Method and device for detecting platform door failure, electronic equipment and storage medium

Abstract

The application discloses a kind of platform door failure detection method and device, electronic equipment and storage medium.Therein, the method comprises: the runtime data of target platform door is acquired in real time;Extract the operation data characteristics of the runtime data;Call sample feature library, wherein the sample feature library includes normal sample feature library and abnormal sample feature library;The running data characteristics are compared using the sample feature library, and the fault detection result of the target platform door is output based on the comparison result.By the application, the technical problem of low accuracy of platform door fault identification, long fault identification time and high computing resource occupancy in related art is solved, false positives are not easy to occur, and the real-time performance and accuracy are better compared with traditional deep learning algorithms.

Description

Technical Field

[0001] This invention relates to the field of safety supervision, and more specifically, to a method and apparatus for detecting platform screen door malfunctions, electronic equipment, and storage medium. Background Technology

[0002] With the rise of artificial intelligence technology, in urban rail transit signaling systems based on vehicle-to-vehicle communication (TACS), the control and status acquisition of platform screen doors are completed by the trackside target controller. When the target controller acquires a true status for the platform screen door to be closed and locked, it means that the platform screen door is closed and locked, and the train can move freely inside the platform. When this status is false, it means that the platform screen door may be open, and the train cannot move in the platform area.

[0003] When detecting faults in platform screen doors, the relevant technologies are generally based on vision and deep learning detection methods. Fault detection using deep learning technology requires data labeling, which is a supervised learning process and requires a large number of samples. It is difficult to balance recognition accuracy and computation time. When using deep learning algorithms such as convolutional neural networks, the accuracy is high, but the computation time is long. When using simple formulas, the computation time is short, but the accuracy is low.

[0004] In terms of fault data acquisition, the main method is to increase the number of sensors. This method is highly dependent on multi-source data and requires the relevant data to be labeled, which results in high labor costs.

[0005] No effective solutions have yet been found to address the aforementioned problems in the relevant technologies. Summary of the Invention

[0006] This invention provides a method and apparatus for detecting platform screen door malfunctions, as well as electronic equipment and storage medium.

[0007] According to one aspect of the embodiments of this application, a method for detecting platform screen door faults is provided. The method includes: acquiring runtime sequence data of a target platform screen door in real time; extracting runtime data features from the runtime sequence data; retrieving a sample feature library, wherein the sample feature library includes a normal sample feature library and an abnormal sample feature library; comparing the runtime data features with the sample feature library, and outputting a fault detection result of the target platform screen door based on the comparison result.

[0008] Furthermore, comparing the operational data features with the sample feature library and outputting the fault detection result of the target platform gate based on the comparison result includes: comparing the operational data features with the normal sample feature library to obtain a first feature difference value; determining whether the first feature difference value exceeds a first threshold; and if the first feature difference value does not exceed the first threshold, outputting the normal operation result of the target platform gate.

[0009] Furthermore, after determining whether the first feature difference value exceeds the first threshold, the method further includes: if the first feature difference value exceeds the first threshold, comparing the operating data features with the abnormal sample feature library to obtain a second feature difference value; determining whether the second feature difference value exceeds the second threshold; if the second feature difference value does not exceed the second threshold, outputting the fault operation result of the target platform door.

[0010] Furthermore, outputting the fault operation result of the target platform gate includes: classifying and acquiring multiple sub-operational data features contained in the operation data features, wherein the multiple sub-operational data features include data features of at least one of the following physical quantities: speed curve, kinetic energy curve, input current, input voltage, and motor speed; for each target sub-operational data feature in the sub-operational data features, sequentially inputting it into a gated loop unit (GRU), wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; acquiring the incremental estimate of the target sub-operational data feature output by the GRU; adding the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; acquiring the actual physical quantity corresponding to the target sub-operational data feature, performing adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sub-operational data feature in the current time series period; and calculating the mean absolute error of the target sub-operational data feature based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. Let i be the current updated value at time i, and N be the time series length of the target sub-running data feature; calculate the difference between the MAE and the preset difference; determine the abnormal physical quantity whose difference exceeds the third threshold, and locate the fault type of the target platform door based on the abnormal physical quantity.

[0011] Furthermore, after determining whether the second feature difference value exceeds the second threshold, the method further includes: if the second feature difference value exceeds the second threshold, updating the sample feature library; acquiring the runtime sequence data of the target platform door in the next cycle in real time; comparing the runtime sequence data of the next cycle with the updated sample feature library, and outputting the fault detection result of the target platform door based on the comparison result.

[0012] Furthermore, before retrieving the sample feature library, the method further includes: acquiring historical time-series data generated by the platform gate during historical operation, wherein the data dimensions of the historical time-series data include at least one of the following: speed curve, kinetic energy curve, input current, input voltage, and motor speed; using the historical time-series data as training data, and dividing the training data into multiple sample data segments according to the time-series period to obtain a sample dataset; extracting the time-series features of each sample data segment in the sample dataset to obtain a sample feature set; and using an unsupervised classification model to identify the sample feature set into a normal sample feature library and an abnormal sample feature library.

[0013] Furthermore, the unsupervised classification model is used to identify the sample feature set into a normal sample feature library and an abnormal sample feature library, including: for each target sample feature in the sample feature set, inputting it sequentially into a gated recurrent unit (GRU) according to the time series period, wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; obtaining the incremental estimate of the target sample feature output by the GRU; adding the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current time series period; obtaining the actual physical quantity of the target sample data segment corresponding to the target sample feature, performing adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sample data segment in the current time series period; and calculating the mean absolute error of the target sample data segment based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. Let i be the current update value at time i, and N be the time period length of the target sample feature; compare the MAE with a preset difference; if the MAE is greater than the preset difference, output the target sample feature to the abnormal sample feature library; if the MAE is less than or equal to the preset difference, output the target sample feature to the normal sample feature library.

[0014] According to another aspect of the embodiments of this application, a platform screen door fault detection device is also provided, comprising: a first acquisition module, configured to acquire runtime sequence data of a target platform screen door in real time; a first extraction module, configured to extract runtime data features from the runtime sequence data; a retrieval module, configured to retrieve a sample feature library, wherein the sample feature library includes a normal sample feature library and an abnormal sample feature library; and a detection module, configured to compare the runtime data features with the sample feature library and output a fault detection result of the target platform screen door based on the comparison result.

[0015] Furthermore, the detection module includes: a first comparison unit, used to compare the operating data features with the normal sample feature library to obtain a first feature difference value; a first judgment unit, used to judge whether the first feature difference value exceeds a first threshold; and a first output unit, used to output the normal operation result of the target platform door if the first feature difference value does not exceed the first threshold.

[0016] Furthermore, the detection module further includes: a second comparison unit, used to compare the operating data features with the abnormal sample feature library to obtain a second feature difference value after the first judgment unit determines whether the first feature difference value exceeds the first threshold; a second judgment unit, used to determine whether the second feature difference value exceeds the second threshold; and a second output unit, used to output the fault operation result of the target platform door if the second feature difference value does not exceed the second threshold.

[0017] Further, the second output unit includes: a first acquisition subunit, used to classify and acquire multiple sub-operational data features included in the operation data features, wherein the multiple sub-operational data features include data features of at least one of the following physical quantities: velocity curve, kinetic energy curve, input current, input voltage, and motor speed; an input subunit, used to sequentially input each target sub-operational data feature in the sub-operational data features into a gated loop unit (GRU), wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; a second acquisition subunit, used to acquire the incremental estimate of the target sub-operational data feature output by the GRU; a first calculation subunit, used to add the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; a third acquisition subunit, used to acquire the actual physical quantity corresponding to the target sub-operational data feature, and perform adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sub-operational data feature in the current time series period; and a second calculation subunit, used to calculate the mean absolute error of the target sub-operational data feature based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. The current updated value of time i is N, and the time series length of the target sub-running data feature is N; the third operation sub-unit is used to calculate the difference between the MAE and the preset difference; the positioning sub-unit is used to determine the abnormal physical quantity whose difference exceeds the third threshold, and locate the fault type of the target platform door based on the abnormal physical quantity.

[0018] Furthermore, the detection module further includes: an update unit, used to update the sample feature library if the second feature difference value exceeds the second threshold after the second judgment unit determines whether the second feature difference value exceeds the second threshold; an acquisition unit, used to acquire the runtime sequence data of the target platform door in the next cycle in real time; and a third comparison unit, used to compare the runtime sequence data of the next cycle with the updated sample feature library, and output the fault detection result of the target platform door based on the comparison result.

[0019] Furthermore, the device further includes: a second acquisition module, used to acquire historical time-series data generated by the platform gate during historical operation before the retrieval module retrieves the sample feature library, wherein the data dimensions of the historical time-series data include at least one of the following: speed curve, kinetic energy curve, input current, input voltage, and motor speed; a segmentation module, used to use the historical time-series data as training data and divide the training data into multiple sample data segments according to the time-series period to obtain a sample dataset; a second extraction module, used to extract the time-series features of each sample data segment in the sample dataset to obtain a sample feature set; and an identification module, used to identify the sample feature set into a normal sample feature library and an abnormal sample feature library using an unsupervised classification model.

[0020] Further, the identification module includes: an input unit, used to sequentially input each target sample feature in the sample feature set into a gated recurrent unit (GRU) according to a time series period, wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; a first acquisition unit, used to acquire the incremental estimate of the target sample feature output by the GRU; a first calculation unit, used to add the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; a second acquisition unit, used to acquire the actual physical quantity of the target sample data segment corresponding to the target sample feature, and perform adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sample data segment in the current time series period; and a second calculation unit, used to calculate the mean absolute error of the target sample data segment based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. The current update value is given by time i, and N is the time period length of the target sample feature; a comparison unit is used to compare the MAE with a preset difference; an output unit is used to output the target sample feature to the abnormal sample feature library if the MAE is greater than the preset difference, and to output the target sample feature to the normal sample feature library if the MAE is less than or equal to the preset difference.

[0021] According to another aspect of the embodiments of this application, a storage medium is also provided, the storage medium including a stored program that executes the above steps when the program is run.

[0022] According to another aspect of the embodiments of this application, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; wherein: the memory is used to store computer programs; and the processor is used to execute the steps in the above method by running the programs stored in the memory.

[0023] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the steps in the above-described method.

[0024] This invention acquires real-time runtime sequence data of the target platform gate, extracts runtime data features from the runtime sequence data, and retrieves a sample feature library, which includes a normal sample feature library and an abnormal sample feature library. The runtime data features are compared using the sample feature library, and the fault detection result of the target platform gate is output based on the comparison result. By acquiring runtime sequence data, the changes in physical quantities of the platform gate over a complete time period can be detected. Unsupervised fault detection based on time sequence data can be achieved using the normal and abnormal sample feature libraries. This invention solves the technical problems of low fault identification accuracy, long fault identification time, and high computational resource consumption in related technologies for platform gates. It is less prone to false alarms and has better real-time performance and accuracy compared to traditional deep learning algorithms. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 This is a hardware structure block diagram of a computer according to an embodiment of the present invention;

[0027] Figure 2 This is a flowchart of a platform door fault detection method according to an embodiment of the present invention;

[0028] Figure 3 This is a logic diagram of the sample classification algorithm in this embodiment of the invention;

[0029] Figure 4 This is a schematic diagram of the classification results in the implementation of this invention;

[0030] Figure 5 This is a schematic diagram of the platform door fault detection method according to an embodiment of the present invention;

[0031] Figure 6 This is a structural block diagram of a platform door fault detection device according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all of them. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present application can be combined with each other.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] Example 1

[0035] The method embodiment provided in Embodiment 1 of this application can be executed in a controller, server, computer, tablet, or similar computing scheduling device. Taking running on a computer as an example, Figure 1 This is a hardware structure block diagram of a computer according to an embodiment of the present invention. For example... Figure 1 As shown, a computer may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. Optionally, the computer may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer described above. For example, the computer may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0036] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to a platform door fault detection method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the aforementioned method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0037] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a computer's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0038] This embodiment provides a method for detecting platform screen door malfunctions. Figure 2 This is a flowchart of a platform screen door fault detection method according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0039] Step S202: Obtain the runtime sequence data of the target platform door in real time;

[0040] In this embodiment, the runtime sequence data refers to the sequence of physical quantities over a continuous period of time when the target platform gate is working. The physical quantities can be the speed curve, kinetic energy curve, input current, input voltage, motor speed, etc. of the platform gate.

[0041] Step S204: Extract runtime data features from runtime sequence data;

[0042] Optionally, the time-series feature extraction tool tsfresh can be used to extract runtime data features from runtime time-series data.

[0043] Step S206: Retrieve the sample feature library, which includes a normal sample feature library and an abnormal sample feature library;

[0044] The sample feature library in this embodiment is obtained in advance from positive sample training data and negative sample training data.

[0045] Step S208: The features of the running data are compared with the sample feature library, and the fault detection results of the target platform door are output based on the comparison results.

[0046] The fault detection results in this embodiment include whether the platform screen door is normal or faulty. In addition, if the platform screen door is faulty, the fault type can be further identified.

[0047] Through the above steps, the runtime sequence data of the target platform gate is acquired in real time, the runtime data features are extracted, and a sample feature library is retrieved. The sample feature library includes a normal sample feature library and an abnormal sample feature library. The runtime data features are compared using the sample feature library, and the fault detection result of the target platform gate is output based on the comparison result. By acquiring runtime sequence data, the changes in physical quantities of the platform gate over a complete time period can be detected. Unsupervised fault detection based on time sequence data can be achieved using the normal sample feature library and the abnormal sample feature library. This solves the technical problems of low fault identification accuracy, long fault identification time, and high computational resource consumption of platform gates in related technologies. It is less likely to generate false alarms and has better real-time performance and accuracy compared with traditional deep learning algorithms.

[0048] In one embodiment of this example, comparing operational data features with a sample feature library and outputting the fault detection result of the target platform gate based on the comparison result includes:

[0049] S11, compare the features of the running data with the features of the normal sample feature library to obtain the first feature difference value;

[0050] Optionally, the correlation or similarity between the feature library of normal samples and the features of running data can be compared. The smaller the feature difference value, the stronger the correlation, and vice versa.

[0051] S12, determine whether the difference value of the first feature exceeds the first threshold;

[0052] S13, if the first feature difference value does not exceed the first threshold, output the normal operation result of the target platform door.

[0053] In some implementation scenarios, after determining whether the difference value of the first feature exceeds the first threshold, the following steps are also included:

[0054] S14. If the first feature difference value exceeds the first threshold, the features of the running data are compared with the abnormal sample feature library to obtain the second feature difference value.

[0055] S15, determine whether the difference value of the second feature exceeds the second threshold;

[0056] S16, if the difference value of the second feature does not exceed the second threshold, output the fault operation result of the target platform door.

[0057] In one example, the output of the target platform gate's fault operation result includes: classifying and obtaining multiple sub-operational data features contained in the operation data features, wherein the multiple sub-operational data features include data features of at least one of the following physical quantities: velocity curve, kinetic energy curve, input current, input voltage, and motor speed; for each target sub-operational data feature, sequentially inputting it into a gated recurrent unit (GRU), wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; obtaining the incremental estimate of the target sub-operational data feature output by the GRU; adding the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; obtaining the actual physical quantity corresponding to the target sub-operational data feature, performing adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sub-operational data feature in the current time series period; and calculating the mean absolute error (MAE) of the target sub-operational data feature based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. Let i be the current updated value at time i, and N be the time series length of the target sub-running data feature; calculate the difference between MAE and the preset difference; determine the abnormal physical quantity whose difference exceeds the third threshold, and locate the target platform door fault type based on the abnormal physical quantity.

[0058] In one example, after determining whether the difference value of the second feature exceeds the second threshold, the method further includes: if the difference value of the second feature exceeds the second threshold, updating the sample feature library; acquiring the runtime sequence data of the target platform door in the next cycle in real time; comparing the runtime sequence data of the next cycle with the updated sample feature library, and outputting the fault detection result of the target platform door based on the comparison result.

[0059] In this example, if the difference value of the second feature exceeds the second threshold, and the correlation coefficient exceeds the threshold, it is considered that the model is not applicable or the amount of data is insufficient. The training database needs to be updated and the detection needs to be repeated.

[0060] In this embodiment, before retrieving the sample feature library, the method further includes: acquiring historical time-series data generated by the platform gate during historical operation, wherein the data dimensions of the historical time-series data include at least one of the following: speed curve, kinetic energy curve, input current, input voltage, and motor speed; using the historical time-series data as training data, and dividing the training data into multiple sample data segments according to the time-series period to obtain a sample dataset; extracting the time-series features of each sample data segment in the sample dataset to obtain a sample feature set; and using an unsupervised classification model to identify the sample feature set into a normal sample feature library and an abnormal sample feature library.

[0061] Optionally, using an unsupervised classification model to identify the sample feature set into normal and abnormal sample feature libraries includes: for each target sample feature in the sample feature set, inputting it sequentially into a gated recurrent unit (GRU) according to the time series, wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; obtaining the incremental estimate of the target sample feature output by the GRU; adding the incremental estimate to the historical update value of the previous time series to obtain the estimate value of the current time series; obtaining the actual physical quantity of the target sample data segment corresponding to the target sample feature, performing adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sample data segment in the current time series; and calculating the mean absolute error of the target sample data segment based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. is the current update value at time i, and N is the time period length of the target sample feature; compare MAE with the preset difference; if MAE is greater than the preset difference, output the target sample feature to the abnormal sample feature library; if MAE is less than or equal to the preset difference, output the target sample feature to the normal sample feature library.

[0062] In one example, the physical quantity of the target sample feature is the velocity value of the velocity curve, the time period length is 1 minute, the velocity value of the target sample data segment corresponding to the target sample feature is the velocity value sampled at 10-second intervals, which are u1~u6 respectively, then N is 6, and i takes the value 1~6.

[0063] Figure 3 This is a logic diagram of the sample classification algorithm in this embodiment of the invention, including:

[0064] Step 1: Import the time series data generated during the historical operation of the platform screen door into the model as training data. The training data includes the platform screen door speed curve, kinetic energy curve, input current, input voltage, motor speed, etc.

[0065] Step 2: Divide the imported training time series data into periodic time periods and extract features. You can use the time series feature extraction tool tsfresh and select the features with strong correlation to retain.

[0066] Step 3: Design an unsupervised sample classification method to identify platform door training data as normal and abnormal samples. This algorithm combines Adaptive Kalman Filter (AKF) and GRU to solve the gradient vanishing and gradient exploding phenomena during the training process of time-series detection models. Compared with traditional periodic detection methods, this method has the advantages of high recognition accuracy and low false positive rate. The algorithm logic is as follows: Figure 3 As shown.

[0067] Taking platform screen door speed as an example, the features filtered by tsfresh are input into a GRU, passing through the input layer, hidden layer, and output layer, and the output is the incremental estimate. The estimated value is obtained by adding the updated value to the actual physical quantity. The estimated value is then processed with the actual physical quantity using AKF to obtain the updated value of the platform gate speed. The updated value is then subtracted from the actual physical quantity to obtain the difference between the two. This difference is then used for post-processing. The post-processing method involves calculating the mean absolute error (MAE) as the criterion for judging the normal sample feature library and the abnormal sample feature library. The formula for calculating the mean absolute error (MAE) is as follows:

[0068] ;

[0069] When the MAE is greater than the preset difference, the samples in that time period are considered abnormal samples, and the corresponding sample features are output to the abnormal sample feature library. When the MAE is less than or equal to the preset difference, the samples in that time period are considered normal samples, and the corresponding sample features are output to the normal sample feature library.

[0070] Figure 4 This is a schematic diagram of the classification results in the implementation of the present invention, illustrating the classification results of normal samples and abnormal samples, wherein the slope of abnormal samples changes more significantly.

[0071] After step 3, the data model training is complete, and real-time testing can be performed based on the model to detect the operating status of the platform gate, and then proceed to step 4.

[0072] Step 4: Import the time-series data generated during the real-time operation of the platform gate into the model as test data. The dimensions of the test data are consistent with those of the training data.

[0073] Step 5: Extract features from the imported real-time runtime sequence data. The features used are those retained in Step 2.

[0074] Step 6: Compare the features of the running data with the features of the normal sample library. This comparison method is a general correlation coefficient comparison method. Determine whether the correlation coefficient between the features of the running data and the features of the normal sample exceeds the threshold. If the coefficient is within the threshold range, proceed to Step 7; if the coefficient exceeds the threshold, proceed to Step 8.

[0075] Step 7: If the coefficients in Step 6 are within the threshold range, the system is considered to be operating normally, and the normal operation result is output.

[0076] Step 8: For running data that does not match the normal sample feature library, compare it with the abnormal sample feature library. Similarly, this comparison method is the general correlation coefficient comparison method. Determine whether the correlation coefficient between the running data features and the abnormal sample features exceeds the threshold. If the coefficient is within the threshold range, proceed to step 9; if the coefficient exceeds the threshold, proceed to step 10.

[0077] Step 9: When the correlation coefficient is within the threshold range, the platform screen door is considered faulty. The fault status result is output, and the fault type is determined. Fault type determination method: Various time-series data generated during real-time operation (platform screen door speed, platform screen door kinetic energy, input voltage, input current, motor speed, etc.) are imported into the sample classification method in Step 3 to obtain the specific physical quantities whose differences exceed the threshold. The corresponding points are the types of faults that have occurred.

[0078] Step 10: When the correlation coefficient exceeds the threshold, it is considered that the model is not applicable or the amount of data is insufficient. The training database needs to be updated, and the process returns to Step 1.

[0079] Figure 5 This is a schematic diagram of a platform screen door fault detection method according to an embodiment of the present invention, including:

[0080] Step 1: Import platform door training time series data from the platform door training database as the training set;

[0081] Step 2: Extract features from the imported training time series data;

[0082] Step 3: Use an unsupervised learning algorithm to identify the input training time series data as normal samples and abnormal samples, and classify them into a normal sample feature library and an abnormal sample feature library.

[0083] Step 4: Import platform door runtime sequence data from the platform door operation database as a test set;

[0084] Step 5: Extract features from the imported runtime sequence data;

[0085] Step 6: Compare the features of the running data with the feature library of normal samples, and determine whether the difference exceeds the threshold. If the difference is within the threshold range, proceed to step 7; if the difference exceeds the threshold, proceed to step 8.

[0086] Step 7: Output the normal operation result of the platform screen door;

[0087] Step 8: Compare the features of the running data with the feature library of abnormal samples, and determine whether the difference exceeds the threshold. If the difference is within the threshold range, proceed to step 9; if the difference exceeds the threshold, proceed to step 10.

[0088] Step 9: The platform door is in a fault state, and the fault type is determined;

[0089] Step 10: Update the training database and return to Step 1.

[0090] This embodiment provides a platform screen door fault detection method. It uses an improved AKF+GRU combined algorithm to classify platform screen door fault levels, solving the problems of low fault identification accuracy, long fault identification time, and high computational resource consumption in traditional platform screen door operation and maintenance. The fault detection method provided in this embodiment uses an unsupervised learning method to learn from small sample data, resulting in low cost. It includes an unsupervised classification method for dividing platform screen door training samples into normal and abnormal samples, a platform screen door fault type judgment method, and a platform screen door fault detection method based on time-series data. This embodiment acquires relevant physical quantities and extracts relevant features based on existing equipment, realizing an unsupervised fault detection method based on time-series data, and can determine the fault type of platform screen doors.

[0091] Compared with existing algorithms such as SeasonalAD and sARIMA, the anomaly identification method proposed in this embodiment is for detection of complete time periods, which is less likely to produce false alarms. Compared with traditional deep learning algorithms, this embodiment has better real-time performance and accuracy.

[0092] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0093] Example 2

[0094] This embodiment also provides a platform screen door malfunction detection device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0095] Figure 6 This is a structural block diagram of a platform door fault detection device according to an embodiment of the present invention, as shown below. Figure 6 As shown, the device includes: a first acquisition module 60, a first extraction module 62, a retrieval module 64, and a detection module 66, wherein...

[0096] The first acquisition module 60 is used to acquire the runtime sequence data of the target platform door in real time;

[0097] The first extraction module 62 is used to extract the runtime data features of the runtime sequence data;

[0098] The retrieval module 64 is used to retrieve the sample feature library, wherein the sample feature library includes a normal sample feature library and an abnormal sample feature library;

[0099] The detection module 66 is used to compare the running data features with the sample feature library and output the fault detection result of the target platform door based on the comparison result.

[0100] Optionally, the detection module includes: a first comparison unit, used to compare the operating data features with the normal sample feature library to obtain a first feature difference value; a first judgment unit, used to judge whether the first feature difference value exceeds a first threshold; and a first output unit, used to output the normal operation result of the target platform door if the first feature difference value does not exceed the first threshold.

[0101] Optionally, the detection module further includes: a second comparison unit, configured to, after the first judgment unit determines whether the first feature difference value exceeds the first threshold, compare the operating data features with the abnormal sample feature library to obtain a second feature difference value if the first feature difference value exceeds the first threshold; a second judgment unit, configured to determine whether the second feature difference value exceeds the second threshold; and a second output unit, configured to, if the second feature difference value does not exceed the second threshold, output the fault operation result of the target platform door.

[0102] Optionally, the second output unit includes: a first acquisition subunit, used to classify and acquire multiple sub-operational data features included in the operation data features, wherein the multiple sub-operational data features include data features of at least one of the following physical quantities: velocity curve, kinetic energy curve, input current, input voltage, and motor speed; an input subunit, used to sequentially input each target sub-operational data feature in the sub-operational data features into a gated loop unit (GRU), wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; a second acquisition subunit, used to acquire the incremental estimate of the target sub-operational data feature output by the GRU; a first calculation subunit, used to add the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; a third acquisition subunit, used to acquire the actual physical quantity corresponding to the target sub-operational data feature, and perform adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sub-operational data feature in the current time series period; and a second calculation subunit, used to calculate the mean absolute error of the target sub-operational data feature based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. The current updated value of time i is N, and the time series length of the target sub-running data feature is N; the third operation sub-unit is used to calculate the difference between the MAE and the preset difference; the positioning sub-unit is used to determine the abnormal physical quantity whose difference exceeds the third threshold, and locate the fault type of the target platform door based on the abnormal physical quantity.

[0103] Optionally, the detection module further includes: an update unit, configured to update the sample feature library if the second feature difference value exceeds the second threshold after the second judgment unit determines whether the second feature difference value exceeds the second threshold; an acquisition unit, configured to acquire the runtime sequence data of the target platform door in the next cycle in real time; and a third comparison unit, configured to compare the runtime sequence data of the next cycle with the updated sample feature library, and output the fault detection result of the target platform door based on the comparison result.

[0104] Optionally, the device further includes: a second acquisition module, configured to acquire historical time-series data generated by the platform gate during historical operation before the retrieval module retrieves the sample feature library, wherein the data dimensions of the historical time-series data include at least one of the following: speed curve, kinetic energy curve, input current, input voltage, and motor speed; a segmentation module, configured to use the historical time-series data as training data and divide the training data into multiple sample data segments according to the time-series period to obtain a sample dataset; a second extraction module, configured to extract the time-series features of each sample data segment in the sample dataset to obtain a sample feature set; and an identification module, configured to use an unsupervised classification model to identify the sample feature set into a normal sample feature library and an abnormal sample feature library.

[0105] Optionally, the recognition module includes: an input unit, used to sequentially input each target sample feature in the sample feature set into a gated recurrent unit (GRU) according to a time series period, wherein the GRU includes an input layer, a hidden layer, and an output layer connected sequentially; a first acquisition unit, used to acquire the incremental estimate of the target sample feature output by the GRU; a first calculation unit, used to add the incremental estimate to the historical update value of the previous time series period to obtain the estimate value of the current period; a second acquisition unit, used to acquire the actual physical quantity of the target sample data segment corresponding to the target sample feature, and perform adaptive Kalman filtering on the estimate value and the actual physical quantity to obtain the current update value of the target sample data segment in the current time series period; and a second calculation unit, used to calculate the mean absolute error of the target sample data segment based on the actual physical quantity and the current update value using the following formula: ,in, Let i be the actual physical quantity at time i. The current update value is given by time i, and N is the time period length of the target sample feature; a comparison unit is used to compare the MAE with a preset difference; an output unit is used to output the target sample feature to the abnormal sample feature library if the MAE is greater than the preset difference, and to output the target sample feature to the normal sample feature library if the MAE is less than or equal to the preset difference.

[0106] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0107] Example 3

[0108] Embodiments of the present invention also provide a storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above method embodiments when running.

[0109] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0110] S1, real-time acquisition of runtime sequence data of the target platform door;

[0111] S2, extract the runtime data features of the runtime sequence data;

[0112] S3, retrieve the sample feature library, wherein the sample feature library includes a normal sample feature library and an abnormal sample feature library;

[0113] S4, compare the operational data features with the sample feature library, and output the fault detection result of the target platform door based on the comparison result.

[0114] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0115] Embodiments of the present invention also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0116] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0117] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0118] S1, real-time acquisition of runtime sequence data of the target platform door;

[0119] S2, extract the runtime data features of the runtime sequence data;

[0120] S3, retrieve the sample feature library, wherein the sample feature library includes a normal sample feature library and an abnormal sample feature library;

[0121] S4, compare the operational data features with the sample feature library, and output the fault detection result of the target platform door based on the comparison result.

[0122] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0123] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0124] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0125] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0126] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0127] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0128] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0129] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method of detecting a failure of a platform door, characterized by The method comprises: acquiring running timing data of a target platform door in real time; extracting running data features of the running timing data; calling a sample feature library, wherein the sample feature library comprises a normal sample feature library and an abnormal sample feature library; comparing the running data features with the sample feature library, and outputting a fault detection result of the target platform door based on a comparison result; comparing the running data features with the sample feature library, and outputting a fault detection result of the target platform door based on a comparison result comprises: comparing the running data features with the normal sample feature library to obtain a first feature difference value; determining whether the first feature difference value exceeds a first threshold value; if the first feature difference value does not exceed the first threshold value, outputting a normal running result of the target platform door; after determining whether the first feature difference value exceeds the first threshold value, the method further comprises: if the first feature difference value exceeds the first threshold value, comparing the running data features with the abnormal sample feature library to obtain a second feature difference value; determining whether the second feature difference value exceeds a second threshold value; if the second feature difference value does not exceed the second threshold value, outputting a fault running result of the target platform door; outputting the fault running result of the target platform door comprises: acquiring a plurality of sub-running data features contained in the running data features in a classified manner, wherein the plurality of sub-running data features comprise data features of at least one of the following physical quantities: a speed curve, a kinetic energy curve, an input current, an input voltage, and a motor speed; for each target sub-running data feature in the sub-running data features, sequentially inputting a gated recurrent unit (GRU), wherein the GRU comprises an input layer, a hidden layer, and an output layer connected in sequence; acquiring an increment estimate output by the target sub-running data feature through the GRU; adding the increment estimate to a historical update value of a previous timing period to obtain an estimated value of a current period; acquiring an actual physical quantity corresponding to the target sub-running data feature, and performing adaptive Kalman filtering on the estimated value and the actual physical quantity to obtain a current update value of the target sub-running data feature in the current timing period; based on the actual physical quantity and the current update value, calculating a mean absolute error (MAE) of the target sub-running data feature using the following formula: wherein, is the actual physical quantity at time i, is the current updated value at time i, and N is the time series length of the target sub-operation data feature; calculating a difference between the MAE and a preset difference value; determining an abnormal physical quantity whose difference value exceeds a third threshold value, and positioning a fault type of the target platform door based on the abnormal physical quantity.

2. The method of claim 1, wherein, after determining whether the second feature difference value exceeds the second threshold value, the method further comprises: if the second feature difference value exceeds the second threshold value, updating the sample feature library; acquiring running timing data of the target platform door in a next period in real time; comparing the running timing data in the next period with the updated sample feature library, and outputting a fault detection result of the target platform door based on a comparison result.

3. The method of claim 1, wherein, before calling the sample feature library, the method further comprises: acquire historical time series data generated by the platform door at a historical time, wherein a data dimension of the historical time series data comprises at least one of a speed curve, a kinetic energy curve, an input current, an input voltage, and a motor speed; divide the historical time series data into a plurality of sample data segments as training data according to a time sequence period, to obtain a sample data set; extract time sequence features of each sample data segment in the sample data set, to obtain a sample feature set; identify the sample feature set into a normal sample feature library and an abnormal sample feature library using an unsupervised classification model.

4. The method of claim 3, wherein, identifying the sample feature set into a normal sample feature library and an abnormal sample feature library using an unsupervised classification model comprises: for each target sample feature in the sample feature set, sequentially input a gated recurrent unit (GRU) according to a time sequence period, wherein the GRU comprises an input layer, a hidden layer, and an output layer connected in sequence; acquire an increment estimate output by the GRU for the target sample feature; add the increment estimate to a historical update value of a previous time sequence period to obtain an estimated value of a current period; acquire an actual physical quantity of a target sample data segment corresponding to the target sample feature, and perform adaptive Kalman filtering on the estimated value and the actual physical quantity to obtain a current update value of the target sample data segment in the current time sequence period; based on the actual physical quantity and the current update value, calculate a mean absolute error (MAE) of the target sample data segment using the following formula: wherein, is the actual physical quantity at time i, is the current updated value at time i, and N is the length of the time series period of the target sample feature. compare the MAE with a preset difference value; if the MAE is greater than the preset difference value, output the target sample feature to the abnormal sample feature library; if the MAE is less than or equal to the preset difference value, output the target sample feature to the normal sample feature library.

5. A device for detecting a failure of a platform door, characterized in that comprises: a first acquisition module for acquiring real-time running time sequence data of a target platform door; a first extraction module for extracting running data features of the running time sequence data; a calling module for calling a sample feature library, wherein the sample feature library comprises a normal sample feature library and an abnormal sample feature library; a detection module for comparing the running data features with the sample feature library, and outputting a fault detection result of the target platform door based on a comparison result, comprising: comparing the running data features with the normal sample feature library to obtain a first feature difference value; determining whether the first feature difference value exceeds a first threshold value; if the first feature difference value does not exceed the first threshold value, output a normal running result of the target platform door; after determining whether the first feature difference value exceeds the first threshold value, further comprising: if the first feature difference value exceeds the first threshold value, comparing the running data features with the abnormal sample feature library to obtain a second feature difference value; determining whether the second feature difference value exceeds a second threshold value; if the second feature difference value does not exceed the second threshold value, output a fault running result of the target platform door; outputting the fault running result of the target platform door comprises: The operation data features are classified to obtain a plurality of sub-operation data features, wherein the plurality of sub-operation data features include at least one of the following physical quantity data features: a speed curve, a kinetic energy curve, an input current, an input voltage, and a motor speed; For each target sub-operation data feature in the sub-operation data features, a gated recurrent unit (GRU) is sequentially inputted, wherein the GRU includes an input layer, a hidden layer, and an output layer connected in sequence; An increment estimation output by the GRU for the target sub-operation data feature is obtained; The increment estimation is added to a historical update value of a previous time sequence period to obtain an estimation value of a current period; An actual physical quantity corresponding to the target sub-operation data feature is obtained, and the estimation value is adaptively Kalman filtered with the actual physical quantity to obtain a current update value of the target sub-operation data feature in the current time sequence period; Based on the actual physical quantity and the current update value, a mean absolute error (MAE) of the target sub-operation data feature is calculated by using the following formula: wherein, is the actual physical quantity at time i, is the current updated value at time i, and N is the time series length of the target sub-operation data feature; A difference between the MAE and a preset difference value is calculated; An abnormal physical quantity whose difference exceeds a third threshold value is determined, and a target station door fault type is located based on the abnormal physical quantity.

6. An electronic device, comprising: The system comprises a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory are in communication with each other through the communication bus; wherein: The memory is used to store a computer program; The processor is used to execute the method steps of any one of claims 1 to 4 by running the program stored in the memory.

7. A storage medium, characterized by The storage medium includes a stored program, wherein the program executes the method steps of any one of claims 1 to 4 when running. The storage medium includes a stored program, wherein the program executes the method steps of any one of claims 1 to 4 when running.

Images