Elevator fault detection method, device and computer readable storage medium

Abstract

The present application discloses an elevator fault detection method, device and computer-readable storage medium, which belongs to the field of fault detection technology. The method comprises: collecting the original time series signal of the elevator to be detected, classifying and processing the original time series signal according to a preset period, obtaining at least one first classification subsequence and at least one second classification subsequence, wherein the preset period is determined according to the operation state of the elevator, performing preliminary processing on the first classification subsequence to obtain a first preliminary subsequence, and performing preliminary processing on the second classification subsequence to obtain a second preliminary subsequence, performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence through a pre-trained elevator fault detection model, and obtaining the fault detection result of the elevator. The present application can effectively improve the accuracy of fault detection, that is, efficiently detect elevator faults, thereby improving the safety of elevator operation.

Description

Technical Field

[0001] The present application relates to the technical field of fault detection, and in particular to an elevator fault detection method, device and computer-readable storage medium. Background Art

[0002] With the development of science and technology, elevators have gradually become an indispensable means of vertical transportation. With the continuous increase in the number of elevators and the continuous aging of elevators, the frequency of elevator failures is also increasing, which seriously affects the safety of elevators. In related technologies, elevators are usually maintained and inspected by regular manual inspection and maintenance. However, the manual inspection and maintenance of elevators will result in low accuracy of fault detection, which will leave a great safety hazard for the operation of elevators. Summary of the invention

[0003] The present application aims to solve at least one of the technical problems in the related art to a certain extent. To this end, the present application proposes an elevator fault detection method, device and computer-readable storage medium, which can efficiently detect elevator faults, thereby improving the safety of elevator operation.

[0004] In a first aspect, an embodiment of the present application provides an elevator fault detection method, the method comprising:

[0005] Collect the original time series signal of the elevator to be tested;

[0006] Classifying and processing the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence; wherein the preset period is determined according to the operating state of the elevator;

[0007] Performing preliminary processing on the first classified subsequence to obtain a first preliminary subsequence, and performing the preliminary processing on the second classified subsequence to obtain a second preliminary subsequence;

[0008] The first preliminary subsequence and the second preliminary subsequence are processed for fault detection by using a pre-trained elevator fault detection model to obtain a fault detection result of the elevator.

[0009] The technical solution of the first aspect of the present application has at least one of the following advantages or beneficial effects: first, the original time series signal of the elevator to be detected is collected, and then the original time series signal is classified and processed according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence, wherein the preset period is determined according to the operating state of the elevator to be detected, and then the first classification subsequence is preliminarily processed to obtain a first preliminary subsequence, and the second classification subsequence is preliminarily processed to obtain a second preliminary subsequence, and then, the first preliminary subsequence and the second preliminary subsequence are processed by a pre-trained elevator fault detection model to obtain the fault detection result of the elevator. After the present application classifies the original time series signal by a preset period to obtain the first classification subsequence and the second classification subsequence, the first preliminary subsequence and the second preliminary subsequence obtained after the preliminary processing are processed by the elevator fault detection model to obtain the fault detection result of the elevator to be detected, thereby effectively improving the accuracy of fault detection, that is, being able to efficiently detect elevator faults, thereby improving the safety of elevator operation.

[0010] In some embodiments of the present application, the collecting of the original time series signal of the elevator includes:

[0011] The original time series signal is collected by a non-invasive sensor; wherein the original time series signal includes at least one of the following: an elevator braking system current signal, a traction machine current signal, a door opening and closing control signal, a safety system signal, and an elevator relative height signal.

[0012] In some embodiments of the present application, the preset period includes a first period and a second period, the operating state includes a car motion state and a door opening and closing state, and the original time series signal is classified and processed according to the preset period to obtain at least one first classification subsequence and at least one second classification subsequence, including:

[0013] The original time series signal is cut according to the first period to obtain at least one of the first classification subsequences; wherein each of the first periods is the duration from the beginning to the end of each of the car motion states;

[0014] The original time series signal is cut according to the second period to obtain at least one second classification subsequence; wherein each of the second periods is the duration of the door opening and closing state from the start of each car movement state to the start of the next car movement state.

[0015] In some embodiments of the present application, the elevator fault detection model includes a first fault detection model and a second fault detection model; before performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence by the pre-trained elevator fault detection model to obtain the fault detection result of the elevator, the method further includes: training the elevator fault detection model, specifically including:

[0016] Acquire a preset reference training sequence set; wherein the reference training sequence set includes a first classification reference sequence and a second classification reference sequence;

[0017] Dividing the first classification reference sequence according to a preset ratio to obtain a first training set and a first test set;

[0018] Training a preset first neural network model according to the first training set to obtain a first preliminary training model;

[0019] Using the first test set to verify the first preliminary training model to obtain the first fault detection model;

[0020] Dividing the second classification reference sequence according to the preset ratio to obtain a second training set and a second test set;

[0021] Training a preset second neural network model according to the second training set to obtain a second preliminary training model;

[0022] The second preliminary training model is verified using the second test set to obtain the second fault detection model.

[0023] In some embodiments of the present application, the first neural network model includes a sequence network, a fully convolutional network and a classifier; the first preliminary training model is obtained by training the preset first neural network model according to the first training set, including:

[0024] Inputting the first training set into the sequence network to perform time correlation extraction processing to obtain a first training feature;

[0025] Inputting the first training set into the fully convolutional network for feature extraction to obtain second training features;

[0026] Performing feature splicing processing on the first training feature and the second training feature to obtain a target splicing feature;

[0027] Inputting the target splicing features into the classifier for fault classification processing to obtain a predicted fault classification result;

[0028] The first neural network model is trained according to the predicted fault classification result to obtain the first preliminary training model.

[0029] In some embodiments of the present application, the training of the first neural network model according to the predicted fault classification result to obtain the first preliminary training model includes:

[0030] The model parameters of the first neural network model are adjusted according to the predicted fault classification result to train the first neural network model to obtain the first preliminary training model.

[0031] In some embodiments of the present application, adjusting the model parameters of the first neural network model according to the predicted fault classification result includes:

[0032] Obtaining actual fault classification results of the elevator;

[0033] Calculating the loss function of the first neural network model according to the predicted fault classification result and the actual fault classification result to obtain a target loss value;

[0034] The target loss value is used as the back-propagation amount to adjust the model parameters of the first neural network model.

[0035] In a second aspect, an embodiment of the present application further provides an elevator fault detection device, the device comprising:

[0036] A data acquisition module, used for acquiring the original time series signal of the elevator;

[0037] A classification processing module, used for classifying and processing the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence; wherein the preset period is determined according to the running state of the elevator;

[0038] a data processing module, configured to perform preliminary processing on the first classified subsequence to obtain a first preliminary subsequence, and perform preliminary processing on the second classified subsequence to obtain a second preliminary subsequence;

[0039] A fault detection module is used to perform fault detection processing on the first preliminary subsequence and the second preliminary subsequence through a pre-trained elevator fault detection model to obtain a fault detection result of the elevator.

[0040] The technical solution of the second aspect of the present application has at least one of the following advantages or beneficial effects: first, the data acquisition module is used to collect the original time series signal of the elevator to be detected, and then the classification processing module classifies and processes the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence, wherein the preset period is determined according to the running state of the elevator to be detected, and then the data processing module performs preliminary processing on the first classification subsequence to obtain a first preliminary subsequence, and performs preliminary processing on the second classification subsequence to obtain a second preliminary subsequence, and finally, the fault detection module performs fault detection processing on the first preliminary subsequence and the second preliminary subsequence through a pre-trained elevator fault detection model to obtain the fault detection result of the elevator. In the present application, after the classification processing module classifies and processes the original time series signal according to a preset period to obtain the first classification subsequence and the second classification subsequence, the fault detection module performs fault detection processing on the first preliminary subsequence and the second preliminary subsequence obtained after preliminary processing through the elevator fault detection model to obtain the fault detection result of the elevator to be detected, thereby effectively improving the accuracy of fault detection, that is, being able to efficiently detect elevator faults, thereby improving the safety of elevator operation.

[0041] In the third aspect, an embodiment of the present application further provides an elevator fault detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any embodiment of the first aspect when executing the computer program.

[0042] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute a method as in any one of the embodiments of the first aspect.

[0043] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0045] Figure 1 is a first flow chart of an elevator fault detection method provided by some embodiments of the present application;

[0046] Figure 2 yes Figure 1 Flow chart of step S120;

[0047] Figure 3 is a second flow chart of an elevator fault detection method provided by some embodiments of the present application;

[0048] Figure 4 yes Figure 3 Flow chart of step S330;

[0049] Figure 5 is a third flow chart of an elevator fault detection method provided by some embodiments of the present application;

[0050] Figure 6 This is a block diagram of the module structure of an elevator fault detection device provided in some embodiments of the present application. DETAILED DESCRIPTION

[0051] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0052] In the description of this application, "several" means more than one, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0053] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0054] With the development of science and technology, elevators have gradually become an indispensable means of vertical transportation. With the continuous increase in the number of elevators and the continuous aging of elevators, the frequency of elevator failures is also increasing, which seriously affects the safety of elevators. In related technologies, elevators are usually maintained and inspected by regular manual inspection and maintenance. However, the manual inspection and maintenance of elevators will result in low accuracy of fault detection, which will leave a great safety hazard for the operation of elevators.

[0055] Based on this, the present application provides an elevator fault detection method, device and computer-readable storage medium. The method is as follows: first, the original time series signal of the elevator to be detected is collected, and then the original time series signal is classified and processed according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence, wherein the preset period is determined according to the operating state of the elevator to be detected, and then the first classification subsequence is preliminarily processed to obtain a first preliminary subsequence, and the second classification subsequence is preliminarily processed to obtain a second preliminary subsequence, and then, the first preliminary subsequence and the second preliminary subsequence are processed by a pre-trained elevator fault detection model to obtain the fault detection result of the elevator. After the present application classifies the original time series signal by a preset period to obtain the first classification subsequence and the second classification subsequence, the first preliminary subsequence and the second preliminary subsequence obtained after the preliminary processing are processed by the elevator fault detection model to obtain the fault detection result of the elevator to be detected, thereby effectively improving the accuracy of fault detection, that is, being able to efficiently detect elevator faults, thereby improving the safety of elevator operation.

[0056] The embodiments of the present application are further described below in conjunction with the accompanying drawings.

[0057] refer to Figure 1 Some embodiments of the present application provide an elevator fault detection method, which includes but is not limited to step S110, step S120, step S130, and step S140.

[0058] Step S110, collecting the original time series signal of the elevator to be detected.

[0059] In some embodiments, the original time series signal of the elevator to be detected is first collected to facilitate the subsequent acquisition of the fault detection result of the elevator.

[0060] In some embodiments, the original time series signal of the elevator can be collected by non-invasive sensors, thereby avoiding the elevator shutdown during the collection process and reducing the impact on the normal service of the elevator during the collection process. Among them, the collected original time series signals include the elevator brake system current signal, the traction machine current signal, the door opening and closing control signal, the safety system signal, and the elevator relative height signal, but are not limited to this. Among them, the elevator brake system current signal is the brake coil current measured by the current sensor, which is used to reflect the working state of the brake; the traction machine current signal is the elevator AC asynchronous motor single-phase current signal, which is used to reflect the working state of the elevator traction machine; the door opening and closing control signal is the stator current signal of the door motor. By detecting the stator current of the door motor, the number of times the elevator door is opened and other information can be observed; the safety system signal is the series chain current signal of the safety relay, power supply, etc., which is used to avoid elevator failure. The disconnection of the safety connection will cause the elevator system to be immediately interrupted. As an important indicator of the elevator operation status, the safety system signal can potentially reflect the elevator operation status; the elevator relative height signal is the distance signal from the top of the elevator to the ground collected by the laser sensor. In practical applications, the sampling frequency of the original time series signal can be set to 20Hz. It should be noted that the sampling frequency can be selected according to actual needs and should not be construed as a limitation to the present application.

[0061] Step S120: classify the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence.

[0062] In some embodiments, the collected original time series signals are classified and processed according to a preset period, wherein the preset period is determined according to the running state of the elevator. The corresponding preset period is determined according to different running states of the elevator, and then the original time series signals are segmented and processed respectively according to each preset period, so as to classify the original time series signals, and better extract the fault sequence features in the original time series signals.

[0063] Step S130 , performing preliminary processing on the first classified subsequence to obtain a first preliminary subsequence, and performing preliminary processing on the second classified subsequence to obtain a second preliminary subsequence.

[0064] In some embodiments, the first classification subsequence obtained by the classification process is preliminarily processed to obtain a first preliminary subsequence, and the second classification subsequence obtained by the classification process is preliminarily processed to obtain a second preliminary subsequence. This embodiment preliminarily processes the first classification subsequence and the second classification subsequence to facilitate subsequent acquisition of input data that conforms to the elevator fault detection model to obtain a fault detection result.

[0065] Specifically, in some embodiments, the preliminary processing includes length consistency of data, filtering and denoising of data, label calibration, and calculation of speed signal based on elevator relative height signal and deletion of abnormal data. In this embodiment, for ease of expression, the subsequence is set to include a first preliminary subsequence and a second preliminary subsequence. Among them, length consistency includes setting the maximum length of the subsequence to 2000, and the subsequence exceeding the set maximum length is regarded as an abnormal sequence, which is deleted from the corresponding subsequence set, that is, the abnormal data is deleted. For the subsequence whose length is less than the set maximum length, zero is padded at the end of the subsequence to reach the set maximum length; Kalman filtering can be used to filter and denoise the subsequence, so that the sequence is smoother, which is conducive to the feature extraction of elevator faults; the speed signal of the elevator can be calculated based on the measured relative height signal of the elevator and the sampling frequency; the relationship between the sequence and the classification category can be found through the rule-base method, and the subsequence is calibrated.

[0066] Step S140, performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence through a pre-trained elevator fault detection model to obtain an elevator fault detection result.

[0067] In some embodiments, a pre-trained elevator fault detection model is used to perform fault detection processing on the first preliminary subsequence and the second preliminary subsequence obtained after preliminary processing to obtain an elevator fault detection result, that is, the first preliminary subsequence and the second preliminary subsequence are input into the trained elevator fault detection model to obtain the output elevator fault detection result, and thus the fault detection of the elevator to be detected is completed.

[0068] In some embodiments, Figure 2 As shown, Figure 2 It is a detailed step flow chart of step S120, where step S120 includes but is not limited to step S210 and step S220.

[0069] Step S210, cutting the original time series signal according to the first period to obtain at least one first classification subsequence;

[0070] Step S220: cutting the original time series signal according to the second period to obtain at least one second classification subsequence.

[0071] Specifically, the preset period includes a first period and a second period, the running state of the elevator includes a car motion state and a door opening and closing state, the car motion state corresponds to the first period, each first period is specifically the duration of each car motion state from the beginning to the end, the door opening and closing state corresponds to the second period, each second period is specifically the duration of the door opening and closing state from the beginning of each car motion state to the beginning of the next car motion state. The original time series signal is cut according to the first period to obtain at least one first classification subsequence, and the original time series signal is cut according to the second period to obtain at least one second classification subsequence. Therefore, by cutting the original time series signal according to the preset period corresponding to the motion state, the fault sequence characteristics of the elevator to be detected can be better extracted, and the accuracy of elevator fault detection can be improved.

[0072] Furthermore, in some embodiments, the elevator to be detected will move from the 1st floor to the 8th floor, that is, from the 1st floor to the 8th floor, and will stop once on the 2nd floor and the 5th floor respectively, accompanied by the door opening and closing action, and finally stop on the 8th floor and remain still after the door opening and closing. In this process, the elevator starts to move from the 1st floor to the 2nd floor and stops moving, and before reaching the 2nd floor without opening the door, this process is the elevator car movement, and the movement state of the elevator during this period is the car movement state, and the duration of the car movement state is a first cycle. Similarly, the elevator starts to move from the 2nd floor to the 5th floor and stops moving, and before reaching the 5th floor without opening the door is the elevator car movement, and the duration of the car movement is also a first cycle. Therefore, by analogy, it is easy to clearly understand that there are three first cycles in the entire process of the elevator to be detected moving from the 1st floor to the 8th floor. In addition, when the elevator starts moving from the 1st floor to the time when the next elevator starts moving, that is, the elevator starts from the car moving state on the 1st floor to the time when the car moves on the 2nd floor, that is, when the elevator stops moving from the 2nd floor to the time when the door opening and closing action is performed and starts moving, the door opening and closing process during this period is the door opening and closing movement of the elevator, and the movement state of the elevator during the door opening and closing movement is the door opening and closing state, and the duration of the door opening and closing state is a second cycle, and so on, there is a second cycle for the elevator from the car moving state on the 2nd floor to the car moving state on the 5th floor, and for the elevator from the car moving state on the 5th floor to the car moving state on the 8th floor. Therefore, it is easy to clearly understand that there are three second cycles in the entire process of the elevator to be tested moving from the 1st floor to the 8th floor. It is understandable that the above description, such as "when the elevator starts moving from the 1st floor to the next elevator movement, that is, the elevator starts moving from the 1st floor car state to the 2nd floor car state, that is, when the elevator stops moving from the 2nd floor to the door opening and closing action before starting to move" is just a different statement to express the meaning of the second period more clearly and without doubt, but it is essentially the same. Afterwards, the original time series signal of the elevator moving from the 1st floor to the 8th floor is cut according to the first period to obtain three first classification subsequences, and the original time series signal is cut according to the second period to obtain three second classification subsequences.

[0073] In some embodiments, Figure 3 As shown, before the first preliminary subsequence and the second preliminary subsequence are subjected to fault detection processing by a pre-trained elevator fault detection model to obtain an elevator fault detection result, an elevator fault detection method provided in the present application further includes but is not limited to the step of "training an elevator fault detection model", which step includes but is not limited to step S310, step S320, step S330, step S340, step S350, step S360, and step S370.

[0074] Step S310, obtaining a preset reference training sequence set;

[0075] Step S320, dividing the first classification reference sequence according to a preset ratio to obtain a first training set and a first test set;

[0076] Step S330, training a preset first neural network model according to a first training set to obtain a first preliminary training model;

[0077] Step S340, verifying the first preliminary training model using the first test set to obtain a first fault detection model;

[0078] Step S350, dividing the second classification reference sequence according to a preset ratio to obtain a second training set and a second test set;

[0079] Step S360, training a preset second neural network model according to a second training set to obtain a second preliminary training model;

[0080] Step S370: Use the second test set to verify the second preliminary training model to obtain a second fault detection model.

[0081] In some embodiments, an elevator fault detection model is trained by first obtaining a preset reference training sequence set, which is composed of a large amount of time series data obtained after preliminary processing, and the reference training sequence set includes a first classification reference sequence and a second classification reference sequence. Then, the first classification reference sequence is divided according to a preset ratio to obtain a first training set and a first test set, and the second classification reference sequence is divided according to a preset ratio to obtain a second training set and a second test set. Then, the preset first neural network model is trained according to the preset first training set to obtain a first preliminary training model, and the first preliminary training model is verified by the first test set to obtain a first fault detection model. Similarly, the preset second neural network model is trained according to the preset second training set to obtain a second preliminary training model, and the second preliminary training model is verified by the second test set to obtain a second fault detection model. The first fault detection model and the second fault detection model are used to detect corresponding elevator faults.

[0082] Further, in some embodiments, the first classification reference sequence is obtained by classifying and processing the collected original time series data according to the first period corresponding to the elevator car motion state and performing preliminary processing, and the second classification reference sequence is obtained by classifying and processing the collected original time series data according to the second period corresponding to the elevator door opening and closing state and performing preliminary processing. The first classification reference sequence contains a large number of subsequences corresponding to the car motion state, and the second classification reference sequence contains a large number of subsequences corresponding to the door opening and closing state. For example, the first classification reference sequence and the second classification reference sequence both contain 33514 periodic subsequences and corresponding category labels. The first classification reference sequence and the second classification reference sequence are multidimensional time series data sets, which are formed by arranging the statistics of multiple behavioral features of the elevator in a set interval time period, such as every hour, in chronological order and integrating them into a multidimensional time series data set within a complete time period, such as two months. Afterwards, 50% of the first classification reference sequence is randomly used as the first training set, and the remaining 50% is used as the first test set; 50% of the second classification reference sequence is randomly used as the second training set, and the remaining 50% is used as the second test set. The first training set is used to train the preset first neural network model to obtain a first preliminary training model, and the first test set is used to verify the first preliminary training model to obtain a first fault detection model. The first fault detection model obtained in this embodiment is used to detect abnormalities in car movement. Accordingly, the second training set is used to train the preset second neural network model to obtain a second preliminary training model, and the first test set is used to verify the first preliminary training model to obtain the first fault detection model. The second fault detection model obtained in this embodiment is used to detect abnormalities in elevator door opening and closing.

[0083] In some embodiments, Figure 4 As shown, Figure 4 It is a detailed step flow chart of step S330, and step S330 includes but is not limited to step S410, step S420, step S430, step S440, and step S450.

[0084] Step S410, inputting the first training set into the sequence network to perform time correlation extraction processing to obtain a first training feature;

[0085] Step S420, inputting the first training set into a fully convolutional network for feature extraction to obtain second training features;

[0086] Step S430, performing feature splicing processing on the first training feature and the second training feature to obtain a target splicing feature;

[0087] Step S440, inputting the target splicing features into a classifier for fault classification processing to obtain a predicted fault classification result;

[0088] Step S450: training the first neural network model according to the predicted fault classification result to obtain a first preliminary training model.

[0089] In some embodiments, the first neural network model includes a sequence network, a full convolution network and a classifier. The preset first neural network model is trained according to the first training set to obtain a first preliminary training model. First, the first training set is input into the value sequence network for time correlation extraction processing to obtain a first training feature, and the first training feature includes the time correlation feature of the first training set. At the same time, the first training set is input into the full convolution network for feature extraction processing to obtain a second training feature, and the second feature includes elevator fault features other than time correlation. Afterwards, the first training feature and the second training feature are spliced ​​to obtain a target splicing feature, and the target splicing feature is input into the classifier for fault classification processing to obtain a predicted fault classification result. Finally, the first neural network model is trained according to the predicted fault classification result to obtain a first preliminary training model.

[0090] Further, in some embodiments, the first neural network model is a multidimensional LSTM-FCN deep neural network model that introduces an attention mechanism, LSTM-FCN is a long short-term memory full convolutional network, and the first neural network model specifically includes a sequence network, a full convolutional network, and a classifier. The sequence network is built on a multidimensional LSTM unit that introduces an attention mechanism, and is used to mine the time correlation features in the multidimensional time series, that is, the time correlation features in the first training set. Among them, LSTM is a long short-term memory network, and the long short-term memory network is a special RNN (recurrent neural network), which is mainly used to solve the gradient disappearance and gradient explosion problems in the long sequence training process. In short, compared with ordinary RNN, LSTM can have better performance in longer sequences. The full convolutional network is composed of three time convolution blocks and a global pooling layer stacked together, as a feature extractor for a multidimensional time series, that is, a feature extractor for the first training set, wherein the number of filters of the convolution block, that is, the number of convolution kernels, can be set to 128-256-128 in sequence, and the convolution kernel size kernel size can be set to 8-5-3 in sequence. The input shape of the model is set to (N, T, M), where N is the number of samples, T is the maximum subsequence length set, and M is the input sequence dimension. In addition, the classifier can use a Softmax classifier, which is evaluated by using a loss function and adjusts the correctness of the classification. In this embodiment, the input shape (16757, 2000, 5) is used as the input, and the input shape is input into the sequence network and the full convolution network respectively, and the output of the sequence network and the full convolution network are connected to obtain the target splicing features, which are input into the Softmax classifier. The number of neurons of the Softmax classifier is set to the number of classification categories, and the number of classification categories can be set to 2 (0-normal sequence, 1-abnormal sequence). The output result is used as the detection result of the first neural network model for the fault category, that is, the predicted fault classification result. Finally, the first neural network model is trained according to the predicted fault classification result to obtain the first preliminary training model.

[0091] It can be understood that the process of step S360 "training the preset second neural network model according to the second training set to obtain a second preliminary training model" is substantially the same as the specific training process of the above-mentioned steps S410, S420, S430, S440, and S450, and will not be repeated here.

[0092] In some embodiments, step S450 also includes but is not limited to the step of "adjusting the model parameters of the first neural network model according to the predicted fault classification result to train the first neural network model to obtain a first preliminary training model."

[0093] Specifically, the model parameters of the first neural network model can be adjusted according to the predicted fault classification result, the training of the first neural network model can be completed, and the first preliminary training model can be obtained to gradually improve the prediction accuracy of the first preliminary training model.

[0094] In some embodiments, Figure 5 As shown, Figure 5 It is a detailed step flow chart of the step of "adjusting the model parameters of the first neural network model according to the predicted fault classification results to train the first neural network model and obtain a first preliminary training model", which includes but is not limited to step S510, step S520, and step S530.

[0095] Step S510, obtaining the actual fault classification result of the elevator;

[0096] Step S520, calculating the loss function of the first neural network model according to the predicted fault classification result and the actual fault classification result to obtain a target loss value;

[0097] Step S530, using the target loss value as the back propagation amount to adjust the model parameters of the first neural network model.

[0098] Specifically, the actual fault classification result of the elevator is obtained according to the category label corresponding to the first training set, and then the loss function of the first neural network model is calculated according to the predicted fault classification result and the actual fault classification result to obtain the target loss value, and the target loss value is used as the back propagation amount to adjust the model parameters of the first neural network model. It should be noted that the gradient descent algorithm can be used to update the model parameters, and through step-by-step iterative solution, the minimized loss function and model parameter value can be obtained, that is, the minimum target loss value and model parameter can be obtained, thereby improving the prediction accuracy of the first preliminary training model.

[0099] In some embodiments, the first preliminary training model is verified using the first test set, and the first preliminary training model that is verified to meet the preset requirements is saved to obtain a first fault detection model. Among them, the results obtained by selecting one verification are displayed for reference only. For example, the results of car motion fault detection are shown in Table 1 and Table 2. Table 1 is the confusion matrix of elevator car motion fault classification, and Table 2 is the elevator car motion fault classification result:

[0100] 0 - Normal 1- Abnormal 0 - Normal 16702 0 1- Abnormal 0 55

[0101] Table 1

[0102] category Accuracy Recall Balanced F-score Number of samples 0 - Normal 1 1 1 16702 1- Abnormal 1 1 1 55

[0103] Table 2

[0104] In the above table, each column of the confusion matrix represents the predicted fault classification result, and each row represents the actual fault classification result. The accuracy of category 0-normal represents the proportion of truly normal samples among the samples identified as normal by the model. The higher the accuracy, the better the model effect; the recall of category 0-normal represents the ratio of the number of samples correctly identified as normal by the model to the total number of normal samples among the truly normal samples. The higher the recall, the better the model effect; the balanced F score of category 0-normal represents the harmonic mean of the corresponding accuracy and recall. The higher the balanced F score, the better the model effect; the number of samples represents the number of samples corresponding to the actual fault classification result. It can be understood that the meanings of the accuracy, recall, and balanced F score corresponding to category 1-abnormal are similar to those of the accuracy, recall, and balanced F score corresponding to category 0-normal. For example, the accuracy of category 1-abnormal represents the proportion of truly abnormal samples among the samples identified as abnormal by the model. The higher the accuracy, the better the model effect, so it will not be repeated here. In addition, it can also be understood that the minimum value of precision, recall, and balanced F-score is 0, and the maximum value is 1.

[0105] In some embodiments, the second preliminary training model is verified using the second test set, and the second preliminary training model that is verified to meet the preset requirements is saved to obtain a second fault detection model. Among them, the results obtained by selecting one verification are displayed for reference only. For example, the results of elevator switch door fault detection are shown in Table 3 and Table 4. Table 3 is the confusion matrix of elevator switch door fault classification, and Table 4 is the elevator switch door fault classification result.

[0106] 0 - Normal 1- Abnormal 0 - Normal 15685 76 1- Abnormal 46 950

[0107] Table 3

[0108] category Accuracy Recall Balanced F-score Number of samples 0 - Normal 0.997 0.995 0.996 15761 1- Abnormal 0.926 0.954 0.940 996

[0109] Table 4

[0110] In the above table, each column of the confusion matrix represents the predicted fault classification result, and each row represents the actual fault classification result. The meanings of accuracy, recall, balanced F score, and number of samples have been described above and will not be repeated here.

[0111] refer to Figure 6Some embodiments of the present application also provide an elevator fault detection device, which includes a data acquisition module 610, a classification processing module 620, a data processing module 630, and a fault detection module 640. The data acquisition module 610 acquires the original time series signal of the elevator to be detected, and then the classification processing module 620 classifies and processes the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence, wherein the preset period is determined according to the running state of the elevator to be detected, and then the data processing module 630 performs preliminary processing on the first classification subsequence to obtain a first preliminary subsequence, and performs preliminary processing on the second classification subsequence to obtain a second preliminary subsequence, and finally, the fault detection module 640 performs fault detection processing on the first preliminary subsequence and the second preliminary subsequence through a pre-trained elevator fault detection model to obtain the fault detection result of the elevator. In the present application, after the classification processing module 620 classifies the original time series signal according to a preset period to obtain a first classification subsequence and a second classification subsequence, the fault detection module 640 performs fault detection processing on the first preliminary subsequence and the second preliminary subsequence obtained after the preliminary processing through the elevator fault detection model to obtain the fault detection result of the elevator to be detected, thereby effectively improving the accuracy of fault detection, that is, being able to efficiently detect elevator faults, thereby improving the safety of elevator operation.

[0112] Some embodiments of the present application also provide an elevator fault detection device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method in any of the above embodiments is implemented, for example, the method described above is executed. Figure 1 Steps S110 to S140 of the method, Figure 2 Steps S210 to S220 of the method, Figure 3 Steps S310 to S370 of the method, Figure 4 Steps S410 to S450 of the method, Figure 5 Method steps S510 to S530 in.

[0113] In addition, one embodiment of the present application further provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method in any of the above embodiments, for example, to execute the above described Figure 1 Steps S110 to S140 of the method, Figure 2 Steps S210 to S230 of the method, Figure 3 Steps S310 to S330 of the method, Figure 4 Steps S410 to S430 of the method, Figure 5 Method steps S510 to S540 in.

[0114] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0115] The embodiments of the present application are described in detail above in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the purpose of the present application. In addition, the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

Claims

1. An elevator fault detection method, characterized in that: include: Collect the original time series signal of the elevator to be tested; Classifying and processing the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence; wherein the preset period is determined according to the operating state of the elevator; Performing preliminary processing on the first classified subsequence to obtain a first preliminary subsequence, and performing the preliminary processing on the second classified subsequence to obtain a second preliminary subsequence; Performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence by using a pre-trained elevator fault detection model to obtain a fault detection result of the elevator; The preset period includes a first period and a second period, the running state includes a car motion state and a door opening and closing state, and the original time series signal is classified and processed according to the preset period to obtain at least one first classification subsequence and at least one second classification subsequence, including: The original time series signal is cut according to the first period to obtain at least one of the first classification subsequences; wherein each of the first periods is the duration from the beginning to the end of each of the car motion states; The original time series signal is cut according to the second period to obtain at least one second classification subsequence; wherein each second period is the duration of the door opening and closing state from the start of each car movement state to the start of the next car movement state; The elevator fault detection model includes a first fault detection model and a second fault detection model; before performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence by the pre-trained elevator fault detection model to obtain the fault detection result of the elevator, the method further includes: training the elevator fault detection model, specifically including: Acquire a preset reference training sequence set; wherein the reference training sequence set includes a first classification reference sequence and a second classification reference sequence; Dividing the first classification reference sequence according to a preset ratio to obtain a first training set and a first test set; Training a preset first neural network model according to the first training set to obtain a first preliminary training model; Using the first test set to verify the first preliminary training model to obtain the first fault detection model; Dividing the second classification reference sequence according to the preset ratio to obtain a second training set and a second test set; Training a preset second neural network model according to the second training set to obtain a second preliminary training model; Using the second test set to verify the second preliminary training model to obtain the second fault detection model; The first neural network model includes a sequence network, a fully convolutional network and a classifier, and the first neural network model is trained according to the first training set to obtain a first preliminary training model, including: Inputting the first training set into the sequence network to perform time correlation extraction processing to obtain a first training feature; Inputting the first training set into the fully convolutional network for feature extraction to obtain second training features; Performing feature splicing processing on the first training feature and the second training feature to obtain a target splicing feature; Inputting the target splicing features into the classifier for fault classification processing to obtain a predicted fault classification result; The first neural network model is trained according to the predicted fault classification result to obtain the first preliminary training model.

2. The method according to claim 1, characterized in that: The collecting of the original time series signal of the elevator comprises: The original time series signal is collected by a non-invasive sensor; wherein the original time series signal includes at least one of the following: an elevator braking system current signal, a traction machine current signal, a door opening and closing control signal, a safety system signal, and an elevator relative height signal.

3. The method according to claim 1, characterized in that The step of training the first neural network model according to the predicted fault classification result to obtain the first preliminary training model includes: The model parameters of the first neural network model are adjusted according to the predicted fault classification result to train the first neural network model to obtain the first preliminary training model.

4. The method according to claim 3, characterized in that: The adjusting the model parameters of the first neural network model according to the predicted fault classification result includes: Obtaining actual fault classification results of the elevator; Calculating the loss function of the first neural network model according to the predicted fault classification result and the actual fault classification result to obtain a target loss value; The target loss value is used as the back-propagation amount to adjust the model parameters of the first neural network model.

5. An elevator fault detection device, characterized in that: include: A data acquisition module, used for acquiring the original time series signal of the elevator; A classification processing module, used for classifying and processing the original time series signal according to a preset period to obtain at least one first classification subsequence and at least one second classification subsequence; wherein the preset period is determined according to the running state of the elevator; a data processing module, configured to perform preliminary processing on the first classified subsequence to obtain a first preliminary subsequence, and perform preliminary processing on the second classified subsequence to obtain a second preliminary subsequence; a fault detection module, configured to perform fault detection processing on the first preliminary subsequence and the second preliminary subsequence by using a pre-trained elevator fault detection model to obtain a fault detection result of the elevator; The preset period includes a first period and a second period, the running state includes a car motion state and a door opening and closing state, and the original time series signal is classified and processed according to the preset period to obtain at least one first classification subsequence and at least one second classification subsequence, including: The original time series signal is cut according to the first period to obtain at least one of the first classification subsequences; wherein each of the first periods is the duration from the beginning to the end of each of the car motion states; The original time series signal is cut according to the second period to obtain at least one second classification subsequence; wherein each second period is the duration of the door opening and closing state from the start of each car movement state to the start of the next car movement state; The elevator fault detection model includes a first fault detection model and a second fault detection model; before performing fault detection processing on the first preliminary subsequence and the second preliminary subsequence by the pre-trained elevator fault detection model to obtain the fault detection result of the elevator, it also includes: training the elevator fault detection model, specifically including: Acquire a preset reference training sequence set; wherein the reference training sequence set includes a first classification reference sequence and a second classification reference sequence; Dividing the first classification reference sequence according to a preset ratio to obtain a first training set and a first test set; Training a preset first neural network model according to the first training set to obtain a first preliminary training model; Using the first test set to verify the first preliminary training model to obtain the first fault detection model; Dividing the second classification reference sequence according to the preset ratio to obtain a second training set and a second test set; Training a preset second neural network model according to the second training set to obtain a second preliminary training model; Using the second test set to verify the second preliminary training model to obtain the second fault detection model; The first neural network model includes a sequence network, a fully convolutional network and a classifier, and the first neural network model is trained according to the first training set to obtain a first preliminary training model, including: Inputting the first training set into the sequence network to perform time correlation extraction processing to obtain a first training feature; Inputting the first training set into the fully convolutional network for feature extraction to obtain second training features; Performing feature splicing processing on the first training feature and the second training feature to obtain a target splicing feature; Inputting the target splicing features into the classifier for fault classification processing to obtain a predicted fault classification result; The first neural network model is trained according to the predicted fault classification result to obtain the first preliminary training model.

6. An elevator fault detection device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method according to any one of claims 1 to 4.

Images