Railway vehicle underframe equipment mechanical fault monitoring system and method

Abstract

The application discloses a rail vehicle chassis equipment mechanical fault monitoring system and method, a microphone array is arranged on a vehicle frame, various noises in a fault monitoring area are collected in real time when the vehicle runs, abnormal noises are extracted through a signal processing means, and positioning of the abnormal sound sources is realized by using an array signal processing technology, so that the fault equipment of the vehicle bogie (vehicle walking component) is quickly determined and positioned, and thus the abnormal conditions of various equipment can be effectively monitored.

Description

Technical Field

[0001] This invention relates to the field of rail transit fault monitoring technology, and in particular to a mechanical fault monitoring system and method for rail vehicle underframe equipment. Background Technology

[0002] Currently, the railway industry uses acoustic-based train bearing fault monitoring equipment installed at the trackside. However, this method can only monitor trains passing the monitoring point for a short period, failing to achieve real-time online monitoring, and requires trains to travel at low speeds. Furthermore, to reduce false alarm rates, the equipment only triggers an alarm after detecting 3-5 consecutive faults in the corresponding bearing of the same train, resulting in a high false negative rate. Additionally, it relies on core technology imported from the United States, leading to high deployment costs, all of which hinder its widespread application. Addressing the shortcomings of the linear array-based TADS system, Anhui University proposed a train bearing monitoring scheme based on a surface array. This also involves trackside monitoring equipment that incorporates spatial information about the pitch angle (A High-Speed ​​Train Bearing Fault Diagnosis Method Based on Uniform Microphone Surface Array Filtering, 201810727759.8). Existing online monitoring technologies primarily use accelerometers placed on train bearings or motor bearings to collect and analyze vibration signals at corresponding measuring points, thereby achieving real-time monitoring of faults in the corresponding bearing components. In reality, the running gear of a train is a complex system, and the number of measurement points for vibration analysis is limited. The above methods cannot effectively monitor other abnormalities such as equipment malfunctions, loose installations, and cracks in the bogie structure. Summary of the Invention

[0003] This invention provides a mechanical fault monitoring system and method for rail vehicle underframe equipment, which can at least solve one of the problems mentioned in the background art.

[0004] A mechanical fault monitoring system for rail vehicle underframe equipment, comprising:

[0005] Several microphone array modules are set in different fault monitoring areas to collect abnormal sounds emitted by the vehicle; the microphone array module includes several microphone units installed on the vehicle body to collect sound signals.

[0006] The signal processing module extracts abnormal noise from the sound signals acquired by the microphone array module and locates the source of the abnormal noise, thereby discovering the fault and confirming the fault location.

[0007] A method for monitoring mechanical faults in the underframe equipment of a rail vehicle includes the following steps:

[0008] Step 1: The sound signal acquired by the microphone array module is processed through a time-domain filter bank w i(n) The sound signals of each fault monitoring area are obtained by filtering. The sound signals of the target area of ​​the fault monitoring area are picked up by scanning and the sound signals of other non-fault monitoring areas are suppressed to achieve spatial filtering and sound source localization.

[0009] Step 2: Calculate the power spectrum of the sound signal when the wheelset passes over the non-track joint, treat the power spectrum as background noise, and continuously update iteratively;

[0010] Calculate the power spectrum of the current frame data, divide it by the background noise power spectrum, and obtain the power spectrum quotient;

[0011] Based on the characteristics of the signal power spectrum distribution, the average value of the power spectrum quotient of a certain sub-band is calculated to obtain the pulse index;

[0012] By combining the kurtosis and spectral centroid of the sound signal, abnormal signals can be identified;

[0013] In step one, the sound signal collected by the microphone array module is processed by a time-domain filter bank w i (n) The method for obtaining the sound signals of each fault monitoring area through filtering is as follows:

[0014] Acquire sound signal r i (n) and the received signal d from the microphone unit k (n), the impulse response from each fault monitoring area to the microphone unit is calculated, and the calculation formula used is:

[0015]

[0016] Where i corresponds to the sound signal emitted in the i-th region, n represents the sampling time, l represents the sampling point of the calculated impulse response, and d k (nl) represents the received signal at time nl;

[0017] The impulse response is then converted to the frequency response using a Fourier transform. The conversion formula is as follows:

[0018] G ik (jω)=FFT{g ik (l)};

[0019] By combining the frequency responses, the transfer function matrix from each fault monitoring region to the microphone unit is obtained:

[0020] G i (jω)=[G i1 (jω)...G ik (jω)], and

[0021] G i (jω)=[G1(jω)...Gi-1 (jω)G i+1 (jω)...G I (jω)];

[0022] According to G i (jω) and G i (jω) yields the filter banks corresponding to each fault monitoring region:

[0023]

[0024] Where PR{} is the operation to find the eigenvector corresponding to the largest eigenvalue of the matrix, W i (jω)=[w i1 (jω) w i2 (jω)... w iK [jω] is a K-times 1st order vector, corresponding to K microphone units;

[0025] Then, the response is converted to a time-domain impulse response using an inverse Fourier transform. The conversion formula is as follows:

[0026] w ik (n) = IFFT{w ik (jω)};

[0027] The signal x received by the microphone unit i (n) The received signals of each fault monitoring area are obtained by filtering with a filter bank:

[0028]

[0029] in, Represents the convolution operation;

[0030] The y values ​​of each fault monitoring area were calculated. i (n), thereby achieving the separation and localization of sound sources in each fault monitoring area and outside the fault monitoring area;

[0031] The calculation of the spectral pulse index in step two is as follows:

[0032] Extract the data x′ between two adjacent track joints of the wheelset. i (n), calculate its power spectrum P′ i (k):

[0033]

[0034] The background noise power spectrum P″ is obtained by iteratively updating the spectrum using the following formula. i Estimation of (k):

[0035] P″ i (k)=P″i (k)*α+P′ i (k)*(1-α);

[0036] Where α is the forgetting factor, a coefficient between 0 and 1;

[0037] For the monitoring data x′ of the current frame j (n), according to the above calculation method, the power spectrum estimate P′ is obtained. j (k), then calculate the power spectral quotient relative to the background noise:

[0038] D j (k)=P j (k) / P′ i (k);

[0039] Further calculation of the mean of its power spectral quotient within a certain bandwidth yields the spectral pulse index:

[0040]

[0041] This allows for the monitoring of abnormal noise and the location of its occurrence.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention is based on the microphone array module deployed on the train underframe to collect the sound information emitted by the train during operation. It collects various noises in the fault monitoring area in real time when the vehicle is running, extracts abnormal noises through signal processing, and uses array signal processing technology to locate the abnormal sound source. This enables the rapid identification and location of faulty equipment in the vehicle bogie (vehicle running parts), thereby effectively monitoring the abnormal conditions of various equipment. Attached Figure Description

[0043] Figure 1 This is a system architecture diagram of the present invention;

[0044] Figure 2 This is an image of the measured time-domain waveform of the on-board device of the present invention;

[0045] Figure 3 This is an image of the measured acoustic signal spectrum pulse index FP1 of the vehicle-mounted device of the present invention;

[0046] Figure 4 This is an image of the measured acoustic signal spectrum pulse index FP2 of the vehicle-mounted device of the present invention;

[0047] Figure 5 This is a flowchart of the monitoring method of the present invention. Detailed Implementation

[0048] The following detailed description of a specific embodiment of the present invention is provided in conjunction with the accompanying drawings. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiment.

[0049] like Figures 1 to 5 As shown in the figure, an embodiment of the present invention provides a mechanical fault monitoring system for rail vehicle underframe equipment, comprising:

[0050] Several microphone array modules are set in different fault monitoring areas (such as gearbox area, wheel area, etc.) to collect abnormal sounds emitted by the vehicle; the microphone array module includes several microphone units installed on the car body (specifically at the bogie at the bottom of the subway car) to collect sound signals.

[0051] The signal processing module extracts abnormal noise from the sound signal acquired by the microphone array module and locates the source of the abnormal noise, thereby discovering the fault and confirming the fault location.

[0052] The microphone array module is fixedly mounted on the vehicle body via a microphone bracket. This installation method is an existing structure and will not be described in detail here.

[0053] The microphone array module and the signal processing module transmit signals via cable or wireless communication. The signal processing module can be an embedded system, including ARM systems, DSP systems, etc. Figure 1 As shown, the signal processing module (processor) receives sound signals through the A / D conversion interface, stores abnormal sound signal data, and sends the abnormal monitoring information to the driver's cab or the vertical vehicle operation control center via network transmission. This communication method adopts the existing communication design, so it will not be described in detail here.

[0054] A monitoring method based on a mechanical fault monitoring system for rail vehicle underframe equipment includes the following steps:

[0055] Step 1: The sound signal acquired by the microphone array module is processed through a time-domain filter bank w i (n) The sound signals of each fault monitoring area are obtained by filtering. The sound signals of the target area of ​​the fault monitoring area are picked up by scanning and the sound signals of other non-fault monitoring areas are suppressed to achieve spatial filtering and sound source localization.

[0056] Step 2: Calculate the power spectrum of the sound signal when the wheelset passes over the non-track joint, treat the power spectrum as background noise, and continuously update iteratively;

[0057] Calculate the power spectrum of the current frame data, divide it by the background noise power spectrum, and obtain the power spectrum quotient;

[0058] Based on the characteristics of the signal power spectrum distribution, the average value of the power spectrum quotient of a certain sub-band is calculated to obtain the pulse index;

[0059] By combining kurtosis and spectral centroid, anomalous signals can be identified;

[0060] In step one, the time-domain filter w i (n) Pre-design based on the actual conditions of the train, the design steps include:

[0061] 1. Determine the area where fault diagnosis is needed based on the actual situation;

[0062] 2. The transfer function matrix G from the sound source to the microphone array was obtained by actual measurement. i (jω), i = 1...I, where I is the total number of regions;

[0063] 3. Design frequency domain received signal filtering matrices for different regions based on the measured frequency domain transfer function matrix;

[0064] 4. Based on the frequency received signal filtering matrix, the time-domain received signal filtering matrix w is obtained through Fast Fourier Transform. i (n);

[0065] In step one, the sound signal acquired by the microphone array module is processed by the time-domain filter bank w i (n) The method for obtaining the sound signals of each fault monitoring area through filtering is as follows:

[0066] Acquire sound signal r i (n) and the received signal d from the microphone unit k (n), the impulse response from each fault monitoring area to the microphone unit is calculated, and the calculation formula used is:

[0067]

[0068] Where i corresponds to the sound signal emitted in the i-th region, n represents the sampling time, l represents the sampling point of the calculated impulse response, and d k (nl) represents the received signal at time nl;

[0069] The impulse response is then converted to the frequency response using a Fourier transform. The conversion formula is as follows:

[0070] G ik (jω)=FFT{g ik (l)};

[0071] By combining the frequency responses, the transfer function matrix from each fault monitoring region to the microphone unit is obtained:

[0072] G i(jω)=[G i1 (jω)...G ik (jω)], and

[0073] G i (jω)=[G1(jω)...G i-1 (jω)G i+1 (jω)...G I (jω)];

[0074] According to G i (jω) and G i (jω) yields the filter banks corresponding to each fault monitoring region:

[0075]

[0076] Where PR{} is the operation to find the eigenvector corresponding to the largest eigenvalue of the matrix, W i (jω)=[w i1 (jω) w i2 (jω)... w iK [jω] is a K-times 1st order vector, corresponding to K microphone units;

[0077] Then, the response is converted to a time-domain impulse response using an inverse Fourier transform. The conversion formula is as follows:

[0078] w ik (n) = IFFT{w ik (jω)};

[0079] The signal x received by the microphone unit i (n) The received signals of each fault monitoring area are obtained by filtering with a filter bank:

[0080]

[0081] in, Represents the convolution operation;

[0082] The y values ​​of each fault monitoring area were calculated. i (n), thereby achieving the separation and localization of sound sources in each fault monitoring area and outside the fault monitoring area;

[0083] The calculation of the spectral pulse index in step two is as follows:

[0084] Extract the data x′ between two adjacent track joints of the wheelset. i (n), calculate its power spectrum P′ i (k):

[0085]

[0086] The background noise power spectrum P″ is obtained by iteratively updating the spectrum using the following formula. i Estimation of (k):

[0087] P″ i (k)=P″ i (k)*α+P′ i (k)*(1-α);

[0088] Where α is the forgetting factor, a coefficient between 0 and 1;

[0089] For the monitoring data x′ of the current frame j (n), according to the above calculation method, the power spectrum estimate P is obtained. j (k), then calculate the power spectral quotient relative to the background noise:

[0090] D j (k)=P j (k) / P′ i (k);

[0091] Further calculation of the mean power spectral quotient within a certain bandwidth reveals that the spectral pulse index is highly sensitive to impulse signals, which is beneficial for monitoring abnormal pulse signals submerged in background noise. Specifically, in this embodiment of the invention, different frequency bands are selected to obtain two spectral pulse index parameters:

[0092]

[0093]

[0094] This allows for the monitoring of abnormal noise and the pinpointing of the time and location of its occurrence.

[0095] The proposed spectral pulse index is beneficial for monitoring abnormal pulse signals under low signal-to-noise ratio conditions. Combined with parameters such as kurtosis and spectral centroid, it enables rapid classification of abnormal signals (abnormal noise signals).

[0096] Figure 2 The sound signal was collected on a subway traveling at 40 km / h. It contained three abnormal noises caused by loosening. Because the abnormal noises were very weak, they were not visible at all on the time domain diagram. Only two pulse sounds excited by the two wheelsets of the bogie passing through the track gaps could be seen. Figure 3 and Figure 4 These are the extracted spectral pulse indices FP1 and FP2, respectively. Figure 3 The signal displayed is the one showing the wheelset passing over the track joint. Figure 4 The display shows the signals of the wheelset passing over the track joint and the abnormal noise signals. Figure 4It can clearly pinpoint the time and location of abnormal noises, thereby enabling effective monitoring of faint abnormal noises and their locations. In other words, it can detect faults and their locations. The monitoring system transmits the abnormal monitoring information to the driver's cab or train operation control center via network transmission. The train operation control center or driver's cab will then make a judgment and carry out maintenance after the current operation or make the train stop urgently for maintenance.

[0097] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit and essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0098] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for monitoring mechanical faults in rail vehicle underframe equipment, characterized in that, The mechanical fault monitoring system for rail vehicle underframe equipment includes: Several microphone array modules are set in different fault monitoring areas to collect abnormal sounds emitted by the vehicle; the microphone array module includes several microphone units installed on the vehicle body to collect sound signals. The signal processing module extracts abnormal noise from the sound signal acquired by the microphone array module and locates the source of the abnormal noise, thereby discovering the fault and confirming the fault location. The monitoring method includes the following steps: Step 1: Pass the sound signal acquired by the microphone array module through a time-domain filter bank. The sound signals of each fault monitoring area are obtained by filtering. The sound signals of the target area of ​​the fault monitoring area are picked up by scanning, and the sound signals of other non-fault monitoring areas are suppressed to achieve spatial filtering and sound source localization. Step 2: Calculate the power spectrum of the sound signal when the wheelset passes over the non-track joint, treat the power spectrum as background noise, and continuously update iteratively; Calculate the power spectrum of the current frame data, divide it by the background noise power spectrum, and obtain the power spectrum quotient; Based on the characteristics of the signal power spectrum distribution, the average value of the power spectrum quotient of a certain sub-band is calculated to obtain the pulse index; By combining the kurtosis and spectral centroid of the sound signal, abnormal signals can be identified.

2. The method for monitoring mechanical faults in rail vehicle underframe equipment as described in claim 1, characterized in that, In step one, the sound signal collected by the microphone array module is processed by a time-domain filter bank. The method for filtering to obtain the sound signals of each fault monitoring area is as follows: Acquire sound signals r i (n) and the received signal of the microphone unit d k (n) The impulse response from each fault monitoring area to the microphone unit was calculated, and the calculation formula used was as follows: ； in i Corresponding to the i The sound signal from each area Represents the sampling time. This represents the sampling points of the calculated impulse response. Indicates the first The received signal at any given moment; The impulse response is then converted to the frequency response using a Fourier transform. The conversion formula is as follows: ； By combining the frequency responses, the transfer function matrix from each fault monitoring region to the microphone unit is obtained: ,and ； according to as well as The filter banks corresponding to each fault monitoring area are obtained: ； in, To find the eigenvector corresponding to the largest eigenvalue of a matrix, the following operations are performed. It is a K-times 1st order vector, corresponding to K microphone units; Then, the response is converted to a time-domain impulse response using an inverse Fourier transform. The conversion formula is as follows: ； Signal received by the microphone unit The received signals for each fault monitoring area are obtained by filtering with a filter bank: ； in, Represents the convolution operation; The calculations for each fault monitoring area are as follows This enables the separation and localization of sound sources in each fault monitoring area and outside the fault monitoring area.

3. The method for monitoring mechanical faults in rail vehicle underframe equipment as described in claim 2, characterized in that, The calculation of the spectral pulse index in step two is as follows: Extract data between two adjacent track joints of the wheelset. Calculate its power spectrum : ； The background noise power spectrum is obtained by iteratively updating the spectrum using the following formula. Estimate: ； in, It is the forgetting factor, a coefficient between 0 and 1; For the monitoring data of the current frame Based on the above calculation method, the power spectrum estimate is obtained. P j ( k Then calculate the power spectral quotient relative to the background noise: ； Further calculation of the mean of its power spectral quotient within a certain bandwidth yields the spectral pulse index: ； This allows for the monitoring of abnormal noise and the location of its occurrence.

Images