Gate fault prediction method and device and computing equipment

Abstract

The embodiment of the invention provides a gate fault prediction method and device and computing equipment, and the method comprises the steps: obtaining a target parameter sequence corresponding to at least two fault influence dimensions before the current time for a to-be-monitored gate; inputting the target parameter sequences corresponding to the at least two fault influence dimensions into a graph aggregation layer of the fault detection model, and fusing the target parameters of the same time range and different fault influence dimensions in each target parameter sequence through the graph aggregation layer to obtain a fused influence representation, the fault detection model is a time series data model obtained by pre-training; and inputting the fusion influence representation into a time sequence processing layer of a fault detection model, and obtaining the fault probability of the gate to be monitored after the current time. According to the method, the dependency relationship among the fault influence factors of multiple dimensions is considered during fault prediction, the data basis of gate fault prediction is more comprehensive, and the prediction accuracy of the fault probability of the to-be-monitored gate is improved.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of rail transit technology, and in particular to a gate fault prediction method, device, and computing equipment. Background Art

[0002] With the rapid development of computer technology and intelligent equipment control technology, accelerated urbanization, and growing transportation demand, rail transit technology has rapidly evolved. Its operational efficiency and safety directly impact the overall performance of the transportation system. In rail transit scenarios, multiple gates are often installed at the intersection of restricted and unrestricted areas. Once a user passes verification at a gate, the gate opens, allowing the user to enter the restricted area through the corresponding channel. As a key device for users to enter and exit restricted and unrestricted areas, the stability and reliability of the gate are particularly important.

[0003] In existing technology, when a gate machine malfunctions, an alarm is issued and maintenance personnel are dispatched to the site to repair the faulty gate machine. However, there is a certain time delay between the gate machine issuing the fault warning and the maintenance personnel arriving on site. Furthermore, gate machine malfunctions can cause harm to users and affect the normal operation of rail transit. Therefore, a timely and accurate gate machine malfunction prediction solution is urgently needed. Summary of the Invention

[0004] In view of this, embodiments of this specification provide a gate fault prediction method. One or more embodiments of this specification also relate to a gate fault prediction device, a computing device, a computer-readable storage medium, and a computer program product to address technical deficiencies in the prior art.

[0005] According to a first aspect of an embodiment of this specification, a gate fault prediction method is provided, comprising: For the gate to be monitored, obtain the target parameter sequence corresponding to at least two fault impact dimensions before the current time; Inputting the target parameter sequences corresponding to the at least two fault impact dimensions into a graph aggregation layer of a fault detection model, and fusing the target parameters of the target parameter sequences in the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, wherein the fault detection model is a pre-trained time series data model; The fused impact representation is input into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

[0006] According to a second aspect of the embodiments of this specification, a gate fault prediction device is provided, comprising: An acquisition module is configured to acquire, for the gate to be monitored, a target parameter sequence corresponding to at least two fault impact dimensions before a current time; a fusion module configured to input the target parameter sequences corresponding to the at least two fault impact dimensions into a graph aggregation layer of a fault detection model, and fuse the target parameters of the target parameter sequences within the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, wherein the fault detection model is a pre-trained time series data model; The acquisition module is configured to input the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

[0007] According to a third aspect of an embodiment of this specification, a computing device is provided, including: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the gate fault prediction method are implemented.

[0008] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned gate fault prediction method are implemented.

[0009] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, comprising a computer program / instruction, which implements the steps of the above-mentioned gate fault prediction method when executed by a processor.

[0010] The embodiments of this specification provide a gate machine fault prediction method, which realizes obtaining a target parameter sequence corresponding to at least two fault impact dimensions of the gate machine to be monitored before the current time, and through the graph aggregation layer of the fault detection model, the target parameters of the same time range and different fault impact dimensions in the target parameter sequence corresponding to each fault impact dimension are fused, and then based on the timing processing layer of the fault detection model, the failure probability of the gate machine to be monitored after the current time is obtained. In this way, the parameters under multiple fault influencing dimensions that affect gate machine failure are comprehensively considered to predict the probability of gate machine failure after the current time, so that whether the gate machine may fail can be predicted in advance and maintenance can be carried out in time to avoid gate machine failure causing harm to users and affecting the normal operation of rail transit; in addition, the graph aggregation layer of the fault detection model is used to spatially fuse the target parameters of the same time range and different fault influencing dimensions, and the mutual dependence between multiple fault influencing dimensions under the same time range is comprehensively analyzed. Then, the spatial fusion results are subjected to time series analysis to determine the failure probability of the gate machine to be monitored after the current time. The spatial and temporal dependence between the fault influencing factors of multiple dimensions is taken into account in fault prediction. The data basis for gate machine failure prediction is more comprehensive, which improves the prediction accuracy of the failure probability of the gate machine to be monitored. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flow chart of a gate fault prediction method provided by one embodiment of this specification; Figure 2 This is a process flow chart of a gate fault prediction method provided by an embodiment of this specification; Figure 3 This is a schematic diagram of the structure of a gate fault prediction device provided by an embodiment of this specification; Figure 4 This is a structural block diagram of a computing device provided by one embodiment of this specification. DETAILED DESCRIPTION

[0012] The following description sets forth many specific details to facilitate a thorough understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the scope of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0013] The terms used in one or more embodiments of this specification are for the purpose of describing specific embodiments only and are not intended to limit one or more embodiments of this specification. The singular forms "a," "an," and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0014] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0015] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0016] It should be noted that with the acceleration of urbanization and the growth of transportation demand, the rail transit industry has ushered in opportunities for rapid development. As a vital component of public transportation, the operational efficiency and safety of the subway directly impact the overall performance of the urban transportation system. As key equipment for passenger entry and exit, the stability and reliability of subway gates are particularly important. Current maintenance methods are often based on passive response after a failure occurs, lacking preventative measures. If a gate malfunction occurs, there is a certain time delay between the discovery of the problem and the arrival of maintenance personnel. Failures may also cause harm to passengers. For example, damage to the gate control panel may cause passengers to be pinched. Failures may even cause subway operations to be interrupted, affecting passenger travel and even causing traffic congestion.

[0017] In practice, the development of artificial intelligence (AI), particularly models for processing time-series data, has made predictive maintenance of gate machines possible. This technology, through real-time monitoring and data analysis, can predict potential gate failures and predict whether a gate is likely to fail. The development of gate failure prediction technology is crucial for improving the safety and efficiency of the rail transit industry. With continued technological advancement, more intelligent, accurate, and secure gate failure prediction and management can be achieved.

[0018] In one implementation, gate operating status parameters such as voltage, phase, and pressure can be collected and analyzed. Based on the analysis results, abnormalities in the gate's operating status can be predicted. This allows for early prediction of the normal operating status of the gate's core components, allowing maintenance personnel to be notified promptly for repairs. However, these methods often simply collect parameters such as voltage, phase, and pressure, comparing a single set of data at a specific moment with a threshold range. This fails to consider the interactions between multiple fault-influencing parameters, resulting in low prediction accuracy.

[0019] In another implementation, a gate control system can be constructed based on a computer. The data acquisition module acquires real-time data from the monitoring site. The index analysis module analyzes the fault impact index and monitors the real-time status index in real time. The warning characteristic value analysis module analyzes the warning characteristic value. The signal transmission module transmits the warning signal based on the warning characteristic value to the back-end platform management subsystem. The warning display module displays the warning signal to the front-end subsystem via the network transmission subsystem. However, the above method is suitable for parking gate fault prediction, not for subway gate fault prediction based on image recognition. It also does not consider the correlation between multiple intentional influencing factors, and the prediction accuracy needs to be further improved.

[0020] The embodiments of this specification provide a gate machine fault prediction solution, which predicts the probability of gate machine failure through historical time series data of gate machine fault influencing parameters; in addition, the influence of the gate machine's environment and the gate machine's own operating parameters are considered in the gate machine fault influencing parameters, and the hidden dependencies between parameters under multiple gate machine fault influencing dimensions are comprehensively analyzed during the prediction process to improve the accuracy of gate machine fault prediction.

[0021] In this specification, a gate fault prediction method is provided. This specification also involves a gate fault prediction device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0022] See also Figure 1 , Figure 1 A flow chart of a gate fault prediction method provided according to an embodiment of this specification is shown, which specifically includes the following steps.

[0023] Step 102: For the gate to be monitored, obtain a target parameter sequence corresponding to at least two fault impact dimensions before the current time.

[0024] Among them, the gate to be monitored refers to any gate that needs to be monitored in the rail transit system. The gate refers to an important device used to control the user's entry and exit into restricted areas and non-restricted areas. It can verify the user by reading the ticket, QR code or other form of electronic certificate held by the user, and automatically open or close the channel according to the verification result, thereby realizing automated management.

[0025] The fault impact dimension refers to the dimension that may affect the failure of the gate machine, that is, the parameters under the fault impact dimension can reflect the current operating status of the gate machine, thereby indicating whether the gate machine may fail in the future. For example, the fault impact dimension can include the gate machine surrounding environment dimension and the gate machine's own operating parameter dimension. The gate machine surrounding environment dimension can also include the ambient temperature and / or ambient humidity dimension. The gate machine's own operating parameter dimension includes the operating voltage and / or operating current dimension, the fan door switch dimension, the CPU temperature and / or the motherboard temperature dimension.

[0026] One fault impact dimension may correspond to one target parameter sequence, and the target parameter sequence includes target parameters corresponding to each time point within a set time period before the current time for the corresponding fault impact dimension.

[0027] In actual implementation, the parameters corresponding to at least two fault impact dimensions of the gate to be monitored within a set time period before the current time can be obtained, and the target parameter sequence corresponding to each fault impact dimension can be obtained respectively. The target parameter sequence is a time series composed of historical parameters of the set time period before the current time. Subsequently, the target parameter sequence corresponding to each fault impact dimension is analyzed, and the probability of the gate to be monitored failing after the current time can be predicted. That is, based on the historical data of each fault impact dimension within a period of time before the current time of the gate to be monitored, the probability of future failures can be predicted, thereby realizing early warning of failures of the gate to be monitored.

[0028] In an optional implementation of this embodiment, for the gate to be monitored, obtaining a target parameter sequence corresponding to at least two fault impact dimensions before the current time includes: Using at least two sensors configured on the gate to be monitored, target impact parameters of at least two fault impact dimensions at each time point are collected and stored; Obtain target impact parameters of at least two fault impact dimensions at each time point before the current time and within a set time period, align them according to time, and obtain target parameter sequences corresponding to at least two fault impact dimensions before the current time.

[0029] It should be noted that various sensors configured on the gate to be monitored can be used to collect and store corresponding target impact parameters in real time. The parameters of a fault impact dimension can be obtained by collecting at least one sensor. For example, if the fault impact dimension is the dimension of the surrounding environment of the gate, the parameters of the fault impact dimension can be collected by a temperature sensor and / or a humidity sensor.

[0030] When collecting and storing target impact parameters corresponding to the fault impact dimension, each parameter can be structured data with a unified timestamp format. The collection frequencies of different sensors can be the same or different.

[0031] In actual implementation, the parameters of each sensor can be collected in real time, and fault detection can be performed periodically. Specifically, every first time period, the target impact parameters of at least two fault impact dimensions at each time point before the current time and within the set time period can be obtained, and then the target impact parameters of each fault impact dimension can be aligned according to the time point to obtain the target parameter sequence corresponding to at least two fault impact dimensions before the current time.

[0032] In addition, after obtaining the target impact parameters of at least two fault impact dimensions at each time point before the current time and within a set time period, preprocessing operations such as data cleaning, outlier detection and removal can be performed on the target impact parameters of each fault impact dimension at each time point before time series alignment is performed. Specifically, data cleaning can include missing value processing and / or duplicate record processing, such as checking and processing missing data points, using interpolation methods (such as linear interpolation) or neighbor-based filling methods to fill missing values, and removing identical duplicate records to avoid bias in analysis; outlier detection and removal can be performed through statistical methods, domain knowledge, machine learning models, etc.

[0033] In specific implementation, since the time intervals for collecting target impact parameters corresponding to different fault impact dimensions may be consistent or inconsistent, and noise data at some abnormal time points may also be collected, resulting in inconsistent time points for target impact parameters corresponding to different fault impact dimensions, subsequent fusion according to time points cannot be performed. Therefore, the parameters of each fault impact dimension at each time point can be aligned according to the timestamp, and the parameters of redundant time points can be deleted to obtain the target parameter sequence corresponding to at least two fault impact dimensions before the current time, wherein the redundant time point refers to the time point at which at least one fault impact dimension does not exist, ensuring that the time points included in the target parameter sequence obtained under each fault impact dimension are the same.

[0034] For example, the target impact parameter 1 of fault impact dimension 1 collected by sensor 1 at each time point before the current time is: 00:00-X1, 00:03-X2, 00:06-X3, 00:09-X4, 00:12-X5. The target impact parameter 2 of fault impact dimension 2 collected by sensor 2 at each time point before the current time is: 00:00-Y1, 00:02-Y2, 00:04-Y3, 00:06-Y4, 00:08-Y5, 00:10-Y6, 00:12-Y7. The target impact parameter 3 of the fault impact dimension 3 collected by sensor 3 at each time point before the current time is: 00:00-Z0, 00:01-Z1, 00:02-Z2, 00:03-Y3, 00:04-Z4, 00:05-Z5, 00:06-Z6, 00:07-Z7, 00:08-Z8, 00:09-Z9, 00:10-Z10, 00:11-Z11, 00:12-Z12. Perform time alignment on the target impact parameters 1-3 and delete the target impact parameters at redundant time points to obtain the target parameter sequence 1 corresponding to fault impact dimension 1: 00:00-X1, 00:06-X3, 00:12-X5; the target parameter sequence 2 corresponding to fault impact dimension 2: 00:00-Y1, 00:06-Y4, 00:12-Y7; and the target parameter sequence 2 corresponding to fault impact dimension 3: 00:00-Z0, 00:06-Z6, 00:12-Z12.

[0035] In the embodiments of this specification, a variety of sensors configured on the gate to be monitored can be used to obtain target impact parameters of at least two fault impact dimensions at various time points before the current time and within a set time period, and time alignment is performed to obtain target parameter sequences corresponding to each fault impact dimension, so that the time points included in each target parameter sequence are corresponding, thereby enabling the parameters of different fault impact dimensions in the same time range to be fused subsequently to improve the accuracy of fault prediction.

[0036] In an optional implementation of this embodiment, the at least two fault impact dimensions include a gate surrounding environment dimension and a gate operating parameter dimension. The gate surrounding environment dimension includes an ambient temperature and / or ambient humidity dimension. The gate operating parameter dimension includes an operating voltage and / or operating current dimension, a door switch dimension, a CPU temperature, and / or a mainboard temperature dimension. Utilize at least two sensors configured on the gate to be monitored to collect and store target impact parameters of at least two fault impact dimensions at each time point, including at least one of the following: Using the temperature sensor and / or humidity sensor configured on the gate to be monitored, collect the temperature value and / or humidity value at each time point under the ambient temperature and / or ambient humidity dimensions, and use the temperature value and / or humidity value as the target influencing parameter; Using the voltage sensor and / or current sensor configured on the gate to be monitored, collect the voltage value and / or current value at each time point under the working voltage and / or working current dimensions, and use the voltage value and / or current value as the target influencing parameter; Using the door switch sensor configured on the gate to be monitored, the cumulative number of door openings and closings at each time point in the door opening and closing dimension is collected, and the cumulative number of door openings and closings is used as the target influencing parameter; Utilize the CPU temperature sensor and / or motherboard sensor configured on the gate to be monitored to collect the CPU temperature values and / or motherboard temperature values at each time point under the CPU temperature and / or motherboard temperature dimensions, and use the CPU temperature values and / or motherboard temperature values as target influencing parameters.

[0037] It should be noted that the gate to be monitored can be configured with at least one of a temperature sensor, a humidity sensor, a voltage sensor, a current sensor, a door switch sensor, a CPU temperature sensor, and a motherboard temperature sensor; at least two fault impact dimensions can include at least one of the ambient temperature and / or ambient humidity dimension, the working voltage and / or working current dimension, the door switch dimension, the CPU temperature and / or the motherboard temperature dimension; the target impact parameter can include at least one of the temperature value of the ambient temperature, the humidity value of the ambient humidity, the voltage value of the working voltage, the current value of the working current, the cumulative number of door openings and closings, the CPU temperature value, and the motherboard temperature value.

[0038] In the embodiments of this specification, the required sensors can be configured on the gate to be monitored based on the actual scenario and monitoring requirements. The corresponding numerical values can be collected based on the configured sensors as response parameters under the corresponding fault impact dimension, so as to facilitate subsequent comprehensive analysis of the target parameter sequence composed of target impact parameters of at least two fault impact dimensions at each time point to achieve fault probability prediction. The fault impact dimension can be defined and configured based on actual needs, and the monitoring flexibility is higher, which can be adapted to a variety of different actual application scenarios.

[0039] Step 104: Input the target parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the fault detection model, and fuse the target parameters of the same time range and different fault impact dimensions in each target parameter sequence through the graph aggregation layer to obtain a fusion impact representation, wherein the fault detection model is a pre-trained time series data model.

[0040] Specifically, the fault detection model can be a time series data model pre-trained based on sample parameter sequences of at least two fault impact dimensions. This fault detection model can refer to a multivariate time series prediction model and can include a graph aggregation layer and a time series processing layer. The graph aggregation layer is used to perform graph aggregation on multiple input target parameter sequences, analyze their dependencies, and fuse target parameters within the same time range and different fault impact dimensions within the input target parameter sequences to obtain a fused impact representation; this fused impact representation can be input into the time series processing layer.

[0041] As an example, the fault detection model can be SageFormer, a time series prediction model that combines a graph neural network (GNN) and the Transformer architecture. It is designed to process data with complex spatial dependencies and temporal dynamics. SageFormer's key feature is its use of graph structures to model spatial relationships between data points and its use of Transformers to capture long-range temporal dependencies. The graph aggregation layer can be the dynamic graph learning module in SageFormer, and the time series processing layer is the spatiotemporal network in SageFormer.

[0042] It should be noted that the graph aggregation layer of the fault detection model is used to spatially fuse the target parameters of different fault impact dimensions in the same time range, and comprehensively analyze the mutual dependence between multiple fault impact dimensions in the same time range. The dependence between the influencing factors of multiple dimensions is taken into account during fault prediction. The data basis for gate fault prediction is more comprehensive, which improves the prediction accuracy of the failure probability of the gate to be monitored.

[0043] In an optional implementation of this embodiment, target parameter sequences corresponding to at least two fault impact dimensions are input into a graph aggregation layer of a fault detection model. The graph aggregation layer fuses target parameters of the target parameter sequences within the same time range and different fault impact dimensions to obtain a fused impact representation, including: Input the target parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the fault detection model. In the graph aggregation layer, a global label is added before the target parameter sequence corresponding to each fault impact dimension to obtain an updated parameter sequence for each fault impact dimension. The global label is used to extract the global information of the corresponding target parameter sequence. Encode the update parameter sequence of each fault impact dimension into a graph node to obtain the corresponding graph structure; Aggregate each graph node in the graph structure and fuse the target parameters of the same time range and different fault impact dimensions to obtain the fused impact representation.

[0044] It should be noted that the global label is a global token. The target parameter sequences are marked with the global label and encoded as graph nodes. The time series features under different fault impact dimensions are then integrated through graph aggregation. This is a method that combines the attention mechanism in natural language processing and graph neural networks to capture the complex relationships between multiple target parameter sequences and the interactions between different target parameter sequences.

[0045] In actual implementation, after obtaining the target parameter sequences corresponding to at least two fault impact dimensions, the target parameter sequences corresponding to at least two fault impact dimensions can be input into the graph aggregation layer of the fault detection model. In the graph aggregation layer, a global label is added in front of the target parameter sequence corresponding to each fault impact dimension. Based on the global label, the global information of the corresponding target parameter sequence can be extracted to realize graph aggregation of each target parameter sequence.

[0046] Specifically, to ensure the stability of the input data for the graph aggregation layer, the target parameter sequences corresponding to each fault impact dimension are typically standardized or normalized. If the target parameter sequences are long, they can be split into shorter segments as needed. A special "global label" is then added to the beginning of each target parameter sequence. This global label does not directly correspond to any specific data point in the target parameter sequence, but rather serves as an abstract representation of the entire sequence. The global label can be initialized in some way, such as randomly or based on sequence statistics.

[0047] Each target parameter sequence and its global label can then be encoded as a graph node. For each target parameter sequence, in addition to its own data, it also includes the corresponding global label, i.e., the updated parameter sequence. When encoding the updated parameter sequence as a graph node, the feature vectors of each node can be constructed using a variety of methods, such as simple feature extraction or the encoder output of the Transformer architecture.

[0048] Afterwards, the relationships (edges) between graph nodes can be defined. These can be manually set based on domain knowledge or dynamically generated by calculating the similarity between nodes. Edges can be bidirectional or directional. Graph convolutional networks (GCNs), graph attention networks (GATs), or other types of graph neural networks can then be used to update the features of graph nodes. These networks can learn the information transfer patterns between adjacent graph nodes and integrate the information from each graph node through graph aggregation operations (such as summation, averaging, or more complex pooling functions). These networks then fuse target parameters within the same time range and across different fault impact dimensions to obtain the final fused impact representation.

[0049] In the embodiments of this specification, global tags can participate in the process of each graph node in the aggregation graph structure, thereby capturing the overall characteristics of all relevant target parameter sequences. By acting as a bridge between different target parameter sequences, global tags can be used to achieve the fusion of target parameters of the same time range and different fault impact dimensions in each target parameter sequence, fully explore the potential spatial dependency relationship between multiple target parameter sequences, and ensure the accuracy of the prediction of the subsequent gate failure probability to be monitored.

[0050] Step 106: Input the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

[0051] Among them, the timing processing layer is used to analyze the relationship between the fusion impact representation of the input in the timing to obtain the final failure probability.

[0052] In actual implementation, the data input to the fault detection model is the target parameter sequence corresponding to each fault impact dimension within a set time period before the current time (e.g., 168 hours, or 7 days). After the fault detection model analyzes the spatial and temporal dependencies, it outputs the probability of a fault occurring in the gate to be monitored within a second time period after the current time (e.g., 96 hours, or 4 days). This set time period and second time period can be configured during model training of the fault detection model, meaning that they remain consistent during the training and application phases. For example, the fault detection model is trained based on the sample parameter sequence corresponding to each fault impact dimension within 7 days before the target time point, as well as the label probability of a gate fault occurring within 4 days after the target time point. After training is complete, the fault detection model can output the probability of a fault occurring in the gate to be monitored within 4 days after the current time point, based on the target parameter sequence corresponding to each fault impact dimension within 7 days before the current time point.

[0053] Specifically, the set duration corresponding to the input target parameter sequence and the predicted time period of the fault probability (i.e., the second duration) can be configured based on actual business needs. In other words, the time range of the target parameter sequence input to the fault detection model and the predicted future time range can be flexibly adjusted.

[0054] It should be noted that by using the graph aggregation layer of the fault detection model to spatially fuse the target parameters of the same time range and different fault impact dimensions, the spatial dependencies between multiple fault impact dimensions in the same time range can be comprehensively analyzed. Then, a time series analysis is performed on the spatial fusion results to determine the time series dependencies between multiple fault impact dimensions, so as to determine the failure probability of the gate to be monitored after the current time. At the same time, the dependency between the spatial and time dimensions is considered, making the data basis for subsequent gate fault prediction more comprehensive and improving the prediction accuracy of the failure probability of the gate to be monitored.

[0055] In an optional implementation of this embodiment, the time series processing layer includes an attention mechanism and a classification layer; the fused impact representation is input into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time, including: Input the fusion influence representation into the attention mechanism of the temporal processing layer to obtain the corresponding temporal fusion encoding; The time series fusion code is input into the classification layer of the time series processing layer to obtain the failure probability of the gate to be monitored after the current time.

[0056] Among them, the Attention Mechanism is a technology widely used in deep learning, especially when processing sequence data. It allows the model to dynamically focus on different parts of the input sequence when processing the data at each time step, thereby improving the model's ability to capture important information; the classification layer is a layer in the neural network used to output the final classification result. In multi-class classification tasks, the classification layer is usually located at the end of the model, responsible for mapping the features extracted by the previous layers to various categories and outputting the probability distribution of each category.

[0057] In practical implementation, the self-attention mechanism in the Transformer can be used to capture the long-range temporal dependencies of the fused influence representation. Specifically, the input fused influence representation can be converted into a query vector, a key vector, and a value vector. These vectors are typically obtained through linear transformations. The dot-product attention mechanism is used to calculate the similarity score between the query vector and the key vector. The value vector is then weighted and summed based on the attention score to obtain the final attention output. To enhance the expressiveness of the model, a multi-head attention mechanism is often used, executing multiple attention functions in parallel and then concatenating the results.

[0058] In addition, to preserve the sequential information in the fused influence representation, position encoding can be added to each time step in the fused influence representation. Position encoding can be generated using sine and cosine functions or learned. The attention output processed by the attention mechanism can then be input into the encoder. The encoder typically consists of multiple layers of self-attention layers and feedforward neural networks, with residual connections and layer normalization after each layer. The output of the encoder is a context vector containing rich spatiotemporal information. This context vector is the final temporal fusion code, which integrates the information at each time in the fused influence representation.

[0059] The classification layer can include a fully connected layer, an activation function, and an output layer. The obtained time series fusion code is input into one or more fully connected layers (also called dense layers) to further refine the features and map them to the output space. An appropriate activation function is applied after the fully connected layer. For binary classification problems, the Sigmoid activation function is usually used, and for multi-classification problems, the Softmax activation function is used. The output layer produces the final prediction result. For fault probability prediction, the output layer is usually a single-node fully connected layer, whose output range is between 0 and 1, indicating the probability of fault occurrence.

[0060] In the embodiments of this specification, the attention mechanism can be used to fuse the temporal information in the input fusion impact representation, and then the failure probability of the gate to be monitored after the current time is obtained based on the classification layer. The attention mechanism is introduced to analyze the temporal dependency between multiple fault impact dimensions, so that the data basis for subsequent gate failure prediction is more comprehensive, and the prediction accuracy of the failure probability of the gate to be monitored is improved.

[0061] In an optional implementation of this embodiment, the fault detection model is trained and obtained by the following method: Obtaining sample data and fault probability labels corresponding to the sample gate, wherein the sample data includes sample parameter sequences corresponding to at least two fault impact dimensions before the sample time; Input the sample parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the initial detection model. The graph aggregation layer fuses the sample parameters of the sample parameter sequences in the same time range and different fault impact dimensions to obtain a sample fusion representation. Input the sample fusion representation into the time series processing layer of the initial detection model to obtain the predicted failure probability of the sample gate after the sample time; According to the fault probability label and the predicted fault probability, the initial detection model is trained until the training stop condition is reached, and a trained fault detection model is obtained.

[0062] It should be noted that the sample data corresponding to the sample gate includes a sample parameter sequence corresponding to at least two fault impact dimensions before the sample time. The collection process of the sample parameter sequence can refer to the collection process of the target parameter sequence in the above application stage, and this embodiment of the specification will not be repeated here.

[0063] In addition, in addition to collecting sample data corresponding to the sample gate, information on whether the sample gate has a fault within the second time period after the sample time can also be collected. For example, if a fault occurs, it is marked as 1, and if no fault occurs, it is marked as 0. This information can be collected once every third time period (such as 10 minutes). Specifically, the probability of the sample gate having a fault within the second time period after the sample time can be manually marked as a label, or the probability of its fault can be obtained by analyzing and calculating the fault information detected within the second time period after the sample time as a label. Alternatively, the fault information of multiple sample gates corresponding to the same sample data can be obtained within the second time period after the sample time, thereby calculating the fault probability corresponding to the sample data as a label.

[0064] It should be noted that "the sample parameter sequences corresponding to at least two fault impact dimensions are input into the graph aggregation layer of the initial detection model, and the sample parameters of the same time range and different fault impact dimensions in each sample parameter sequence are fused through the graph aggregation layer to obtain a sample fusion representation; the sample fusion representation is input into the timing processing layer of the initial detection model to obtain the predicted failure probability of the sample gate after the current time". The specific implementation process can be referred to the above-mentioned "the target parameter sequences corresponding to at least two fault impact dimensions are input into the graph aggregation layer of the fault detection model, and the target parameters of the same time range and different fault impact dimensions in each target parameter sequence are fused through the graph aggregation layer to obtain a fusion impact representation; the fusion impact representation is input into the timing processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time". The embodiments of this specification will not be repeated here.

[0065] In actual implementation, the failure probability label is the true value of the sample data, representing the probability of a sample gate failure occurring within the second period after the sample time. The predicted failure probability is the predicted value output by the time series processing layer of the initial detection model, representing the predicted probability of a sample gate failure occurring within the second period after the sample time. Based on the failure probability label and the predicted failure probability, the model loss is calculated. Backward gradient propagation based on the loss value adjusts the model parameters of the initial detection model until the training stop condition is met, resulting in a fully trained fault detection model.

[0066] In one possible implementation, whether the training stop condition is met can be determined only based on the relationship between the loss value and the loss threshold. If the loss value is greater than or equal to the loss value threshold, it means that the training stop condition has not been met. If the loss value is less than the loss value threshold, it means that the training stop condition is met, and the training is stopped to obtain the trained fault detection model, where the loss value threshold is the critical value of the loss value.

[0067] In another possible implementation, in addition to comparing the relationship between the loss value and the loss value threshold, the number of iterations can also be combined to determine whether the training stop condition has been met. If the loss value is greater than the loss value threshold, it can be further determined whether the current number of iterations has reached the preset number of iterations. If the current number of iterations has not reached the preset number of iterations, it can be determined that the training stop condition has not been met. If the preset number of iterations has been reached, it is determined that the training stop condition has been met, and a trained fault detection model is obtained. The preset number of iterations is set based on actual conditions. When the number of training iterations reaches the preset number of iterations, it indicates that the initial detection model has been trained sufficiently. At this point, the prediction results of the initial detection model are extremely close to the actual results, and training can be stopped.

[0068] In the embodiments of this specification, the difference between the predicted results and the actual results of the initial detection model can be intuitively shown by calculating the loss value. The specific training situation of the initial detection model can be judged according to the loss value, and the initial detection model can be trained in a targeted manner based on the difference, and the parameters of the initial detection model can be adjusted, thereby effectively improving the training rate and training effect of the initial detection model.

[0069] In practical implementation, multiple sample data points and corresponding failure probability labels for multiple gates can be obtained as a sample dataset. This dataset can then be divided into training, validation, and test datasets in a predetermined ratio (e.g., 6:2:2). The initial detection model architecture is then configured, specifying the duration of the input sample parameter sequence and the predicted time period for the predicted failure probability. For example, if the input is a sample parameter sequence from the 7 days before time t, the output is the predicted failure probability of the gate occurring within the next 4 days after time t. The initial detection model is trained by adjusting hyperparameters such as the number of training rounds, optimizer, step size, batch_size, and graph_depth to obtain a trained candidate detection model. The candidate detection model's training performance is then tested on the validation dataset. Based on the MSE (mean squared error) and MAE (mean absolute error) metrics, the training weights of the model with the best training performance are selected to obtain the trained fault detection model. Afterwards, the test dataset is used to finally evaluate the generalization ability of the fault detection model obtained through training. The testing process is independent of the training and validation processes to ensure the authenticity and reliability of the evaluation results.

[0070] In an optional implementation of this embodiment, after inputting the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time, the following steps are further included: According to the set maintenance judgment strategy and failure probability, the maintenance prompts of the gate to be monitored are determined.

[0071] The maintenance judgment strategy is a pre-configured constraint that determines whether the fault probability output by the fault detection model requires maintenance. For example, the maintenance judgment strategy includes issuing a maintenance alarm when the gate machine failure probability exceeds a probability threshold. The maintenance prompt is used to inform staff whether the gate machine under monitoring requires maintenance. In other words, the maintenance prompt can include a corresponding maintenance alarm message when maintenance is required, and can also display a prompt message when maintenance is not required.

[0072] It should be noted that, based on the set maintenance judgment strategy, the fault probability output by the fault detection model can be analyzed and judged to determine whether the gate to be monitored needs maintenance, and a maintenance prompt can be given according to the judgment result. Thus, when the failure probability of the gate to be monitored indicates that maintenance is required, prompts can be given in time, so that maintenance can be carried out in time before the gate to be monitored actually fails, thereby avoiding harm to users caused by gate failure and avoiding affecting the normal operation of rail transit.

[0073] In an optional implementation of this embodiment, the maintenance judgment strategy includes issuing a maintenance alarm when the gate failure probability is greater than a probability threshold; According to the set maintenance judgment strategy and failure probability, the maintenance prompts of the gate to be monitored are determined, including: determining whether the probability of failure is greater than a probability threshold; If it is greater than the probability threshold, the attribute information of the gate to be monitored is obtained, and a maintenance alarm message of the gate to be monitored is generated based on the attribute information, and the maintenance alarm message is pushed to the set recipient.

[0074] The probability threshold is a pre-configured value used to determine whether the probability of failure is too high. This probability threshold can be flexibly adjusted based on actual business needs, for example, the probability threshold can be configured to 0.8, 0.9, etc. The attribute information of the gate to be monitored refers to information used to assist personnel in determining the location, maintenance personnel, maintenance tools, and other maintenance methods of the gate to be monitored, such as the identification, location, and model of the gate to be monitored. The set recipient refers to the terminal device or electronic account corresponding to the personnel who monitor, maintain, or repair the gate to be monitored.

[0075] In actual implementation, it is possible to determine whether the failure probability is greater than the probability threshold. If it is greater than the probability threshold, it means that the probability of the gate to be monitored failing after the current time is high. At this time, the attribute information of the gate to be monitored can be obtained, and a maintenance alarm message of the gate to be monitored can be generated based on the attribute information. The maintenance alarm message is pushed to the set receiver. The set receiver can determine the location, maintenance personnel, maintenance tools and other maintenance methods of the gate to be monitored based on the attribute information in the maintenance alarm message, so as to realize timely maintenance of the gate to be monitored. If it is not greater than the probability threshold, the normal working message of the gate to be monitored can be pushed to the set receiver so that the set receiver can be informed of the working status of the gate to be monitored in a timely manner, or no prompt processing can be performed.

[0076] It should be noted that if the failure probability is greater than the probability threshold, a maintenance alarm message of the gate to be monitored can be generated based on the attribute information of the gate to be monitored and the attribute information, and the maintenance alarm message can be pushed to the set recipient, so that the set recipient can promptly determine the corresponding maintenance method and take corresponding measures in time before the gate to be monitored actually fails.

[0077] The embodiments of this specification provide a gate machine fault prediction method, which comprehensively considers the parameters under multiple fault influencing dimensions that affect the gate machine fault, and predicts the probability of the gate machine fault after the current time, so that it can predict in advance whether the gate machine may fail, and perform maintenance in time to avoid the gate machine failure causing harm to users and avoiding affecting the normal operation of rail transit; and, utilizes the graph aggregation layer of the fault detection model to spatially fuse the target parameters of the same time range and different fault influencing dimensions, comprehensively analyzes the mutual dependence between multiple fault influencing dimensions under the same time range, and then performs time series analysis on the spatial fusion results to determine the failure probability of the gate machine to be monitored after the current time. When predicting the fault, the spatial and temporal dependence between the fault influencing factors of multiple dimensions is considered, the data basis for the gate machine fault prediction is more comprehensive, and the prediction accuracy of the failure probability of the gate machine to be monitored is improved.

[0078] The following combined Figure 2 , taking the application of the gate machine fault prediction method provided in this specification in the subway scene as an example, the gate machine fault prediction method is further explained. Figure 2 A flow chart of the processing process of a gate fault prediction method provided by an embodiment of this specification is shown, which specifically includes the following steps.

[0079] Step 202: Utilize multiple sensors to collect sample parameter sequences corresponding to at least two fault impact dimensions of a sample subway gate machine before a sample time, and obtain a fault probability label of the sample subway gate machine that fails after the sample time.

[0080] Step 204: Divide the collected sample parameter sequence into a training data set, a validation data set, and a test data set according to a set ratio.

[0081] Step 206: Train the initial detection model based on the training data set, select the fault detection model obtained through training based on the validation data set, and test the generalization ability of the fault detection model through the test data set. After passing the test, a trained fault detection model is obtained.

[0082] It should be noted that the above steps 202-206 are the training process of the fault detection model, and the following steps 208-212 are the process of using the trained fault detection model to perform fault probability prediction.

[0083] Step 208: Utilize multiple sensors on the subway gate to be monitored to obtain target parameter sequences corresponding to at least two fault impact dimensions before the current time.

[0084] Step 210: Input the target parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the fault detection model, and fuse the target parameters of the same time range and different fault impact dimensions in each target parameter sequence through the graph aggregation layer to obtain a fused impact representation.

[0085] Step 212: Input the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the subway gate to be monitored after the current time.

[0086] Step 214: When the fault probability is greater than the probability threshold, the attribute information of the subway gate to be monitored is obtained, and a maintenance alarm message of the subway gate to be monitored is generated according to the attribute information, and the maintenance alarm message is pushed to the subway staff.

[0087] The embodiments of this specification provide a gate machine fault prediction method, which comprehensively considers the parameters under multiple fault influencing dimensions that affect the failure of subway gate machines, and predicts the probability of subway gate machine failure after the current time, so that it can predict in advance whether the subway gate machine may fail, and perform maintenance in time to avoid the failure of the subway gate machine causing harm to passengers and avoiding affecting the normal operation of the subway; and, the graph aggregation layer of the fault detection model is used to spatially fuse the target parameters of the same time range and different fault influencing dimensions, comprehensively analyze the mutual dependence between multiple fault influencing dimensions under the same time range, and then perform time series analysis on the spatial fusion results to determine the failure probability of the subway gate machine to be monitored after the current time. The spatial and temporal dependencies between the fault influencing factors of multiple dimensions are considered in the fault prediction, and the data basis for gate machine fault prediction is more comprehensive, which improves the prediction accuracy of the failure probability of the subway gate machine to be detected.

[0088] Corresponding to the above method embodiment, this specification also provides an embodiment of a gate fault prediction device, Figure 3 FIG1 shows a schematic diagram of the structure of a gate fault prediction device provided by an embodiment of this specification. Figure 3 As shown, the device includes: The acquisition module 302 is configured to acquire, for the gate to be monitored, a target parameter sequence corresponding to at least two fault impact dimensions before a current time; The fusion module 304 is configured to input the target parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the fault detection model, and fuse the target parameters of the target parameter sequences within the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, wherein the fault detection model is a pre-trained time series data model; The acquisition module 306 is configured to input the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

[0089] Optionally, the acquisition module 302 is further configured to: Using at least two sensors configured on the gate to be monitored, target impact parameters of at least two fault impact dimensions at each time point are collected and stored; Obtain target impact parameters of at least two fault impact dimensions at each time point before the current time and within a set time period, align them according to time, and obtain target parameter sequences corresponding to at least two fault impact dimensions before the current time.

[0090] Optionally, the at least two fault impact dimensions include a gate machine ambient environment dimension and a gate machine operating parameter dimension, wherein the gate machine ambient environment dimension includes an ambient temperature and / or ambient humidity dimension, and the gate machine operating parameter dimension includes an operating voltage and / or operating current dimension, a door switch dimension, a CPU temperature, and / or a mainboard temperature dimension; The acquisition module 302 is further configured to do at least one of the following: Using the temperature sensor and / or humidity sensor configured on the gate to be monitored, collect the temperature value and / or humidity value at each time point under the ambient temperature and / or ambient humidity dimensions, and use the temperature value and / or humidity value as the target influencing parameter; Using the voltage sensor and / or current sensor configured on the gate to be monitored, collect the voltage value and / or current value at each time point under the working voltage and / or working current dimensions, and use the voltage value and / or current value as the target influencing parameter; Using the door switch sensor configured on the gate to be monitored, the cumulative number of door openings and closings at each time point in the door opening and closing dimension is collected, and the cumulative number of door openings and closings is used as the target influencing parameter; Utilize the CPU temperature sensor and / or motherboard sensor configured on the gate to be monitored to collect the CPU temperature values and / or motherboard temperature values at each time point under the CPU temperature and / or motherboard temperature dimensions, and use the CPU temperature values and / or motherboard temperature values as target influencing parameters.

[0091] Optionally, the fusion module 304 is further configured to: Input the target parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the fault detection model. In the graph aggregation layer, a global label is added before the target parameter sequence corresponding to each fault impact dimension to obtain an updated parameter sequence for each fault impact dimension. The global label is used to extract the global information of the corresponding target parameter sequence. Encode the update parameter sequence of each fault impact dimension into a graph node to obtain the corresponding graph structure; Aggregate each graph node in the graph structure and fuse the target parameters of the same time range and different fault impact dimensions to obtain the fused impact representation.

[0092] Optionally, the time series processing layer includes an attention mechanism and a classification layer; the acquisition module 306 is further configured to: Input the fusion influence representation into the attention mechanism of the temporal processing layer to obtain the corresponding temporal fusion encoding; The time series fusion code is input into the classification layer of the time series processing layer to obtain the failure probability of the gate to be monitored after the current time.

[0093] Optionally, the device further includes a prompt module configured to: According to the set maintenance judgment strategy and failure probability, the maintenance prompts of the gate to be monitored are determined.

[0094] Optionally, the maintenance judgment strategy includes issuing a maintenance alarm when the gate failure probability is greater than a probability threshold; The prompt module is further configured to: determining whether the probability of failure is greater than a probability threshold; If it is greater than the probability threshold, the attribute information of the gate to be monitored is obtained, and a maintenance alarm message of the gate to be monitored is generated based on the attribute information, and the maintenance alarm message is pushed to the set recipient.

[0095] Optionally, the device further includes a training module configured to: Obtaining sample data and fault probability labels corresponding to the sample gate, wherein the sample data includes sample parameter sequences corresponding to at least two fault impact dimensions before the sample time; Input the sample parameter sequences corresponding to at least two fault impact dimensions into the graph aggregation layer of the initial detection model. The graph aggregation layer fuses the sample parameters of the sample parameter sequences in the same time range and different fault impact dimensions to obtain a sample fusion representation. Input the sample fusion representation into the time series processing layer of the initial detection model to obtain the predicted failure probability of the sample gate after the sample time; According to the fault probability label and the predicted fault probability, the initial detection model is trained until the training stop condition is reached, and a trained fault detection model is obtained.

[0096] The embodiments of this specification provide a gate machine fault prediction device, including an acquisition module, a fusion module and an acquisition module. The modules interact with each other to comprehensively consider the parameters under multiple fault influencing dimensions that affect the gate machine fault, and predict the probability of the gate machine fault after the current time, so as to predict in advance whether the gate machine may fail and perform maintenance in time to avoid the gate machine failure causing harm to users and avoiding affecting the normal operation of rail transit; and, the graph aggregation layer of the fault detection model is used to spatially fuse the target parameters of the same time range and different fault influencing dimensions, comprehensively analyze the mutual dependence between multiple fault influencing dimensions in the same time range, and then perform time series analysis on the spatial fusion results to determine the failure probability of the gate machine to be monitored after the current time. The spatial and temporal dependence between the fault influencing factors of multiple dimensions is considered in the fault prediction, and the data basis for the gate machine fault prediction is more comprehensive, which improves the prediction accuracy of the failure probability of the gate machine to be monitored.

[0097] The above is a schematic diagram of a gate fault prediction device according to this embodiment. It should be noted that the technical solution of the gate fault prediction device and the technical solution of the gate fault prediction method described above are based on the same concept. For details not described in detail in the technical solution of the gate fault prediction device, please refer to the description of the technical solution of the gate fault prediction method described above.

[0098] Figure 4 4 shows a block diagram of a computing device according to one embodiment of the present disclosure. Components of the computing device 400 include, but are not limited to, a memory 410 and a processor 420. The processor 420 is connected to the memory 410 via a bus 430, and a database 450 is used to store data.

[0099] Computing device 400 also includes an access device 440 that enables computing device 400 to communicate via one or more networks 460. Examples of such networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. Access device 440 may include one or more of any type of network interface (e.g., a network interface controller (NIC)) whether wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, or a near field communication (NFC) interface.

[0100] In one embodiment of the present specification, the above components of the computing device 400 and Figure 4 Other components not shown in the figure may also be connected to each other, for example, via a bus. Figure 4 The computing device structure block diagram shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art may add or replace other components as needed.

[0101] Computing device 400 can be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, personal digital assistant, laptop computer, notebook computer, netbook computer, etc.), a mobile phone (e.g., a smartphone), a wearable computing device (e.g., a smartwatch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or personal computer (PC). Computing device 400 can also be a mobile or stationary server.

[0102] The processor 420 is configured to execute the following computer executable instructions, which, when executed by the processor, implement the steps of the gate fault prediction method.

[0103] The above is a schematic diagram of a computing device according to this embodiment. It should be noted that the technical solution of the computing device and the technical solution of the gate fault prediction method described above are based on the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the gate fault prediction method described above.

[0104] An embodiment of the present specification further provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the gate fault prediction method.

[0105] The above is a schematic diagram of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium is based on the same concept as the technical solution of the gate fault prediction method described above. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the gate fault prediction method described above.

[0106] An embodiment of the present specification further provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the gate fault prediction method described above.

[0107] The above is a schematic diagram of a computer program according to this embodiment. It should be noted that the technical solution of the computer program and the technical solution of the gate fault prediction method described above are based on the same concept. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the gate fault prediction method described above.

[0108] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0109] Computer instructions include computer program code, which may be in source code, object code, executable files, or some intermediate form. Computer-readable media may include any entity or device capable of carrying computer program code, recording media, USB flash drives, removable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signals, telecommunications signals, and software distribution media. It should be noted that the content of computer-readable media may be appropriately expanded or reduced based on the requirements of patent practice. For example, in some jurisdictions, according to patent practice, computer-readable media does not include electric carrier signals or telecommunications signals.

[0110] It should be noted that for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0111] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0112] The preferred embodiments disclosed above are intended only to help illustrate this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made based on the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that those skilled in the art can better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A gate fault prediction method, characterized in that: include: For the gate to be monitored, obtain the target parameter sequence corresponding to at least two fault impact dimensions before the current time; Inputting the target parameter sequences corresponding to the at least two fault impact dimensions into a graph aggregation layer of a fault detection model, and fusing the target parameters of the target parameter sequences in the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, wherein the fault detection model is a pre-trained time series data model; The fused impact representation is input into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

2. The gate fault prediction method according to claim 1, characterized in that: The step of obtaining target parameter sequences corresponding to at least two fault impact dimensions before the current time for the gate to be monitored includes: Using at least two sensors configured on the gate to be monitored, collecting and storing target impact parameters of at least two fault impact dimensions at various time points; Obtain target impact parameters of at least two fault impact dimensions at each time point before the current time and within a set time period, align them according to time, and obtain target parameter sequences corresponding to the at least two fault impact dimensions before the current time.

3. The gate fault prediction method according to claim 2, characterized in that: The at least two fault impact dimensions include the gate machine surrounding environment dimension and the gate machine's own operating parameter dimension, wherein the gate machine surrounding environment dimension includes the ambient temperature and / or ambient humidity dimension, and the gate machine's own operating parameter dimension includes the operating voltage and / or operating current dimension, the fan door switch dimension, the CPU temperature and / or the mainboard temperature dimension; The method of collecting and storing target impact parameters of at least two fault impact dimensions at each time point using at least two sensors configured on the gate to be monitored includes at least one of the following: Utilizing the temperature sensor and / or humidity sensor configured on the gate to be monitored, collecting the temperature value and / or humidity value at each time point under the ambient temperature and / or ambient humidity dimensions, and using the temperature value and / or humidity value as the target influencing parameter; Using the voltage sensor and / or current sensor configured on the gate to be monitored, collect the voltage value and / or current value at each time point under the working voltage and / or working current dimension, and use the voltage value and / or current value as the target influencing parameter; Using the door switch sensor configured on the gate to be monitored, the cumulative number of door switches at each time point under the door switch dimension is collected, and the cumulative number of door switches is used as the target influencing parameter; Utilize the CPU temperature sensor and / or mainboard sensor configured on the gate to be monitored to collect the CPU temperature value and / or mainboard temperature value at each time point under the CPU temperature and / or mainboard temperature dimensions, and use the CPU temperature value and / or mainboard temperature value as the target influencing parameter.

4. The gate fault prediction method according to claim 1, characterized in that: Inputting the target parameter sequences corresponding to the at least two fault impact dimensions into the graph aggregation layer of the fault detection model, and fusing the target parameters of the target parameter sequences in the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, includes: Inputting the target parameter sequences corresponding to the at least two fault impact dimensions into a graph aggregation layer of a fault detection model, and adding a global label in front of the target parameter sequence corresponding to each fault impact dimension in the graph aggregation layer to obtain an updated parameter sequence for each fault impact dimension, wherein the global label is used to extract global information of the corresponding target parameter sequence; Encoding the update parameter sequences of the various fault impact dimensions into graph nodes to obtain a corresponding graph structure; Aggregate the graph nodes in the graph structure, fuse the target parameters in the same time range and different fault impact dimensions, and obtain the fused impact representation.

5. The gate fault prediction method according to claim 1, characterized in that: The temporal processing layer includes an attention mechanism and a classification layer; Inputting the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time includes: Inputting the fusion influence representation into the attention mechanism of the temporal processing layer to obtain the corresponding temporal fusion encoding; The time series fusion code is input into the classification layer of the time series processing layer to obtain the failure probability of the gate to be monitored after the current time.

6. The gate fault prediction method according to claim 1, characterized in that: After inputting the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time, the method further includes: According to the set maintenance judgment strategy and the failure probability, the maintenance prompt of the gate to be monitored is determined.

7. The gate fault prediction method according to claim 6, characterized in that: The maintenance judgment strategy includes issuing a maintenance alarm when the gate failure probability is greater than a probability threshold; The step of determining the maintenance prompt for the gate to be monitored based on the set maintenance judgment strategy and the failure probability includes: determining whether the failure probability is greater than a probability threshold; If it is greater than the probability threshold, the attribute information of the gate to be monitored is obtained, and a maintenance alarm message of the gate to be monitored is generated according to the attribute information, and the maintenance alarm message is pushed to the set recipient.

8. The gate fault prediction method according to claim 1, characterized in that: The fault detection model is trained in the following way: Obtaining sample data and fault probability labels corresponding to the sample gate, wherein the sample data includes sample parameter sequences corresponding to at least two fault impact dimensions before the sample time; Inputting the sample parameter sequences corresponding to the at least two fault impact dimensions into the graph aggregation layer of the initial detection model, and fusing the sample parameters of the sample parameter sequences in the same time range and different fault impact dimensions through the graph aggregation layer to obtain a sample fusion representation; Inputting the sample fusion representation into the time series processing layer of the initial detection model to obtain the predicted failure probability of the sample gate after the sample time; The initial detection model is trained according to the fault probability label and the predicted fault probability until a training stop condition is reached, thereby obtaining a trained fault detection model.

9. A gate fault prediction device, characterized in that: include: An acquisition module is configured to acquire, for the gate to be monitored, a target parameter sequence corresponding to at least two fault impact dimensions before a current time; a fusion module configured to input the target parameter sequences corresponding to the at least two fault impact dimensions into a graph aggregation layer of a fault detection model, and fuse the target parameters of the target parameter sequences within the same time range and different fault impact dimensions through the graph aggregation layer to obtain a fused impact representation, wherein the fault detection model is a pre-trained time series data model; The acquisition module is configured to input the fused impact representation into the time series processing layer of the fault detection model to obtain the failure probability of the gate to be monitored after the current time.

10. A computing device, characterized in that include: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the gate fault prediction method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium, characterized in that It stores computer-executable instructions, which, when executed by a processor, implement the steps of the gate fault prediction method according to any one of claims 1 to 8.

12. A computer program product, characterized in that The invention comprises a computer program / instruction, which, when executed by a processor, implements the steps of the gate fault prediction method according to any one of claims 1 to 8.

Images