Operation fault monitoring method and system applied to automobile transmission system

Abstract

The present application relates to the technical field of transmission system fault diagnosis, and specifically to an operation fault monitoring method and system applied to automobile transmission systems, which specifically includes: determining the vibration characteristic coefficient of each period based on the periodic change characteristics of the vibration signal of each period when the automobile transmission system is running, the fluctuation of the vibration sequence, and the amplitude difference between different frequency regions of the vibration sequence in the frequency domain; constructing a fault trend coefficient based on the frequency of data fluctuations of the vibration signals of all periods and the change trend of the vibration characteristic coefficients of all periods to adjust the abnormality ratio, and based on the adjusted abnormality ratio, combining the isolation forest algorithm to monitor the operation fault of the automobile transmission system; improving the sensitivity of abnormality detection in the early stage of fault or minor fault, ensuring the dynamic balance between the false alarm rate and the missed alarm rate of transmission system fault detection, and improving the quality of automobile transmission system fault detection.

Description

Technical Field

[0001] The present application relates to the technical field of transmission system fault diagnosis, and in particular to an operation fault monitoring method and system applied to an automobile transmission system. Background Art

[0002] With the rapid development of the modern automotive industry and the increasing complexity and automation of vehicles, ensuring the stability and reliability of vehicle drivetrains has become crucial. As the core component connecting the engine and wheels, the drivetrain not only determines the vehicle's power transmission efficiency but also directly impacts the driving experience and safety. Faced with growing safety demands and higher expectations for vehicle performance, timely and accurate detection and prediction of potential drivetrain failures have become a critical task for manufacturers and service providers. With the advancement of intelligent monitoring technology, real-time online monitoring is driving fault monitoring towards intelligent and adaptive capabilities.

[0003] However, existing automotive transmission system fault monitoring solutions primarily focus on analyzing the instantaneous characteristics of vibration signals. After a fault occurs, the frequency domain characteristics of the instantaneous vibration signal are analyzed. However, this approach often only diagnoses the fault after a significant fault has already occurred, causing wear and tear on other components in the vehicle's transmission system and reducing its service life. In the case of early-stage or minor faults, this diagnostic approach, which focuses solely on transient characteristics, is difficult to identify. Summary of the Invention

[0004] In order to solve the above technical problems, the purpose of this application is to provide a method and system for monitoring operating faults in an automobile transmission system. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for monitoring operating faults in an automobile transmission system, the method comprising the following steps:

[0006] Collect all vibration signals in each cycle of the vehicle transmission system and construct the vibration sequence of each cycle;

[0007] Determine the vibration characteristic coefficient of each period based on the periodic variation characteristics of the elements in the vibration sequence of each period, the fluctuation of the vibration sequence, and the amplitude difference between different frequency regions of the vibration sequence in the frequency domain;

[0008] Construct a total vibration sequence based on the vibration sequences of all cycles; construct a vibration feature sequence based on the vibration characteristic coefficients of all cycles; construct a fault trend coefficient of the automobile transmission system based on the frequency of data fluctuations in the total vibration sequence and the data change trend in the vibration characteristic sequence;

[0009] determining an abnormality ratio adjustment value of the vehicle transmission system based on the fault trend coefficient;

[0010] Based on the abnormal proportion adjustment value, the automobile transmission system operation fault monitoring is performed in combination with the isolation forest algorithm.

[0011] In one embodiment, the process of obtaining the vibration characteristic coefficient of each period is as follows:

[0012] Obtain the autocorrelation coefficient of the vibration sequence of each period through the autocorrelation function, and construct the autocorrelation function graph of each period vibration sequence; construct the period characteristic value of the autocorrelation function graph of each period based on the peak value distribution characteristics in the autocorrelation function graph;

[0013] Calculate the variance of all elements in the vibration sequence of each period, recorded as the first variance;

[0014] Obtain a frequency domain waveform of each periodic vibration sequence using a time-frequency conversion algorithm. In each frequency domain waveform, divide all frequencies into a low-frequency region, a medium-frequency region, and a high-frequency region; calculate the average value of the variance of the amplitudes of all frequencies in the medium-frequency region and the high-frequency region, and record it as a first average value; record frequencies in the frequency domain waveform whose amplitudes are greater than a preset frequency significance threshold as significant frequencies, and record the sequence consisting of the amplitudes of all significant frequencies in the low-frequency region as a low-frequency amplitude sequence; and obtain a low-frequency distribution feature metric for the low-frequency region based on changes in elements in a first-order difference sequence of the low-frequency amplitude sequence;

[0015] A vibration characteristic coefficient of each period is constructed based on the period characteristic value, the first variance, and the difference between the first average value and the low-frequency distribution characteristic metric.

[0016] In one embodiment, the process of obtaining the periodic characteristic value is as follows:

[0017] The peaks in the autocorrelation function graph of each period that are greater than the preset correlation threshold are recorded as significant peaks;

[0018] The hysteresis corresponding to each significant peak value is used as the ordinate, and the order of appearance of each significant peak is used as the abscissa to obtain the coordinate point corresponding to each significant peak; the coordinate points corresponding to all significant peaks in the autocorrelation function graph are fitted using a linear fitting algorithm, and the obtained fitting straight line is recorded as the hysteresis fitting straight line; and the determination coefficient of the hysteresis fitting straight line is calculated;

[0019] The product of the mean value of all significant peak-to-peak values ​​in the autocorrelation function diagram of each period and the determination coefficient is used as the period characteristic value of the autocorrelation function diagram of each period.

[0020] In one embodiment, the low-frequency distribution feature metric is expressed as:

[0021] In the formula, lf i is the low-frequency distribution characteristic measure of the low-frequency area in the frequency domain waveform of the i-th cycle, μ(L i ),σ(L i ) are the mean and variance of all elements in the first-order difference sequence of the low-frequency amplitude sequence in the i-th period, and norm() is the normalization function.

[0022] In one embodiment, the vibration characteristic coefficient is expressed as:

[0023] Where A i is the vibration characteristic coefficient of the i-th period, R i is the periodic characteristic value of the i-th period, σ(V i ) is the first variance of the i-th period, t i is the lag corresponding to the first significant peak in the autocorrelation function diagram of the ith period, mf i is the first average value of the i-th cycle, lf i is the low-frequency distribution characteristic measure of the i-th period, and exp() is an exponential function with the natural constant e as the base.

[0024] In one embodiment, the process of obtaining the failure tendency coefficient of the automobile transmission system is as follows:

[0025] The total vibration sequence is used as the input of the polynomial fitting algorithm, and the output is the fitting curve of the total vibration sequence;

[0026] Using time series decomposition algorithm, the trend sequence of vibration characteristic sequence is obtained;

[0027] A fault trend coefficient of the automobile transmission system is constructed based on changes in the fitting curve and the fitting straight line of the trend sequence.

[0028] In one embodiment, the expression of the failure tendency coefficient is:

[0029] B=lg(X)×exp(-K), where B is the failure trend coefficient of the automobile transmission system, X is the number of data points whose derivative is 0 on the fitting curve, K is the slope of the fitting straight line of the trend sequence, lg() is a logarithmic function with 10 as the real number, and exp() is an exponential function with the natural constant e as the base.

[0030] In one embodiment, the expression of the abnormal ratio adjustment value of the automobile transmission system is:

[0031] Where W is the abnormal ratio adjustment value of the vehicle transmission system, B is the failure trend coefficient of the vehicle transmission system, and arctan() is the inverse tangent function.

[0032] In one embodiment, the automobile transmission system operation fault monitoring is performed based on the abnormal ratio adjustment value in combination with the isolation forest algorithm, specifically:

[0033] The vibration sequences of all cycles are used as input of the isolation forest algorithm, the abnormality ratio adjustment value is used as the abnormality ratio in the isolation forest algorithm, and the abnormality score of each vibration sequence of each cycle is output; and a segmentation threshold value of the abnormality score of the vibration sequences of all cycles is obtained;

[0034] The period in which the abnormality score of the vibration sequence is greater than the segmentation threshold is regarded as an abnormal period; the sequence consisting of all moments corresponding to all abnormal periods is obtained and recorded as an abnormal sequence, and the Hurst index of the first-order difference sequence of the abnormal sequence is calculated. If the Hurst index is greater than the preset fault threshold, then the current vehicle transmission system has an operating fault; otherwise, the current vehicle transmission system is operating normally.

[0035] In a second aspect, an embodiment of the present application also provides an operation fault monitoring system for an automobile transmission system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above methods when executing the computer program.

[0036] The embodiments of the present application have at least the following beneficial effects:

[0037] This application calculates the vibration characteristic coefficient based on the time-frequency domain characteristics of the instantaneous vibration data of each cycle when the automobile transmission system is running, which helps to improve the accuracy of identifying the difference between normal and abnormal states; then, the fault trend coefficient is calculated by combining the time-series variation trend of the vibration data and the vibration characteristic coefficient, which helps to improve the accuracy and sensitivity of subsequent fault diagnosis; finally, the abnormal proportion is adaptively adjusted according to the fault trend coefficient, balancing the accuracy and sensitivity of the isolated forest anomaly detection. In this way, the time-frequency domain characteristics and time-series variation trend of the vibration data are combined to improve the sensitivity of anomaly detection in the early stage of a fault or a minor fault, and it can respond flexibly according to the actual vibration situation, improve the adaptability and robustness of anomaly detection, ensure the dynamic balance between the false alarm rate and the missed alarm rate of the transmission system fault detection, and improve the quality of automobile transmission system fault detection, thereby realizing high-precision and high-sensitivity operation fault monitoring of the automobile transmission system, ensuring the safety of the automobile transmission system and extending the service life to a certain extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0039] Figure 1 A flowchart of a method for monitoring operating faults in an automobile transmission system according to an embodiment of the present application;

[0040] Figure 2 is a schematic diagram of a frequency domain waveform diagram under normal circumstances;

[0041] Figure 3 This is a schematic diagram of the frequency domain waveform under a minor fault;

[0042] Figure 4 This is a schematic diagram of the frequency domain waveform under severe fault conditions;

[0043] Figure 5 Schematic diagram of the autocorrelation function graph. DETAILED DESCRIPTION

[0044] To further illustrate the technical means and effectiveness of this application to achieve the intended invention objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of the method and system for monitoring operational faults in an automotive transmission system proposed in this application. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0045] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0046] The specific scheme of the operation fault monitoring method and system for automobile transmission system provided by the present application is described in detail below with reference to the accompanying drawings.

[0047] See also Figure 1 , which shows a flowchart of a method for monitoring operating faults in an automobile transmission system according to an embodiment of the present application, the method comprising the following steps:

[0048] Step S1 : collecting all vibration signals in each cycle when the automobile transmission system is running, and constructing a vibration sequence of each cycle.

[0049] In this embodiment, a vibration sensor is installed on the gearbox housing of the automobile transmission system to collect vibration signals when the transmission system is running. Specifically:

[0050] When collecting vibration signals, the signal collection frequency is set to 5200 Hz, and the vibration signal collection time for each time is set to 1 second, that is, the collection cycle is 1 second, and the sequence composed of all vibration signals collected within the collection cycle is used as the vibration sequence of each collection cycle.

[0051] It should be noted that this application only provides one setting method for the collection frequency and collection time each time the vibration signal is collected. The implementer can set the collection frequency and collection time each time according to the actual situation. This application does not impose any specific restrictions.

[0052] The vibration sequence of each cycle is used as the input of the wavelet denoising algorithm, and the collected vibration signal is denoised to obtain the denoised vibration sequence. The wavelet denoising algorithm is a well-known technology, and the specific process is not described in detail. It should be noted that for denoising vibration signals, this application only provides one denoising method. There are many existing denoising methods, and implementers can also use other denoising algorithms to denoise vibration signals. This application does not impose any specific restrictions.

[0053] Step S2, determining the vibration characteristic coefficient of each period based on the periodic variation characteristics of the elements in the vibration sequence of each period, the fluctuation of the vibration sequence, and the amplitude difference between different frequency regions of the vibration sequence in the frequency domain.

[0054] In the detection of operational faults in automotive transmission systems, the existing detection method is to determine whether an operational fault exists by analyzing the changes in vibration data during operation and performing time-frequency domain analysis on the collected data. However, this method focuses on the instantaneous changes in the vibration data, which means that only when a more significant fault occurs will the time-frequency domain characteristics of the vibration data show significant changes, thereby determining whether a fault exists. However, in the early stages of a fault or when the fault is relatively minor, the time-frequency domain changes in the vibration data are relatively minor, making it difficult to diagnose based solely on the instantaneous changes. By the time a diagnosis can be made, it indicates that a major fault has already occurred, which will adversely affect other components in the transmission system and shorten the service life of the transmission system. Therefore, it is necessary to improve the sensitivity of fault detection so that faults can be detected in a timely manner even in the early stages or when the fault is relatively minor, thereby increasing the service life of the automotive transmission system.

[0055] First, under normal circumstances, when the transmission system is fault-free, the overall vibration conditions are relatively similar due to the coordinated operation of the mechanical components, friction between the components, meshing, and other factors. The vibration signal shows a certain degree of repeatability and regularity, and the vibration data has small differences. At the same time, because normal vibration is caused by fast-cycling processes, such as gear rotation or shaft rotation, these processes have a short time period, so the overall periodic changes are relatively small. However, when a fault occurs in the transmission system, the vibration mode at the fault location will be different from the normal vibration, resulting in an increase in vibration intensity, which makes the vibration data abnormal and with large differences. At the same time, because the faulty component generates more irregular motion or impact loads, it will change the dynamic characteristics of the system, disturb the vibration period, and cause the vibration period to become longer. For example, due to the non-uniform motion caused by bearing wear or gear damage, the overall periodic changes are relatively large.

[0056] Secondly, under normal circumstances, in the frequency domain, due to the influence of the meshing frequency and harmonic components, a larger amplitude will appear in the low-frequency area, but the meshing frequency is still the main factor. On both sides of the meshing frequency, the energy of the higher-order harmonics decays step by step, that is, the amplitude change is closer to the parabolic distribution; and due to the existence of the fault, it will cause transient shocks, thereby causing modulation. In the early stage of the fault or in relatively minor cases, there is a slight stable modulation, and the amplitude of the sideband is generally high and concentrated. As the fault becomes more serious, it will evolve into an impact modulation, and the sideband will also be parabolic, but the medium and high frequency data will show more obvious fluctuations. Among them, the schematic diagram of the frequency domain waveform under normal circumstances is as follows: Figure 2 As shown, the schematic diagram of the frequency domain waveform under a minor fault is as follows Figure 3 As shown, the schematic diagram of the frequency domain waveform under severe fault is as follows Figure 4 shown.

[0057] (1) The vibration sequence of each period is used as the input of the autocorrelation function, and the output is a series of autocorrelation coefficients of the vibration sequence of each period. Each autocorrelation coefficient corresponds to a different hysteresis. In the embodiment of the present application, the autocorrelation coefficient of the hysteresis value range is [1,100], and the autocorrelation function diagram of the vibration sequence of the period is obtained based on the obtained autocorrelation coefficient. The schematic diagram of the autocorrelation function diagram is shown as follows: Figure 5 As shown. As other embodiments of the present application, the implementer can set the range of the lag amount in the autocorrelation function diagram according to actual conditions. In the autocorrelation function diagram, the peak with a peak value greater than the correlation threshold α is recorded as a significant peak. Preferably, in the embodiment of the present application, the value of α is set to 0.75. As other embodiments of the present application, the implementer can set the value of α according to actual conditions. Among them, the autocorrelation function is a well-known technology, and the specific process will not be repeated.

[0058] Furthermore, for each significant peak in the autocorrelation function graph, the hysteresis corresponding to the peak value of the significant peak is used as the ordinate, and the order of appearance of the significant peaks is used as the abscissa to obtain the coordinate points corresponding to the significant peaks. The coordinate points corresponding to all significant peaks in the autocorrelation function graph are linearly fitted using the least squares method, and the output fitting line is recorded as the hysteresis fitting line. The least squares method is a well-known technique, and the specific process is not repeated here.

[0059] (2) The vibration sequence of each cycle is used as the input of the Fourier transform algorithm, and the frequency domain waveform of the vibration sequence is output. Among them, Fourier transform is a well-known technology, and the specific process is not repeated here.

[0060] It should be noted that for obtaining the frequency domain waveform of the vibration sequence, this application only provides a time-frequency transformation method. There are many existing time-frequency transformation methods. Implementers can also use other time-frequency transformation algorithms to obtain the frequency domain waveform of the vibration sequence. This application does not make specific restrictions.

[0061] Calculate the mean of the amplitudes of all frequencies in the frequency domain waveform as the frequency significance threshold; record the frequencies with amplitudes greater than the frequency significance threshold as significant frequencies. Then divide the entire frequency region in the frequency domain waveform into 3 sections, obtaining 3 sub-intervals, which are recorded as low-frequency region, medium-frequency region and high-frequency region in order of frequency from small to large. Further, arrange the amplitudes of all significant frequencies in the low-frequency region in order of significant frequency from small to large to form a low-frequency amplitude sequence, and obtain the first-order difference sequence of the low-frequency amplitude sequence. Among them, the first-order difference sequence is a well-known technology, and the specific process will not be repeated here.

[0062] (3) Based on the above analysis, the vibration characteristic coefficient of each period is calculated to measure the time-frequency domain characteristics of the vibration signal of each period. Specifically:

[0063] First, to analyze the periodicity of the vibration sequence of each cycle, the product of the mean of all significant peak-to-peak values ​​in the autocorrelation function graph of each cycle and the determination coefficient of the hysteresis fitting line is calculated and recorded as the periodic characteristic value. The calculation of the determination coefficient is a well-known technique and the specific process is not repeated here.

[0064] Furthermore, the variance of all elements in the vibration sequence of each period is calculated and recorded as the first variance to evaluate the fluctuation of the vibration sequence of each period;

[0065] Furthermore, in order to analyze the amplitude difference between different frequency regions of the vibration signal in the frequency domain, the average value of the variance of the amplitudes of all frequencies in the intermediate frequency region and the high frequency region is calculated in the frequency domain waveform of the vibration sequence of each period, and recorded as the first average value; at the same time, the low-frequency distribution characteristics of the low-frequency region in the frequency domain waveform are analyzed based on the changes in the elements in the first-order difference sequence of the low-frequency amplitude sequence. The expression is: In the formula, lf i is the low-frequency distribution characteristic measure of the low-frequency area in the frequency domain waveform of the i-th cycle, μ(L i ),σ(L i ) are the mean and variance of all elements in the first-order difference sequence of the low-frequency amplitude sequence in the i-th period, and norm() is the normalization function.

[0066] Finally, the vibration characteristic coefficient of each period is calculated, and the expression is:

[0067]

[0068] Where A i is the vibration characteristic coefficient of the i-th period, R i is the periodic characteristic value of the i-th period, σ(V i ) is the first variance of the i-th period, t i is the lag corresponding to the first significant peak in the autocorrelation function diagram of the ith period, mf i is the first average value of the i-th cycle, lf i is a measure of the low-frequency distribution characteristics of the low-frequency area in the frequency domain waveform diagram of the ith period, and exp() is an exponential function with the natural constant e as the base.

[0069] Under normal circumstances, vibration data presents high periodic fluctuations with a small period, i.e. R i Larger, t i Small; at the same time, the fluctuation of the vibration sequence is small, that is, σ(V i ) is small; secondly, the significant frequency amplitude in the low-frequency region presents a parabolic distribution, with the largest value in the middle and gradually attenuating on both sides, so the mean μ(L i ) is small, and the variance σ(L i ) is larger, that is, lf i Small; at the same time, the amplitude fluctuation in the medium and high frequency areas is small, that is, mf i is smaller, so exp(mf i -lf i ) is small; thus, under normal circumstances, the vibration characteristic coefficient A i On the contrary, when there is a fault, the periodicity is weakened, the period is small, the vibration data fluctuation increases, and the mid-high frequency amplitude fluctuation increases, so the vibration characteristic coefficient A i The smaller the fault, the more serious the A i The smaller.

[0070] Step S3, constructing a total vibration sequence based on the vibration sequences of all cycles; constructing a vibration feature sequence based on the vibration feature coefficients of all cycles; and constructing a fault trend coefficient of the automobile transmission system based on the frequency of data fluctuations in the total vibration sequence and the data change trend in the vibration feature sequence.

[0071] The vibration characteristic coefficient determines the magnitude of the fault by analyzing the characteristics of the instantaneous vibration data. However, in the early stages of a fault or when the fault is relatively minor, the change in vibration data is not obvious, resulting in a smaller difference in the vibration characteristic coefficient. Therefore, it is difficult to monitor and diagnose the fault based solely on the vibration characteristic coefficient.

[0072] Under normal circumstances with no faults, the fluctuation of vibration data is small and there is a large regularity. At the same time, as analyzed above, the significant frequency amplitude fluctuation in the low-frequency area (sideband amplitude distribution) presents a parabolic distribution, and the medium and high frequency amplitudes are small and relatively stable. Therefore, the overall temporal changes of vibration data are relatively stable, the frequency of vibration fluctuations is small, and there is no obvious trend in the fluctuations.

[0073] In the early stages of a fault, vibration data will fluctuate slightly, increasing fluctuations. The distribution of sideband amplitudes will also change, and high-frequency noise will also be present, causing fluctuations in mid- and high-frequency amplitudes. These characteristics all cause relatively minor data fluctuations, making it difficult to distinguish them from normal data fluctuations based solely on the vibration changes within a single cycle. However, due to the dual effects of the fault and normal data fluctuations, vibration data fluctuates more frequently in time series, with a clear trend of increasing volatility. As the fault becomes more severe, the fault's cause becomes increasingly dominant, reducing the frequency of vibration data fluctuations. However, the fluctuations are more severe, and the trend is more pronounced.

[0074] (1) A sequence consisting of all elements in the vibration sequence of all cycles arranged in ascending time order is used as the total vibration sequence of the vehicle transmission system. The total vibration sequence is used as the input of a polynomial fitting algorithm, and the output is the equation of the fitting curve of the total vibration sequence. The polynomial fitting algorithm is a well-known technique, and the specific process is not described in detail here.

[0075] It should be noted that for the curve fitting of the total vibration sequence, this application only provides a curve fitting method. There are many existing curve fitting methods, and implementers can also use other curve fitting algorithms to perform curve fitting on the total vibration sequence. This application does not make specific restrictions.

[0076] (2) The vibration characteristic coefficients of all cycles arranged in ascending time order are recorded as a vibration characteristic sequence. The vibration characteristic sequence is used as the input of the STL decomposition (Seasonal-Trend Decomposition using LOESS) algorithm to obtain the trend sequence of the vibration characteristic sequence. The trend sequence is used as the input of the least squares linear fitting algorithm, and the output is the fitting line equation of the trend sequence. The STL decomposition algorithm and the least squares linear fitting algorithm are both well-known technologies, and the specific process is not repeated here.

[0077] It should be noted that for the decomposition of vibration feature sequences and the linear fitting of trend sequences, this application only provides a time series decomposition method and a linear fitting method. There are many existing time series decomposition methods and linear fitting methods. Implementers can also use other time series decomposition algorithms and linear fitting algorithms to decompose vibration feature sequences and perform linear fitting on trend sequences respectively. This application does not make specific restrictions.

[0078] (3) Based on the above analysis, the fault trend coefficient is calculated to measure the changing trend of the transmission system vibration. The expression is:

[0079] B = lg(X) × exp(-K), where B is the failure trend coefficient of the automobile transmission system, X is the number of data points with a derivative of 0 on the fitting curve of the total vibration sequence, K is the slope of the fitting straight line of the trend sequence, lg() is a logarithmic function with 10 as the real number, and exp() is an exponential function with the natural constant e as the base.

[0080] Since the frequency of vibration data fluctuations is relatively small both under normal conditions and in cases of more serious faults, the role of the logarithmic function is to reduce the influence of the frequency measure, while the role of the exponential function is to enhance the influence of the trend measure.

[0081] Under normal circumstances, vibration data fluctuates less frequently and lacks a clear trend. Therefore, X and |K| are both small, meaning lg(X) is small and exp(-K) is close to 1, resulting in a small fault trend coefficient, B. In the early stages of a fault, vibration data fluctuates more frequently and exhibits a general trend. Specifically, X is large, K is negative, and |K| increases. Therefore, lg(X) is large and exp(-K) increases, resulting in a large fault trend coefficient, B. As the fault grows, the frequency of vibration data fluctuations decreases, but the trend of vibration changes becomes more pronounced. Specifically, X decreases, K becomes negative, and |K| increases. Therefore, lg(X) is relatively small and exp(-K) is large. However, the exponential growth rate is faster than the logarithm, so the more severe the fault, the larger the fault trend coefficient, B.

[0082] Step S4: determining an abnormality ratio adjustment value of the vehicle transmission system based on the fault trend coefficient.

[0083] After obtaining the time series monitoring data of the automobile transmission system vibration data, the abnormal points therein can be detected by the anomaly detection algorithm. However, in the traditional isolation forest algorithm, when performing anomaly detection, the various parameters set therein are fixed. The abnormal ratio set in this way is often small, which can reduce the false alarm rate of the fault. This leads to a better fault detection effect when the fault characteristics are more obvious. However, in the early stage of the fault or when the fault is relatively minor, the difference between the data is not obvious enough, making it difficult to detect the abnormal points. If the abnormal ratio is set to a larger value, although it can improve the sensitivity of anomaly detection, so that it has better fault identification ability in the early stage of the fault or when the fault is relatively minor, it will increase the false alarm rate of the fault. Therefore, the choice of abnormal ratio affects the sensitivity and accuracy of the isolation forest algorithm for anomaly detection, and is also the key to balancing the application of this algorithm in automobile transmission system fault diagnosis. Therefore, it is necessary to adaptively adjust the abnormal ratio according to the vibration conditions of the automobile transmission system to improve the sensitivity and accuracy of anomaly detection. The expression is:

[0084] Where W is the abnormal ratio adjustment value of the vehicle transmission system, B is the failure trend coefficient of the vehicle transmission system, and arctan() is the inverse tangent function.

[0085] When the fault is more serious, the fault trend coefficient B is larger, the vibration data anomaly is more obvious, and the corresponding anomaly ratio is smaller, which can reduce the false alarm rate; conversely, when the fault is in the early stage or the fault is relatively minor, the abnormal change of the vibration data is smaller, and the corresponding anomaly ratio is larger, which can improve the sensitivity of anomaly identification and reduce the missed alarm rate.

[0086] Step S5: Based on the abnormal ratio adjustment value, the vehicle transmission system operation fault monitoring is performed in combination with the isolation forest algorithm.

[0087] The collected vibration sequences for all cycles are used as input to the Isolation Forest algorithm. The number of samples extracted at each time is 256, and the number of isolated trees in the forest is 100. The calculated abnormality ratio adjustment value W is used as the abnormality ratio in the Isolation Forest algorithm, and the output is the abnormality score of each periodic vibration sequence. The abnormality score of the vibration sequence for all cycles is used as the input of the Otsu method, and the output is the segmentation threshold. The Isolation Forest algorithm and the Otsu method are both well-known technologies, and the specific process will not be repeated here.

[0088] It should be noted that for the setting of the number of samples drawn each time and the number of isolated trees, the embodiment of this application only provides one setting method. The implementer can set the number of samples drawn each time and the number of isolated trees according to actual conditions. This application does not impose any specific restrictions.

[0089] The period in which the abnormal score of the vibration sequence is greater than the segmentation threshold is regarded as an abnormal period. All moments corresponding to all abnormal periods are arranged in ascending time order to form an abnormal sequence, the first-order difference sequence of the abnormal sequence is obtained, and the Hurst index of the first-order difference sequence is calculated. When the Hurst index is greater than the fault threshold, it indicates that the current vehicle transmission system has an operational fault and requires maintenance by relevant maintenance personnel; when the Hurst index is less than or equal to the fault threshold, it indicates that the current vehicle transmission system is operating normally. Preferably, in an embodiment of the present application, the fault threshold is set to 0.5. As other embodiments of the present application, the implementer can set the fault threshold according to actual conditions. Among them, the Hurst index is a well-known technology, and the specific process will not be repeated.

[0090] Based on the same inventive concept as the above-mentioned method, an embodiment of the present application also provides an operation fault monitoring system applied to an automobile transmission system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above-mentioned operation fault monitoring methods applied to an automobile transmission system are implemented.

[0091] In summary, the embodiment of the present application provides an operation fault monitoring method applied to an automobile transmission system. According to the time-frequency domain characteristics of the instantaneous vibration data of each cycle when the automobile transmission system is running, the vibration characteristic coefficient is calculated, which helps to improve the accuracy of identifying the difference between normal and abnormal states; then, the fault trend coefficient is calculated by combining the time-series variation trend of the vibration data and the vibration characteristic coefficient, which helps to improve the accuracy and sensitivity of subsequent fault diagnosis; finally, the abnormal proportion is adaptively adjusted according to the fault trend coefficient, which balances the accuracy and sensitivity of the isolated forest anomaly detection. In this way, by combining the time-frequency domain characteristics and time-series variation trend of the vibration data, the sensitivity of anomaly detection in the early stage of a fault or a minor fault is improved, and it can respond flexibly according to the actual vibration situation, thereby improving the adaptability and robustness of anomaly detection, ensuring the dynamic balance between the false alarm rate and the missed alarm rate of the transmission system fault detection, and improving the quality of automobile transmission system fault detection, thereby realizing high-precision and high-sensitivity operation fault monitoring of the automobile transmission system, ensuring the safety of the automobile transmission system and extending its service life to a certain extent.

[0092] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the above descriptions are of specific embodiments of the present application. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] The various embodiments in this application are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0094] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for monitoring operating faults in an automobile transmission system, characterized in that: The method comprises the following steps: Collect all vibration signals in each cycle of the vehicle transmission system and construct the vibration sequence of each cycle; Determine the vibration characteristic coefficient of each period based on the periodic variation characteristics of the elements in the vibration sequence of each period, the fluctuation of the vibration sequence, and the amplitude difference between different frequency regions of the vibration sequence in the frequency domain; Construct a total vibration sequence based on the vibration sequences of all cycles; construct a vibration feature sequence based on the vibration characteristic coefficients of all cycles; construct a fault trend coefficient of the automobile transmission system based on the frequency of data fluctuations in the total vibration sequence and the data change trend in the vibration characteristic sequence; determining an abnormality ratio adjustment value of the vehicle transmission system based on the fault trend coefficient; Based on the abnormal ratio adjustment value, the vehicle transmission system operation fault monitoring is performed in combination with the isolation forest algorithm; The process of obtaining the vibration characteristic coefficient of each period is as follows: Obtain the autocorrelation coefficient of the vibration sequence of each period through the autocorrelation function, and construct the autocorrelation function graph of each period vibration sequence; construct the period characteristic value of the autocorrelation function graph of each period based on the peak value distribution characteristics in the autocorrelation function graph; Calculate the variance of all elements in the vibration sequence of each period, recorded as the first variance; Obtain a frequency domain waveform of each periodic vibration sequence using a time-frequency conversion algorithm. In each frequency domain waveform, divide all frequencies into a low-frequency region, a medium-frequency region, and a high-frequency region; calculate the average value of the variance of the amplitudes of all frequencies in the medium-frequency region and the high-frequency region, and record it as a first average value; record frequencies in the frequency domain waveform whose amplitudes are greater than a preset frequency significance threshold as significant frequencies, and record the sequence consisting of the amplitudes of all significant frequencies in the low-frequency region as a low-frequency amplitude sequence; and obtain a low-frequency distribution feature metric for the low-frequency region based on changes in elements in a first-order difference sequence of the low-frequency amplitude sequence; A vibration characteristic coefficient of each period is constructed based on the period characteristic value, the first variance, and the difference between the first average value and the low-frequency distribution characteristic metric.

2. The method for monitoring operating faults of an automobile transmission system according to claim 1, wherein: The process of obtaining the periodic characteristic value is as follows: The peaks in the autocorrelation function graph of each period that are greater than the preset correlation threshold are recorded as significant peaks; The hysteresis corresponding to each significant peak value is used as the ordinate, and the order of appearance of each significant peak is used as the abscissa to obtain the coordinate point corresponding to each significant peak; the coordinate points corresponding to all significant peaks in the autocorrelation function graph are fitted using a linear fitting algorithm, and the obtained fitting straight line is recorded as the hysteresis fitting straight line; and the determination coefficient of the hysteresis fitting straight line is calculated; The product of the mean value of all significant peak-to-peak values ​​in the autocorrelation function diagram of each period and the determination coefficient is used as the period characteristic value of the autocorrelation function diagram of each period.

3. The method for monitoring operating faults of an automobile transmission system according to claim 1, wherein: The expression of the low-frequency distribution characteristic metric is: In the formula, lf i is the low-frequency distribution characteristic measure of the low-frequency area in the frequency domain waveform of the i-th cycle, μ(L i ),σ(L i ) are the mean and variance of all elements in the first-order difference sequence of the low-frequency amplitude sequence in the i-th period, and norm() is the normalization function.

4. The method for monitoring operating faults in an automobile transmission system according to claim 2, wherein: The expression of the vibration characteristic coefficient is: Where A i is the vibration characteristic coefficient of the i-th period, R i is the periodic characteristic value of the i-th period, σ(V i ) is the first variance of the i-th period, t i is the lag corresponding to the first significant peak in the autocorrelation function diagram of the ith period, mf i is the first average value of the i-th cycle, lf i is the low-frequency distribution characteristic measure of the i-th period, and exp() is an exponential function with the natural constant e as the base.

5. The method for monitoring operating faults in an automobile transmission system according to claim 1, wherein: The process of obtaining the failure trend coefficient of the automobile transmission system is as follows: The total vibration sequence is used as the input of the polynomial fitting algorithm, and the output is the fitting curve of the total vibration sequence; Using time series decomposition algorithm, the trend sequence of vibration characteristic sequence is obtained; A fault trend coefficient of the automobile transmission system is constructed based on changes in the fitting curve and the fitting straight line of the trend sequence.

6. The method for monitoring operating faults in an automobile transmission system according to claim 5, wherein: The expression of the fault tendency coefficient is: B=lg(X)×exp(-K), where B is the failure trend coefficient of the automobile transmission system, X is the number of data points whose derivative is 0 on the fitting curve, K is the slope of the fitting straight line of the trend sequence, lg() is a logarithmic function with 10 as the real number, and exp() is an exponential function with the natural constant e as the base.

7. The method for monitoring operating faults in an automobile transmission system according to claim 1, wherein: The expression of the abnormal ratio adjustment value of the automobile transmission system is: Where W is the abnormal ratio adjustment value of the vehicle transmission system, B is the failure trend coefficient of the vehicle transmission system, and arctan() is the inverse tangent function.

8. The method for monitoring operating faults in an automobile transmission system according to claim 1, wherein: The automobile transmission system operation fault monitoring is performed based on the abnormal ratio adjustment value and in combination with the isolation forest algorithm, specifically: The vibration sequences of all cycles are used as input of the isolation forest algorithm, the abnormality ratio adjustment value is used as the abnormality ratio in the isolation forest algorithm, and the abnormality score of each vibration sequence of each cycle is output; and a segmentation threshold value of the abnormality score of the vibration sequences of all cycles is obtained; The period in which the abnormality score of the vibration sequence is greater than the segmentation threshold is regarded as an abnormal period; the sequence consisting of all moments corresponding to all abnormal periods is obtained and recorded as an abnormal sequence, and the Hurst index of the first-order difference sequence of the abnormal sequence is calculated. If the Hurst index is greater than the preset fault threshold, then the current vehicle transmission system has an operating fault; otherwise, the current vehicle transmission system is operating normally.

9. An operation fault monitoring system for an automobile transmission system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images