[0040] When the fault trend level exceeds a preset early warning threshold, a risk early warning is performed through a display.

[0041] Another aspect of the embodiment of the application further provides an escalator fault early warning system based on transfer learning, comprising:

[0042] A first module is configured to collect to-be-analyzed feature data of a target escalator, wherein the to-be-analyzed feature data comprises vibration data, temperature data and noise data.

[0043] A second module is configured to input the to-be-analyzed feature data into a fault early warning identification model for analysis and identification to determine a fault trend level.

[0044] A third module is configured to output fault early warning data for fault risk early warning when the fault trend level exceeds a preset early warning threshold.

[0045] The fault early warning identification model is obtained through transfer learning technology.

[0046] Another aspect of the embodiment of the application further provides an escalator fault early warning system based on transfer learning, comprising:

[0047] A sensor module is configured to collect to-be-analyzed feature data, wherein the to-be-analyzed feature data comprises vibration data, temperature data and noise data; and the sensor module comprises a vibration detection sensor, a temperature measurement sensor and a noise collection sensor.

[0048] An Internet of Things module is configured to upload the to-be-analyzed feature data to a cloud server.

[0049] The cloud server is configured to train a fault early warning identification model and deploy the fault early warning identification model.

[0050] An edge computing module is configured to obtain the fault early warning identification model and the to-be-analyzed feature data from the cloud server.

[0051] A fault early warning display is configured to perform a risk early warning when the fault trend level exceeds a preset early warning threshold.

[0052] Another aspect of the embodiment of the application further provides an electronic device comprising a processor and a memory.

[0053] The memory is configured to store a program.

[0054] The processor executes the program to implement the method as described above.

[0055] Another aspect of the embodiment of the present application also provides a computer readable storage medium, the storage medium stores a program, and the program is executed by a processor to implement the method as described above.

[0056] The embodiment of the present application also discloses a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the method as described above.

[0057] The embodiment of the present application collects target escalator feature data to be analyzed, the feature data to be analyzed including vibration data, temperature data and noise data; the feature data to be analyzed is input into a fault early warning identification model for analysis and identification to determine a fault trend level; when the fault trend level exceeds a preset early warning threshold, fault early warning data is output for fault risk early warning. The present application has high stability, high accuracy and low cost. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0059] Figure 1 The system structure schematic diagram provided by the embodiment of the present application is provided.

[0060] Figure 2 The processing flowchart of the escalator fault early warning system provided by the embodiment of the present application is provided.

[0061] Figure 3 The training flowchart of the escalator fault early warning system based on the migration learning provided by the embodiment of the present application is provided. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0063] In view of the problems in the prior art, the application provides an escalator fault early warning system based on transfer learning, which analyzes three working condition characteristics (vibration, noise and temperature) of the escalator respectively, updates a fault trend identification model based on an existing mechanical equipment fault identification model with high similarity and in combination with a transfer learning algorithm, so as to realize fault early warning and predictive maintenance of the escalator.

[0064] Specifically, one aspect of the application provides an escalator fault early warning method based on transfer learning, comprising:

[0065] Collecting feature data to be analyzed of a target escalator, the feature data to be analyzed comprising vibration data, temperature data and noise data;

[0066] Inputting the feature data to be analyzed into a fault early warning identification model for analysis and identification, and determining a fault trend level;

[0067] When the fault trend level exceeds a preset early warning threshold, outputting fault early warning data for fault risk early warning;

[0068] The fault early warning identification model is obtained by training through a transfer learning technique.

[0069] Optionally, the feature data to be analyzed of the target escalator comprises at least one of the following:

[0070] Collecting vibration data of the motor through a vibration detection sensor;

[0071] Collecting vibration data of the speed reducer through a vibration detection sensor;

[0072] Collecting vibration data of the anchor bolt through a vibration detection sensor;

[0073] Collecting vibration data of the main drive wheel through a vibration detection sensor;

[0074] Collecting vibration data of the tensioning sprocket through a vibration detection sensor;

[0075] Collecting temperature data of the speed reducer oil through a temperature sensor;

[0076] Collecting temperature data of the handrail belt through a temperature sensor;

[0077] Collecting temperature data of the environment through a temperature sensor;

[0078] Collecting noise data of the motor through a noise detection sensor;

[0079] Collecting noise data of the main drive wheel through a noise detection sensor;

[0080] Collecting noise data of the tensioning wheel through a noise detection sensor;

[0081] Collecting noise data of the step by the noise detection sensor.

[0082] Optionally, the training process of the escalator fault early warning identification model comprises the following steps:

[0083] Obtaining a fault trend grading model of similar equipment of the escalator equipment, wherein the similar equipment comprises but is not limited to a turbine system, a straight escalator system and a conveyor belt system;

[0084] According to the fault trend grading model of the similar equipment that has been trained, obtaining source domain training data; and obtaining target domain training data according to the escalator sample;

[0085] According to the source domain training data and the target domain training data, constructing a training set, and randomly selecting a subset from the target domain training data as a verification set;

[0086] According to the training set and the verification set, performing transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level, and constructing an escalator fault early warning identification model.

[0087] Optionally, in the step of performing transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level according to the training set and the verification set, and constructing an escalator fault early warning identification model, the transfer learning on the fault trend grading model of the similar equipment from the data level comprises:

[0088] Taking the verification set as an optimization target, performing iterative training according to the training set;

[0089] Adjusting learning weights of each training data in the training set in each iterative training process to realize transfer learning of the fault trend grading model from the source domain to the target domain.

[0090] Optionally, in the step of performing transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level according to the training set and the verification set, and constructing an escalator fault early warning identification model, the transfer learning on the fault trend grading model of the similar equipment from the feature level comprises:

[0091] Finding a common feature representation between the source domain and the target domain;

[0092] Mapping the source domain training data and the target domain training data from an original feature space to a new feature space, wherein data distributions of the source domain training data and the target domain training data in the new feature space are the same;

[0093] Migrate learning is performed on the failure trend grading model of the similar device based on the new feature space.

[0094] Optionally, the method further comprises:

[0095] Collecting, by an Internet of Things module, to-be-analyzed feature data of a target escalator;

[0096] Uploading, by the Internet of Things module, the to-be-analyzed feature data to a cloud server; wherein the cloud server trains and deploys the failure early warning identification model;

[0097] Obtaining, by an edge computing module, the failure early warning identification model and the to-be-analyzed feature data from the cloud server;

[0098] Identifying, by the edge computing module, a failure trend grade of the to-be-analyzed feature data by using the failure early warning identification model;

[0099] When the failure trend grade exceeds a preset early warning threshold, outputting failure early warning data to perform failure risk early warning by a display.

[0100] Another aspect of the embodiment of the present application further provides an escalator failure early warning system based on migration learning, comprising:

[0101] A first module is configured to collect to-be-analyzed feature data of a target escalator, wherein the to-be-analyzed feature data comprises vibration data, temperature data and noise data;

[0102] A second module is configured to input the to-be-analyzed feature data into a failure early warning identification model to perform analysis and identification, and determine a failure trend grade;

[0103] A third module is configured to output failure early warning data to perform failure risk early warning when the failure trend grade exceeds a preset early warning threshold;

[0104] The failure early warning identification model is trained by using a migration learning technology.

[0105] Another aspect of the embodiment of the present application further provides an escalator failure early warning system based on migration learning, comprising:

[0106] A sensor module is configured to collect to-be-analyzed feature data; wherein the to-be-analyzed feature data comprises vibration data, temperature data and noise data; and the sensor module comprises a vibration detection sensor, a temperature measurement sensor and a noise collection sensor;

[0107] An Internet of Things module is configured to upload the to-be-analyzed feature data to a cloud server;

[0108] a cloud server configured to train a fault early warning identification model and deploy the fault early warning identification model;

[0109] an edge computing module configured to obtain the fault early warning identification model and the feature data to be analyzed from the cloud server;

[0110] a fault early warning display configured to perform risk early warning when the fault trend level exceeds a preset early warning threshold.

[0111] Another aspect of the embodiment of the present application further provides an electronic device comprising a processor and a memory.

[0112] The memory is configured to store a program.

[0113] The processor executes the program to realize the method as described above.

[0114] Another aspect of the embodiment of the present application further provides a computer readable storage medium, which stores a program, and the program is executed by a processor to realize the method as described above.

[0115] The embodiment of the present application further discloses a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the method as described above.

[0116] The specific implementation process of the present application will be described in detail below with reference to the accompanying drawings of the specification:

[0117] The limitations and disadvantages of the prior art include the following points:

[0118] 1. The escalator fault identification can only give an alarm after the fault has occurred, which has caused irreparable accidents to the escalator passengers and is likely to cause damage to other components of the escalator, and cannot remind maintenance personnel to make predictive maintenance, thereby causing additional economic losses.

[0119] 2. The existing research only models the fault early warning based on vibration data, and does not consider the importance of noise and temperature characteristics to the identification of the fault trend of the escalator.

[0120] 3. The research on the fault trend early warning of the escalator needs to be based on the process of the occurrence of the fault point of the escalator, and at the present stage, there is a lack of such fault data, and it is difficult to realize the experiment which needs a high cost.

[0121] In view of the problems of the prior art, the working condition characteristics of the escalator sample are acquired by using a vibration detection sensor, a temperature measurement sensor and a noise acquisition sensor respectively, a high-similarity mechanical transmission equipment fault early warning recognition model that has been trained is used, and an escalator fault trend recognition model is established by combining a transfer learning algorithm, so that fault early warning is realized, and thus predictive maintenance work can be carried out before a fault occurs. The scheme system mainly comprises a main control board, sensors, an Internet of Things module, a cloud server, an edge computing module and a fault early warning display.

[0122] As shown in Figure 1 The fault early warning system of the application mainly comprises a main control board, sensors, an Internet of Things module, a cloud server, an edge computing module and a fault early warning display.

[0123] The main control board is used for controlling normal operation of the escalator, simultaneously controlling the sensors to collect working condition characteristic data of the escalator, and transmitting the data to other modules for feature analysis and fault early warning recognition model training, and when the escalator system is about to fail, the corresponding fault information also needs to be transmitted to the fault early warning display, which can be a controller composed of a microcomputer board, a PLC, a single-chip microcomputer, an embedded system and the like.

[0124] The sensors respectively comprise a vibration detection sensor, a temperature measurement sensor and a noise acquisition sensor, which are installed on key components of the escalator and are used for collecting real-time working condition characteristic data of the escalator, such as collecting vibration information of important mechanical components such as a motor, a speed reducer, a foundation bolt, a main drive wheel and a tensioning sprocket through the vibration detection sensor, collecting temperature information of key positions such as a speed reducer oil, a handrail belt and an environment through the temperature sensor, and collecting noise information of key running positions such as a motor, a main drive wheel, a tensioning wheel and a step through the noise detection sensor.

[0125] The Internet of Things module uploads the working condition characteristic data acquired by the sensors to the cloud server, early fault early warning model training needs to be carried out in the server, different fault early warning recognition models will be established for different escalators, and the trained model is reloaded to the edge computing module in the escalator fault early warning system through the Internet of Things module.

[0126] The cloud server is mainly responsible for early warning system modeling and algorithm updating, analyzes source domain data and target domain data required for transfer learning model training, and a series of experiments including feature expression, machine learning modeling, parameter optimization and recognition testing all need to be carried out in the server by using its high-performance configuration and large database; in addition, when the escalator system discovers a fault risk and recognizes an early warning signal, the server also needs to display the early warning signal on the early warning platform in a timely manner.

[0127] The edge computing module needs to implement the test process of the escalator sample, test the real-time acquired escalator sample data by using the established fault early warning model, identify the fault trend level, and return the fault early warning data when the escalator sample fault trend level is close to the set fault point. The early warning information will alert the maintenance personnel through the escalator display, and at the same time, the real-time detected early warning data is uploaded to the cloud server through the Internet of Things module, displayed on the early warning platform, and the relevant department is notified that the escalator has a fault risk at which position, thereby providing protection for the timely development of maintenance work.

[0128] The fault early warning display is used to alert the maintenance personnel, and the edge computing module of the fault early warning display will control the fault early warning display to display which position and component of the escalator will fail through the main control panel when the escalator is about to fail, thereby alerting the maintenance personnel to go to perform maintenance work.

[0129] Most of the existing researches are directed to the identification of escalator faults, but the remedial measures are taken after the fault has occurred, and the potential danger to passengers and the damage to the entire escalator mechanical system are often not enough. Therefore, the escalator fault early warning system is invented in the present application, which is more meaningful to passengers and escalators to perform preventive maintenance before the fault occurs. Figure 2 The flowchart of the escalator fault early warning system is shown, for each type of escalator, if there is no corresponding fault early warning model, then the first use of the fault early warning model needs to train the fault trend level identification model in the early stage. First, the escalator characteristic data (including vibration, temperature and noise) acquired by the sensor are uploaded to the cloud server through the Internet of Things module, and the modeling is performed in the server. The automatic escalator fault early warning system model technical route based on transfer learning is shown in Figure 3 The fault early warning model trained by the server is downloaded to the escalator control system again through the Internet of Things module, and the fault trend level identification is performed in the edge computing module. The edge computing module needs to analyze the vibration, temperature and noise characteristic data collected by the escalator sensor. If the test data is detected in the fault trend identification system and the fault trend level is not close to the set fault point, then the return value is empty, and the cycle is restarted. Once the escalator fault trend level is close to the set fault point, the early warning signal is returned, and the fault display is used to display which position and module of the automatic escalator will fail, thereby alerting the maintenance personnel to go to perform maintenance work in time.

[0130] In order to realize the fault early warning of the escalator, the application proposes to analyze the trend before the escalator fault occurs, and divide the fault trend level into four categories of no fault, mild, moderate and severe before the fault occurs. The working condition characteristics reflected by the mechanical structure of the escalator in the three levels are investigated, the key information for identifying the four trend levels is extracted from the working condition characteristics (vibration, temperature and noise), and a fault trend classification model is established by combining machine learning algorithm. After the model is trained in the early stage, the corresponding fault trend level can be identified based on the real-time working condition data collected by the escalator in the later stage, and the corresponding fault trend level is displayed on the fault display to realize the fault early warning of the escalator, so as to remind the maintenance personnel to carry out corresponding maintenance work. However, at present, most of the escalator early warning research needs to collect a large amount of escalator sample data, and needs to carry out continuous experiments for different fault trend levels, which requires the escalator to collect continuous data before the fault point occurs. However, the time point of the fault of the escalator is difficult to predict, and not only for one escalator, but also for more escalators with the same fault, which is not only time-consuming and costly for escalator manufacturers and researchers.

[0131] In addition, in the traditional machine learning field, it is usually required that the training data and test data of the collected escalator samples meet the same environmental feature space distribution, which is a prerequisite for ensuring the good performance of the escalator fault prediction model in the steady state environment. This condition is very harsh. In the practical application of escalator fault early warning, new scene data formed by escalators with different lifting heights and speeds are usually faced. The old scene data from the original modeling environment may be outdated. The difference of these environmental parameters may cause the feature space and distribution of the collected escalator working condition data to be inconsistent, so that the model trained by the original old scene data may not obtain ideal pattern classification results on new scene data, especially based on scene factors (lifting height and speed) with large differences. Therefore, this paper proposes to use transfer learning to train the escalator fault early warning model.

[0132] In order to solve the problems of large sample demand, high model training difficulty and unstable escalator fault early warning model in the changing scene in the traditional machine learning, this paper proposes an escalator fault early warning based on transfer learning, such as Figure 3The technical route of escalator fault early warning system based on transfer learning is demonstrated. The fault trend grading model of similar equipment (such as turbine, straight elevator and conveyor system, etc.) is combined with transfer learning algorithm to realize the fault early warning of escalator. The training data (source domain) in the fault trend grading model of similar equipment is transferred to the training data (target domain) of escalator from two aspects of data and features, respectively. The transfer learning algorithm is used to fuse the training data of source domain and target domain as the training set, and the machine learning algorithm is combined to train the trend grading model, so as to realize the fault early warning of escalator based on transfer learning.

[0133] When the transfer learning is carried out from the data level, the source domain and target domain training set need to be mixed to form the "source + target domain" training set. A subset is randomly selected from the target domain training set as the verification set. The fault trend grading model trained based on the source domain is used to identify the accuracy rate of the verification set as the optimization target for repeated iteration experiment. Due to the difference in data distribution between the source domain and the target domain, the machine learning model based on the source domain may cause misjudgment when identifying the trend level of the target domain. Therefore, the learning weight of the "source + target domain" training data needs to be adjusted in each iteration. The weight of the source domain and target domain training data that is beneficial to the target task is increased, and the weight of the source domain data that is not conducive to the target task is reduced. The machine learning model is continuously optimized towards the fault trend grading direction of the target domain, so as to realize the transfer learning of the fault trend grading model from the source domain to the target domain.

[0134] In addition, the transfer learning of the fault trend grading model can also be carried out from the feature level. At this time, the common feature representation between the source domain and the target domain needs to be found out. The source domain and target domain training set are mapped from the original feature space to the new feature space, in which the data distribution of the source domain and the target domain is the same. Then, the knowledge transfer is carried out based on the new feature space, so that the machine learning model can better utilize the large amount of source domain training set that is easy to obtain, and indirectly realize the identification of the target domain data, thereby realizing the transfer learning of the fault trend grading model from the source domain to the target domain.

[0135] By the scheme of transfer learning, on the one hand, the sample amount of escalator required by the early warning model is greatly reduced, the cost and time are saved, on the other hand, the experimental resources of similar equipment are reused, and the prior knowledge based on the source domain training data can make the early warning system established in the future more stable and easy to popularize. In addition, after the escalator early warning model is trained in the early stage, when facing new feature spaces formed by new type of escalator working condition data with different lifting heights and speeds, the method of transfer learning can be used for early warning model training. Although the feature data of new and old scenes may be different in distribution, they still have very similar features in the local feature range, so the model of the old scene can learn the key information beneficial to the escalator fault early warning in the new scene from a small amount of new scene data, so that only a small amount of escalator sample data is needed to realize the modeling of the fault early warning system in the new scene.

[0136] In summary, compared with the prior art, the present application has the following advantages:

[0137] 1. The fault early warning model has higher stability. In addition to the vibration characteristics of the escalator, the system increases the temperature and noise characteristics, considers the influence of these variables on the escalator fault early warning model, extracts the most critical feature information for equipment fault identification, and improves the identification accuracy of the early warning system.

[0138] 2. Cost savings. Now only a small amount of escalator samples are needed to train the fault early warning model, while the establishment of a traditional machine learning model requires a large number of escalator samples for training, which is time-consuming and wastes a lot of resources.

[0139] 3. Higher safety. The warning is given before the escalator fault occurs, so that maintenance personnel can timely perform maintenance work, which is of great significance for passenger safety and prolonging the service life of the escalator.

[0140] In some alternative embodiments, the functions / operations mentioned in the block diagram can not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two blocks shown in succession can actually be executed substantially simultaneously or the blocks can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flowcharts of the present application are provided by way of example, with the purpose of providing a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and in which sub-operations described as part of larger operations are independently executed.

[0141] Furthermore, although the present application is described in the context of functional modules, it is to be understood that one or more of the described functions and / or features can be integrated in a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It will also be appreciated that detailed discussion of the actual implementation of each module is not necessary to an understanding of the application. Rather, the actual implementation of the modules, in combination with their attributes, functions, and internal relationships, are to be understood within the context of the devices disclosed herein. Thus, those skilled in the art with access to patents, scientific journals, and other public sources known by those skilled in the art will be able to practice the application as set forth in the claims without undue experimentation, using ordinary skill in the art along with the present disclosure. It is also to be understood that the specific concepts disclosed are merely illustrative and that the scope of the present application is to be determined by the entire scope of the claims, along with all equivalents of the claims and their equivalents.

[0142] If the functions are implemented in software, the functions can be stored in or implemented as one or more computer program products, which can be incorporated into a computer- readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this document, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.

[0143] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of instructions executed in an order indicated in the flowcharts, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this document, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.

[0144] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via optical scanning of the paper or other medium, then compiled, interpreted, or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.

[0145] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following techniques can be used to implement the hardware used in the described embodiments: discrete logic circuitry having logic gates for implementing logic functions upon data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and so forth.

[0146] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples.

[0147] Although embodiments of the present application have been shown and described, it would be recognized by those of ordinary skill in the art that various changes, modifications, substitutions, and alterations can be made hereto without departing from the principles and spirit of the application, and it is intended that the scope of the application be limited only by the claims and equivalents thereof.

[0148] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the described embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the present application, and these equivalent modifications or replacements are all included in the scope defined by the claims of the present application.

Claims

1. A method for escalator fault early warning based on transfer learning, characterized in that, The method comprises the following steps: Collecting target escalator feature data to be analyzed, which includes vibration data, temperature data and noise data; Inputting the feature data to be analyzed into a fault early warning identification model for analysis and identification to determine a fault trend level; When the fault trend level exceeds a preset early warning threshold, outputting fault early warning data for fault risk early warning; The fault early warning identification model is trained by a transfer learning technology; The training process of the fault early warning identification model comprises the following steps: Obtaining a fault trend grading model of similar equipment of the escalator equipment, wherein the similar equipment includes but is not limited to a turbine system, a straight escalator system and a conveyor belt system; According to the already trained fault trend grading model of the similar equipment, obtaining source domain training data; and obtaining target domain training data according to escalator samples; According to the source domain training data and the target domain training data, constructing a training set, and randomly selecting a subset from the target domain training data as a verification set; According to the training set and the verification set, performing transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level to construct a fault early warning identification model of the escalator equipment; The training set is constructed by fusing the source domain training data and the target domain training data through a transfer learning algorithm. 2.The escalator fault early warning method based on transfer learning of claim 1, wherein, The collection of the target escalator feature data to be analyzed comprises at least one of the following: Collecting vibration data of a motor through a vibration detection sensor; Collecting vibration data of a speed reducer through a vibration detection sensor; Collecting vibration data of a foundation bolt through a vibration detection sensor; Collecting vibration data of a main drive wheel through a vibration detection sensor; Collecting vibration data of a tensioning sprocket through a vibration detection sensor; Collecting temperature data of a speed reducer oil through a temperature sensor; Collecting temperature data of a handrail belt through a temperature sensor; Collecting temperature data of an environment through a temperature sensor; Collecting noise data of a motor through a noise detection sensor; Collecting noise data of a main drive wheel through a noise detection sensor; Collecting noise data of a tensioning wheel through a noise detection sensor; Collecting noise data of a step through a noise detection sensor. 3.The escalator fault early warning method based on transfer learning of claim 1, wherein, In the step of performing transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level to construct a fault early warning identification model of the escalator equipment according to the training set and the verification set, the transfer learning on the fault trend grading model of the similar equipment from a data level comprises the following steps: Taking the verification set as an optimization target, iteratively training according to the training set; Adjusting learning weights of each training data in the training set in each iterative training process to realize transfer learning of the fault trend grading model from a source domain to a target domain.

4. The automatic escalator fault early warning method based on transfer learning according to claim 1, wherein In the step of performing transfer learning on the fault trend grading model of the similar equipment from a feature level, the transfer learning on the fault trend grading model of the similar equipment from a feature level comprises the following steps: Finding a feature representation common to the source domain and the target domain; Mapping source domain training data and target domain training data from an original feature space to a new feature space, wherein the data distribution of the source domain training data and the target domain training data in the new feature space is the same; Based on the new feature space, the failure trend grading model of the similar device is migrated.

5. The escalator failure warning method based on transfer learning according to claim 1, wherein the method further comprises: Collecting target escalator feature data to be analyzed through an Internet of Things module; Uploading the feature data to be analyzed to a cloud server through the Internet of Things module; wherein the cloud server trains the failure warning identification model and deploys the failure warning identification model; Obtaining the failure warning identification model and the feature data to be analyzed from the cloud server through an edge computing module; The edge computing module identifies the failure trend level of the feature data to be analyzed through the failure warning identification model; When the failure trend level exceeds a preset warning threshold, a risk warning is performed through a display. Comprising:

6. An escalator failure early warning system based on transfer learning, characterized by, A first module for collecting target escalator feature data to be analyzed, the feature data to be analyzed including vibration data, temperature data, and noise data; A second module for inputting the feature data to be analyzed into a failure warning identification model for analysis and identification to determine a failure trend level; A third module for outputting failure warning data to perform a failure risk warning when the failure trend level exceeds a preset warning threshold; Wherein the failure warning identification model is trained by transfer learning technology; The training process of the failure warning identification model comprises the following steps: Obtaining a failure trend grading model of similar devices of escalator equipment, wherein the similar devices include but are not limited to turbine systems, straight escalator systems, and conveyor belt systems; According to the already trained failure trend grading model of similar devices, source domain training data is obtained; and target domain training data is obtained according to escalator samples; According to the source domain training data and the target domain training data, a training set is constructed, and a subset is randomly selected from the target domain training data as a validation set; According to the training set and the validation set, transfer learning is performed on the failure trend grading model of the similar device from the data level and the feature level to construct a failure warning identification model of the escalator equipment; Wherein, by transfer learning algorithm, the training set is constructed by fusing the source domain training data and the target domain training data. Comprising:

7. An escalator failure early warning system based on transfer learning, characterized by, A sensor module for collecting feature data to be analyzed; wherein the feature data to be analyzed includes vibration data, temperature data, and noise data; the sensor module includes a vibration detection sensor, a temperature measurement sensor, and a noise acquisition sensor; An Internet of Things module for uploading the feature data to be analyzed to a cloud server; A cloud server for training a failure warning identification model and deploying the failure warning identification model; An edge computing module for obtaining the failure warning identification model and the feature data to be analyzed from the cloud server; ​ The fault early warning display is used for early warning when the fault trend level exceeds a preset early warning threshold; The training process of the fault early warning identification model comprises the following steps: Obtain a fault trend grading model of similar equipment of the escalator equipment, wherein the similar equipment comprises, but is not limited to, a turbine system, a straight ladder system and a conveyor belt system; According to the trained fault trend grading model of the similar equipment, obtain source domain training data; and obtain target domain training data according to the escalator sample; According to the source domain training data and the target domain training data, construct a training set, and randomly select a subset from the target domain training data as a verification set; According to the training set and the verification set, perform transfer learning on the fault trend grading model of the similar equipment from a data level and a feature level, and construct a fault early warning identification model of the escalator equipment. Through a transfer learning algorithm, the source domain training data and the target domain training data are fused to construct a training set.

8. An electronic device, comprising: The processor and the memory are included. The memory is used to store a program. The processor executes the program to implement the method of any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to implement the method of any one of claims 1 to 5.

Images