Train sliding plug door fault diagnosis method and system based on acoustic signals

Abstract

The invention provides a train sliding plug door fault diagnosis method and system based on an acoustic signal, and belongs to the technical field of fault diagnosis, and the method specifically comprises the steps: carrying out the filtering processing of a preset type of interference signal in the acoustic signal, obtaining a processing signal, decomposing the processing signal, obtaining a plurality of dimensions of acoustic feature signals, and carrying out the processing of the acoustic feature signals; the method comprises the following steps: acquiring an acoustic characteristic signal of a sliding plug door, sending the acoustic characteristic signal to a prediction model of a Transform-based double-branch model framework, determining the abnormal probability of the sliding plug door under different fault types, and when the fault type of which the abnormal probability does not meet the requirement exists, acquiring a measurement result of a historical acoustic signal in an adjacent time period, according to the method, the consistency of the multiple groups of historical acoustic signals in the adjacent time period and the acoustic signals is determined, and whether an early warning signal needs to be output or not is determined by combining the historical acoustic signals, the abnormal probability of the fault type of which the abnormal probability does not meet the requirement under the acoustic signals and the measurement reliability, so that the accuracy of fault diagnosis processing is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault diagnosis, and in particular relates to a train plug door fault diagnosis method and system based on acoustic signals. Background Art

[0002] High-speed trains have the characteristics of high operating density and dense passenger flow. The frequent opening and closing of the train's sliding doors will increase the friction between the various components of the sliding door system, which is very easy to damage the door components. At the same time, the failure of the train's sliding doors may seriously threaten the safety of passengers and cause train delays. Usually, the failure will not only delay the current train, but also affect the normal operation of subsequent trains, thus causing a large backlog of delays.

[0003] In order to realize fault detection of plug doors, specifically in the invention patent application CN202410143231.1 "Metro plug door fault detection method based on Gram angle field and deep learning", the real-time detection of the upper and lower gear pin displacement, pressure wheel overpressure, V-type inversion fault and V-type forward size exceeding standard fault in the subway door system is realized by utilizing a deep learning network, which can realize automatic real-time detection of faults to ensure driving safety. However, this detection method has the problem that electrical signals are difficult to collect, and the electrical system of the train plug doors may be interfered with during the collection process.

[0004] In the paper "A Sound-Based Fault Diagnosis Method for Railway Point Machines Based on Two-Stage Feature Selection Strategy and Ensemble Classifier", acoustic signals are used to detect plug door faults. However, a fixed omnidirectional single microphone is used. During the actual train operation, there are many sound interference signals. The omnidirectional single microphone has major defects in sound pickup and is also unable to accurately diagnose the plug door faults of the train.

[0005] In response to the above technical problems, the present application specifically provides a train plug door fault diagnosis method and system based on acoustic signals. Summary of the invention

[0006] To achieve the purpose of the present invention, the present invention adopts the following technical solutions: According to one aspect of the present invention, a train plug door fault diagnosis method based on acoustic signals is provided.

[0007] A train plug door fault diagnosis method based on acoustic signals, specifically comprising: S1 uses a microphone array to pick up the acoustic signal of the plug door of the train, and filters the interference signal of a preset type in the acoustic signal to obtain a processed signal and the interference time in the processed signal; S2: when the distribution data of the interference moment in the acoustic signal meets the requirements, the measurement reliability of the processed signal is determined to meet the requirements by using the matching data of the motor current data of the plug sliding door and the processed signal, and then proceed to the next step; S3 decomposes the processed signal to obtain acoustic feature signals of multiple dimensions, and sends the acoustic feature signals to a prediction model based on a dual-branch model framework of Transformer to determine the abnormal probability of the plug door under different fault types. When there is a fault type whose abnormal probability does not meet the requirements, obtain the measurement results of the historical acoustic signals in the adjacent time period; S4 determines the consistency between multiple groups of historical acoustic signals in the adjacent time period and the acoustic signal, and determines whether it is necessary to output a warning signal in combination with the historical acoustic signals and the abnormal probability of the fault type whose abnormal probability under the acoustic signal does not meet the requirements and the measurement reliability.

[0008] The beneficial effects of the present invention are: Whether the measurement reliability of the processed signal meets the requirements is determined by the matching data of the motor current data and the processed signal of the Sierra door, thereby further realizing the matching of the motor current data and the processed signal, realizing the identification of abnormal conditions of the processed signal, avoiding the occurrence of technical problems such as a high degree of abnormality of the processed signal due to interference from external sound signals, and improving the reliability of fault identification and processing of the Sierra door.

[0009] Based on the consistency between multiple groups of historical acoustic signals and acoustic signals in the adjacent time period, the historical acoustic signals and the abnormal probability of the fault type whose abnormal probability under the acoustic signals does not meet the requirements, and the measurement reliability, it is determined whether it is necessary to output a warning signal. This not only takes into account the difference in the reliability of the current acoustic signal caused by the difference in the consistency between the historical acoustic signals and the acoustic signals, but also further combines the historical acoustic signals and the abnormal probability under the acoustic signals and the measurement reliability, thereby realizing the evaluation of the abnormal probability of the fault type from multiple dimensions and improving the reliability of fault identification processing.

[0010] A further technical solution is that the preset types of interference signals include footsteps, rolling sounds of suitcase wheels and human voices.

[0011] A further technical solution is that the similarities in different dimensions include similarities in different time domain features and similarities in frequency domain features.

[0012] A further technical solution is that the method for determining the interference moment in the processing signal is: Acquiring the signal amplitude of the processed signal at different moments, and determining the amplitude abnormality coefficient at different moments based on the signal amplitude at different moments; Determine similarity coefficients between signal characteristics of processed signals and different preset types of interference signals at different moments, and use similarity coefficients with signal characteristics of interference signals of different preset types to determine similar types at different moments, and determine signal interference coefficients at different moments based on the proportion of the number of interference signals of similar types; The interference anomaly coefficient at different moments is determined by taking the average value of the signal interference coefficient and the amplitude anomaly coefficient at different moments, and the interference anomaly coefficient is used to determine whether the moment is an interference moment.

[0013] A further technical solution is that the amplitude anomaly coefficient is determined based on the ratio of the signal amplitude to a preset amplitude.

[0014] A further technical solution is that the acoustic characteristic signal includes a frequency domain characteristic signal and a time domain characteristic signal.

[0015] A further technical solution is that the fault types include door control circuit failure, door leaf damage, loose wires or short circuit, control cabinet equipment failure, door sensor failure, loose door leaf bolts, poor vehicle grounding, and control cabinet equipment failure.

[0016] In a second aspect, the present invention provides a computer system comprising: a memory and a processor that are communicatively connected, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-mentioned train plug door fault diagnosis method based on acoustic signals when running the computer program.

[0017] Other features and advantages will be described in the following description, and partly become apparent from the description, or understood by practicing the invention. The purpose and other advantages of the invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0018] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0020] Figure 1 It is a flow chart of a train plug door fault diagnosis method based on acoustic signals; Figure 2 is a flow chart of a method for determining the moment of interference in a processing signal; Figure 3 It is a flow chart for determining that the distribution data of the interference moment in the acoustic signal meets the requirements; Figure 4 It is a framework diagram of a computer system. DETAILED DESCRIPTION

[0021] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the present invention will be comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar structures, and thus their detailed description will be omitted.

[0022] The terms "a", "an", "the", and "said" are used to indicate the presence of one or more elements / components / etc.; the terms "comprising" and "having" are used to express an open-ended inclusive meaning and mean that additional elements / components / etc. may be present in addition to the listed elements / components / etc.

[0023] Over the years, the failure of train plug doors has caused train delays and suspensions, reduced operational efficiency, and even threatened the personal safety of passengers, so they need to be monitored and maintained. On the other hand, due to its complex structure and high frequency of use, the failure modes of train plug doors are also diverse, which puts high demands on the status monitoring and maintenance of plug doors. In view of the current situation of high-speed railway development in my country, it is urgent to invent a convenient and flexible train plug door status monitoring and fault diagnosis method to achieve accurate identification of plug door status, provide support for the maintenance and repair of train plug doors, and ensure the safe and good operation order of high-speed railways.

[0024] At present, the invention of the train plug door system in China is still in its infancy. The main means of plug door fault diagnosis is still through the traditional manual detection method, which is inefficient and requires high experience and operation level of maintenance personnel. In the fault diagnosis method based on signal processing, most of the invention teams chose to use motor current signals as analysis data. This detection method has the problem that electrical signals are not easy to collect, and the electrical system of the train plug door may be disturbed during the collection process. In contrast to the electrical signal, the acoustic signal adopts a non-contact signal collection method, which will not interfere with the equipment. The acoustic signal collection is convenient and the collection method is flexible. The acoustic signal generated by the interaction of various components of the train plug door during the action process is closely related to its working state, which can intuitively reflect the status information of the system. At the same time, audio signal processing technology has been rapidly developing and maturing for many years, providing powerful technical means and theoretical support for acoustic signal analysis. Based on the above considerations, the present invention proposes a high-speed railway train plug door fault diagnosis method based on acoustic signals, aiming to solve the problems of high failure rate, diversified failures, and difficulty in comprehensive analysis and accurate positioning caused by the complex structure and frequent action of the plug door.

[0025] High-speed railway plug door acoustic fault diagnosis is a complex model classification task, which can be divided into two core stages: model training and classification detection.

[0026] The first is the model training stage: first, acoustic signals covering various common faults are collected, and then these signals are preprocessed, such as format conversion, normalized sampling rate, pre-emphasis, frame segmentation and window function processing, and accurately labeled as high-quality training samples. Next, representative acoustic features are extracted and input into the Transformer model for training. When the model training reaches the preset standard, the solidified model is ready for subsequent applications.

[0027] The second is the classification and detection stage: input the acoustic signal to be detected, perform preprocessing and feature extraction similar to the training stage, and then input these features into the solidified Transformer model. The system outputs the fault category based on the model classification results.

[0028] Example 1 To solve the above problems, according to one aspect of the present invention, Figure 1 As shown, according to one aspect of the present invention, a train plug door fault diagnosis method based on acoustic signals is provided, which specifically includes: S1 uses a microphone array to pick up the acoustic signal of the plug door of the train, and filters the interference signal of a preset type in the acoustic signal to obtain a processed signal and the interference time in the processed signal; Furthermore, the preset types of interference signals include footsteps, rolling sounds of suitcase wheels, and human voices.

[0029] Specifically, the similarities in different dimensions include similarities in different time domain features and similarities in frequency domain features.

[0030] Specifically, Figure 2 As shown, the method for determining the interference moment in the processing signal is: Acquiring the signal amplitude of the processed signal at different moments, and determining the amplitude abnormality coefficient at different moments based on the signal amplitude at different moments; Determine similarity coefficients between signal characteristics of processed signals and different preset types of interference signals at different moments, and use similarity coefficients with signal characteristics of interference signals of different preset types to determine similar types at different moments, and determine signal interference coefficients at different moments based on the proportion of the number of interference signals of similar types; The interference anomaly coefficient at different moments is determined by taking the average value of the signal interference coefficient and the amplitude anomaly coefficient at different moments, and the interference anomaly coefficient is used to determine whether the moment is an interference moment.

[0031] Furthermore, the amplitude anomaly coefficient is determined based on the ratio of the signal amplitude to a preset amplitude.

[0032] Alternatively, the similarity type is a signal feature of an interference signal of a preset type whose similarity coefficient is greater than a preset similarity coefficient.

[0033] In another embodiment, the method for determining the interference moment in the processing signal is: Acquire the signal amplitude of the processed signal at the moment, and based on the signal amplitude at the moment, when the signal amplitude at the moment is less than a preset signal amplitude threshold, determine that the moment is an interference moment; When the signal amplitude at the moment is not less than the preset signal amplitude threshold: Determine a similarity coefficient between the signal characteristics of the processed signal at the moment and different preset types of interference signals, and when there is a signal characteristic of the preset type of interference signal with a similarity coefficient greater than a preset similarity threshold at the moment, determine the moment as an interference moment; When there is no signal feature of a preset type of interference signal whose similarity coefficient is greater than a preset similarity threshold at the moment: When it is determined that no similar type exists in the moment by using a similarity coefficient of a signal feature of an interference signal of a different preset type, it is determined that the moment is not an interference moment; When there is a similar type at the moment in question: Determine the similarity type of the moment by using the similarity coefficient of the signal characteristics of the interference signals of different preset types, and when the number of the similarity types is greater than the number of preset types, determine that the moment is an interference moment; When the number of similar types is not greater than the preset number of types: Determine the signal interference coefficient at the moment based on the proportion of the number of similar types of interference signals and the similarity coefficients of different similar types, and when the signal interference coefficient at the moment does not meet the requirement, determine that the moment is an interference moment; When the signal interference coefficient at the time meets the requirement: Based on the signal amplitude at the moment, the amplitude abnormality coefficient at the moment is determined, the interference abnormality coefficient at the moment is determined by the average value of the signal interference coefficient at the moment and the amplitude abnormality coefficient, and the interference abnormality coefficient is used to determine whether the moment is an interference moment.

[0034] It should be noted that the signal characteristics include amplitude characteristics, phase characteristics and frequency characteristics; S2: when the distribution data of the interference moment in the acoustic signal meets the requirements, the measurement reliability of the processed signal is determined to meet the requirements by using the matching data of the motor current data of the plug sliding door and the processed signal, and then proceed to the next step; Specifically, Figure 3 As shown, determining that the distribution data of the interference moment in the acoustic signal meets the requirements specifically includes: Determine the number of interference moments in the acoustic signal based on the distribution data of interference moments in the acoustic signal, and determine the interference proportional factor of the acoustic signal based on the proportion of the number of interference moments; Determine interference moments whose interval duration is less than a preset duration using interval data of different adjacent interference moments, and determine the distribution clustering coefficient of the acoustic signal according to the proportion of the number of interference moments whose interval duration is less than the preset duration; The signal anomaly coefficient of the acoustic signal is determined by the interference proportional factor and the distribution clustering coefficient of the acoustic signal, and the signal anomaly coefficient is used to determine whether the distribution data of the interference moment in the acoustic signal meets the requirements.

[0035] Furthermore, the signal anomaly coefficient of the acoustic signal is determined according to the product of the interference proportional factor and the distribution clustering coefficient of the acoustic signal.

[0036] It can be understood that when the signal anomaly coefficient of the acoustic signal is greater than a preset anomaly coefficient threshold, it is determined that the distribution data of the interference moment in the acoustic signal does not meet the requirements.

[0037] Specifically, when the distribution data of the interference moment in the acoustic signal does not meet the requirements, it is determined that the abnormal probability of the Sierra door cannot be determined.

[0038] It should be noted that determining whether the measurement reliability of the acoustic signal meets the requirements specifically includes: Determine the signal characteristic matching coefficients of the motor current data and the processing signal at different times based on the matching data of the motor current data of the plug sliding door and the processing signal; The signal feature matching coefficient is used to determine the matching deviation moment of the processed signal, and the number of the matching deviation moments is used to determine whether the measurement reliability of the acoustic signal meets the requirements.

[0039] Furthermore, the signal characteristic matching coefficient between the motor current data and the processed signal is determined based on the matching between the motor current characteristics and the amplitude characteristics of the processed signal at different times. Specifically, the preset signal amplitude characteristics corresponding to the motor current characteristics are determined based on the motor current characteristics, and are determined based on the similarity coefficient between the preset signal amplitude characteristics and the amplitude characteristics of the processed signal.

[0040] It should be noted that the matching deviation moment is the moment when the signal feature matching coefficient is less than the preset matching coefficient.

[0041] Specifically, when the measurement reliability of the acoustic signal does not meet the requirements, it is determined that the abnormal probability of the plug door cannot be determined.

[0042] In another embodiment, determining whether the measurement reliability of the acoustic signal meets the requirement specifically includes: The signal characteristic matching coefficients of the motor current data and the processed signal at different times are determined by the matching data of the motor current data of the plug sliding door and the processed signal. When the signal characteristic matching coefficients are used to determine that there is no matching deviation, it is determined that the measurement reliability of the acoustic signal meets the requirements. When the signal feature matching coefficient is used to determine the time when there is a matching deviation: Acquiring the number of matching deviation moments in the processed signal, and when the number of matching deviation moments in the processed signal does not meet the requirement, determining that the measurement reliability of the acoustic signal does not meet the requirement; When the number of matching deviation moments in the processed signal meets the requirement: Obtaining an average value of the signal feature matching coefficients at different times in the processed signal, and when the average value of the signal feature matching coefficients at different times is less than a preset feature coefficient threshold, determining that the measurement reliability of the acoustic signal does not meet the requirement; When the average value of the signal feature matching coefficients at different times is not less than the preset feature coefficient threshold: Determining the number of matching deviation moments in different divided time periods in the processed signal based on the distribution data between different matching deviation moments, and when there is no divided time period in which the number of matching deviation moments does not meet the requirement, determining that the measurement reliability of the acoustic signal meets the requirement; When there is a partition period where the number of matching deviation moments does not meet the requirement: Taking the divided time periods in which the number of matching deviation moments does not meet the requirement as the distribution aggregation time periods, and when the number of the distribution aggregation time periods does not meet the requirement, determining that the measurement reliability of the acoustic signal does not meet the requirement; When the number of distribution aggregation periods meets the requirement: The deviation coefficients of different divided time periods are determined according to the proportion of the number of matching deviation moments in different divided time periods, the measurement reliability coefficient of the acoustic signal is determined based on the average value of the deviation coefficients of different divided time periods, and the measurement reliability coefficient is used to determine whether the measurement reliability of the acoustic signal meets the requirements.

[0043] S3 decomposes the processed signal to obtain acoustic feature signals of multiple dimensions, and sends the acoustic feature signals to a prediction model based on a dual-branch model framework of Transformer to determine the abnormal probability of the plug door under different fault types. When there is a fault type whose abnormal probability does not meet the requirements, obtain the measurement results of the historical acoustic signals in the adjacent time period; In order to fully retain the characteristic information of acoustic signals in time domain and frequency domain and achieve deep fusion thereof, the present invention proposes to design a dual-branch Transformer model for fusion and classification of time domain and frequency domain features of acoustic signals. This model is named as the dual-branch Transformer model for multi-domain feature fusion of acoustic signals.

[0044] The dual-branch Transformer model for time-domain and frequency-domain feature fusion and classification consists of a parallel time-domain Transformer feature extraction module and a frequency-domain Transformer feature extraction module. Each Transformer feature extraction module consists of three submodules: a linear mapping layer, a Transformer encoder unit, and a multi-layer perceptron. The Transformer encoder unit contains a multi-head attention layer, a normalization layer, and a multi-layer perceptron. Transformer encoder unit.

[0045] The dual-stream Transformer model for time-domain and frequency-domain feature fusion and classification takes the time-domain feature representation and the frequency-domain feature representation as input, extracts the time-domain feature information from the time-domain feature extraction module, and extracts the frequency-domain feature information from the frequency-domain feature extraction module. Finally, the extracted time-domain feature information and frequency-domain feature information are fused in the fusion and classification layer for high-precision classification. The fusion and classification layer consists of a cascade layer and a fully connected layer with Softmax activation. The present invention intends to use multi-classification cross entropy as the loss function, which is defined as follows: The present invention proposes an adversarial transfer learning model for acoustic fault diagnosis of plug doors of high-speed trains based on the Domain-Adversarial Neural Network (DANN) technical framework. Its core purpose is to ensure that the acoustic fault diagnosis of plug doors of high-speed trains can continue to maintain a high degree of accuracy in various changing practical application environments. The transfer learning model for acoustic fault diagnosis of plug doors of high-speed trains mainly focuses on constructing feature representations with strong generalization capabilities, so as to align acoustic fault data in different environments in a shared feature space. The architecture of the adversarial transfer learning model consists of three key components: feature extractor, acoustic fault classifier and domain classifier.

[0046] Furthermore, the acoustic characteristic signal includes a frequency domain characteristic signal and a time domain characteristic signal.

[0047] Specifically, the fault types include door control circuit failure, door leaf damage, loose wires or short circuit, control cabinet equipment failure, door sensor failure, loose door leaf bolts, poor vehicle grounding, and control cabinet equipment failure.

[0048] S4 determines the consistency between multiple groups of historical acoustic signals in the adjacent time period and the acoustic signal, and determines whether it is necessary to output a warning signal in combination with the historical acoustic signals and the abnormal probability of the fault type whose abnormal probability under the acoustic signal does not meet the requirements and the measurement reliability.

[0049] Furthermore, the consistency between the historical acoustic signal and the acoustic signal is determined based on a signal similarity coefficient of an acoustic characteristic signal between the historical acoustic signal and the acoustic signal at different moments.

[0050] It is understandable that determining whether to output a warning signal specifically includes: Determine the signal similarity coefficients of the acoustic characteristic signals of the historical acoustic signal and the acoustic signal at different times based on the consistency between different historical acoustic signals and the acoustic signal, and determine the acoustic signal similarity coefficients of the different historical acoustic signals and the acoustic signal based on the average value of the signal similarity coefficients; The fault type whose abnormal probability does not meet the requirements is taken as the target fault type, and the historical acoustic signal whose abnormal probability of the target fault type is greater than the preset abnormal probability threshold is taken as the screening acoustic signal; The corrected reliability coefficients of different screened acoustic signals are determined by multiplying the measurement reliability of the screened acoustic signal and the acoustic signal similarity coefficient. The comprehensive reliability coefficient is determined according to the sum of the corrected reliability coefficients of different screened acoustic signals. The comprehensive reliability coefficient is used to determine whether it is necessary to output a warning signal.

[0051] It should be noted that when the average abnormal probability of the fault type whose abnormal probability does not meet the requirement is greater than a preset probability threshold, it is determined that a warning signal needs to be output.

[0052] Further, it is determined whether it is necessary to output a warning signal, including: S41 determines the signal similarity coefficients of the acoustic characteristic signals of the historical acoustic signals and the acoustic signal at different times based on the consistency between different historical acoustic signals and the acoustic signal, and determines the acoustic signal similarity coefficients of the different historical acoustic signals and the acoustic signal based on the average value of the signal similarity coefficients; S42 determines the corrected abnormal probability of the fault type whose abnormal probability does not meet the requirement under different historical acoustic signals according to the product of the abnormal probability of the fault type whose abnormal probability does not meet the requirement under different historical acoustic signals, the measurement reliability and the acoustic signal similarity coefficient; S43 determines an abnormal probability factor of the fault type whose abnormal probability does not meet the requirement by correcting the abnormal probability of the fault type under different historical acoustic signals, and uses the abnormal probability factor to determine whether it is necessary to output an early warning signal.

[0053] Optionally, the above step S41 includes the following contents: S411 determines the signal similarity coefficients of the acoustic characteristic signals of the historical acoustic signals and the acoustic signal at different times based on the consistency between different historical acoustic signals and the acoustic signal, and determines the acoustic signal similarity coefficients of the different historical acoustic signals and the acoustic signal based on the average value of the signal similarity coefficients. When there is a historical acoustic signal whose acoustic signal similarity coefficient meets the requirements, the process proceeds to step S412. When there is no historical acoustic signal whose acoustic signal similarity coefficient meets the requirements, it is determined that there is no need to output a warning signal. S412: taking historical acoustic signals whose acoustic signal similarity coefficients meet the requirements as similar acoustic signals; when the number of similar acoustic signals meets the requirements, proceeding to step S42; when the number of similar acoustic signals does not meet the requirements, proceeding to step S413; S413 determines a comprehensive similarity coefficient based on the acoustic signal similarity coefficients of different similar acoustic signals and the number of similar acoustic signals. When the comprehensive similarity coefficient meets the requirements, the process proceeds to step S42. When the comprehensive similarity coefficient does not meet the requirements, it is determined that there is no need to issue a warning signal.

[0054] Optionally, the above step S42 includes the following contents: S421 takes the fault type whose abnormal probability does not meet the requirement as the target fault type. When the abnormal probability of the target fault type under different historical acoustic signals is less than the preset abnormal probability threshold, it is determined that there is no need to issue a warning signal. When there is a historical acoustic signal whose abnormal probability of the target fault type is not less than the preset abnormal probability threshold, the process proceeds to step S422. S422: taking the historical acoustic signals whose abnormal probability of the target fault type is not less than the preset abnormal probability threshold as abnormal acoustic signals; when the number of the abnormal acoustic signals is greater than the preset number of acoustic signals, proceeding to step S43; when the number of the abnormal acoustic signals is not greater than the number of historical acoustic signals of the preset abnormal probability threshold, proceeding to step S423; S423: determining the corrected reliability coefficients under different abnormal acoustic signals by multiplying the measurement reliability of different abnormal acoustic signals by the product of the acoustic signal similarity coefficients; when there is no abnormal acoustic signal whose corrected reliability coefficient is greater than the preset reliability coefficient, determining that it is not necessary to issue an early warning signal; when there is an abnormal acoustic signal whose corrected reliability coefficient is greater than the preset reliability coefficient, proceeding to step S424; S424 determines the corrected abnormal probability of the target fault type under different historical acoustic signals according to the product of the abnormal probability of the target fault type under different historical acoustic signals, the measurement reliability and the acoustic signal similarity coefficient. When the number of historical acoustic signals whose corrected abnormal probability of the target fault type does not meet the requirement is greater than the preset number of acoustic signals, it is determined that a warning signal needs to be issued. When the number of historical acoustic signals whose corrected abnormal probability of the target fault type does not meet the requirement is not greater than the preset number of acoustic signals, the process proceeds to step S43.

[0055] Example 2 Second, as Figure 4 As shown, the present invention provides a computer system, comprising: a memory and a processor that are communicatively connected, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-mentioned train plug door fault diagnosis method based on acoustic signals when running the computer program.

[0056] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0057] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0058] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. A train plug door fault diagnosis method based on acoustic signals, characterized in that: Specifically include: Using a microphone array to pick up the acoustic signal of the plug door of the train, filtering a preset type of interference signal in the acoustic signal to obtain a processed signal and the interference moment in the processed signal; When the distribution data of the interference moment in the acoustic signal meets the requirements, and when the measurement reliability of the processed signal is determined to meet the requirements by using the matching data of the motor current data of the plug sliding door and the processed signal, proceed to the next step; Decomposing the processed signal to obtain acoustic characteristic signals of multiple dimensions, and sending the acoustic characteristic signals to a prediction model based on a dual-branch model framework of Transformer to determine the abnormal probability of the plug door under different fault types, and when there is a fault type whose abnormal probability does not meet the requirements, obtaining the measurement results of the historical acoustic signals in the adjacent time period; Determine the consistency of multiple groups of historical acoustic signals in the adjacent time period with the acoustic signal, and determine whether it is necessary to output a warning signal based on the historical acoustic signals and the abnormal probability of the fault type whose abnormal probability under the acoustic signal does not meet the requirements and the measurement reliability.

2. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The preset types of interference signals include footsteps, rolling sounds of luggage wheels, and human voices.

3. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The similarities in different dimensions include similarities in different time domain features and similarities in frequency domain features.

4. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The method for determining the interference moment in the processing signal is: Acquiring the signal amplitude of the processed signal at different moments, and determining the amplitude abnormality coefficient at different moments based on the signal amplitude at different moments; Determine similarity coefficients between signal characteristics of processed signals and different preset types of interference signals at different moments, and use similarity coefficients with signal characteristics of interference signals of different preset types to determine similar types at different moments, and determine signal interference coefficients at different moments based on the proportion of the number of interference signals of similar types; The interference anomaly coefficient at different moments is determined by taking the average value of the signal interference coefficient and the amplitude anomaly coefficient at different moments, and the interference anomaly coefficient is used to determine whether the moment is an interference moment.

5. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 4, characterized in that: The amplitude anomaly coefficient is determined according to the ratio of the signal amplitude to the preset amplitude.

6. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 4, characterized in that: The similarity type is a signal feature of a preset type of interference signal whose similarity coefficient is greater than a preset similarity coefficient.

7. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The acoustic characteristic signal includes a frequency domain characteristic signal and a time domain characteristic signal.

8. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The fault types include door control circuit failure, door leaf damage, loose wires or short circuit, control cabinet equipment failure, door sensor failure, loose door leaf bolts, poor vehicle grounding, and control cabinet equipment failure.

9. The train plug door fault diagnosis method based on acoustic signals as claimed in claim 1, characterized in that: The consistency between the historical acoustic signal and the acoustic signal is determined based on signal similarity coefficients of acoustic characteristic signals of the historical acoustic signal and the acoustic signal at different moments.

10. A computer system comprising: A memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor executes a train plug door fault diagnosis method based on acoustic signals as described in any one of claims 1-9 when running the computer program.

Images