Automobile fault remote diagnosis method and system

Abstract

The invention discloses an automobile fault remote diagnosis method and system, and particularly relates to the technical field of automobile fault remote diagnosis, an output voltage signal of a throttle position sensor is obtained through a vehicle-mounted diagnosis system, and tiny voltage drift or fluctuation caused by a resistive film at the initial stage of abrasion is effectively identified through time-frequency domain analysis, so that the fault diagnosis accuracy is improved. By calculating the joint entropy fluctuation coefficient and the signal time sequence similar abnormal coefficient, the health condition of the throttle valve sensor is evaluated more comprehensively, the accuracy of hidden gradual change fault diagnosis is improved, and by constructing a hidden gradual change fault evaluation model and outputting an evaluation index, the fault diagnosis accuracy is improved. The method can carry out effective quantitative evaluation on the hidden gradual change fault at the initial wear stage of the resistive film, and can dynamically monitor the evolution trend of the fault and timely carry out early warning by carrying out clustering analysis on the hidden gradual change fault degrees at different time points, thereby avoiding the traditional detection hysteresis based on fault code triggering, and improving the detection accuracy. And a more efficient and sensitive fault identification and prevention means is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of remote diagnosis of vehicle faults. More specifically, the present invention relates to a method and system for remote diagnosis of vehicle faults. Background Art

[0002] With the development of intelligent vehicles and vehicle networking technologies, remote fault diagnosis systems have gradually become the core components of the vehicle's intelligent operation and maintenance system. Such systems typically rely on various sensor data and fault diagnostic codes (DTCs) generated by electronic control units (ECUs) for fault detection and identification. Among them, the throttle position sensor, as a key component in the power control system, is mainly used to monitor the throttle opening in real time and feedback the corresponding position signal to the ECU in the form of an analog voltage to achieve precise control of the engine intake air volume.

[0003] Existing throttle position sensors mostly adopt a potentiometer structure, in which the internal resistance film generates position-related resistance value changes as the throttle shaft rotates, thereby outputting an analog voltage signal corresponding to the throttle opening. This type of sensor design has the advantages of simple structure and fast response. However, during long-term use, especially in the area of the commonly used opening position, the resistance film is prone to local wear or contact point fatigue, resulting in drift or even intermittent non-contact areas in the sensor output characteristics, forming a hidden gradual fault mode. In the early stage of such wear faults, the sensor output signal may only show a slight voltage offset or fluctuation, which has not exceeded the diagnostic threshold set by the ECU, so the standard fault code (DTC) will not be triggered. Such hidden degradation may have limited impact on power performance in the short term, but as the wear gradually intensifies, it will cause a decrease in throttle control accuracy, leading to performance degradation problems such as unstable idling, increased fuel consumption, and delayed power response, and it is very likely to be missed by the existing fault code-driven remote diagnosis system. Therefore, there is an urgent need to construct a new type of vehicle remote fault diagnosis method and system to achieve high-sensitivity identification and non-invasive evaluation of hidden gradual faults of throttle sensors. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method and system for remote diagnosis of vehicle faults to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for remote diagnosis of vehicle faults includes the following steps:

[0007] Step S1, obtaining the output voltage signal of the throttle position sensor through the on-board diagnostic system, and performing joint time-frequency domain analysis on it to identify the initial wear state of the resistance film;

[0008] Step S2, when the resistance film is in the initial wear state, obtain the joint entropy information of the throttle position sensor and other vehicle-mounted sensors, and calculate the joint entropy fluctuation coefficient of the sensing signals according to the joint entropy information of the throttle position sensor and other vehicle-mounted sensors;

[0009] Step S3, when the resistance film is in the initial wear state, obtain the timing similarity information of the throttle position sensor and the accelerator pedal signal, and calculate the signal timing similarity anomaly coefficient according to the timing similarity information of the throttle position sensor signal and the accelerator pedal signal;

[0010] Step S4, construct a hidden gradual fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signals and the signal timing similarity anomaly coefficient, output a hidden gradual fault evaluation index, and evaluate the degree of hidden gradual fault in the initial wear stage of the resistance film;

[0011] Step S5, perform cluster analysis on the degrees of hidden gradual faults at different time points, and give an early warning of the hidden gradual faults in the initial wear stage of the resistance film.

[0012] In a preferred embodiment, in step S1, the output voltage signal of the throttle position sensor is collected in real time through the on-vehicle diagnostic system, and the output voltage signal is filtered and denoised;

[0013] Perform joint time-frequency domain analysis on the filtered and denoised output voltage signal of the throttle position sensor to identify the initial wear state of the resistance film, specifically as follows:

[0014] Select a wavelet basis function to perform wavelet packet transform on the output voltage signal Scl(t). After the transform, multiple decomposed sub-signals are obtained. Each sub-signal corresponds to a different frequency band. Mark the decomposed sub-signals as C j (t), where j = {1, 2,..., 2 J} represents different frequency bands, J is the number of transform layers, and each C j (t) contains signal components with different frequency ranges;

[0015] Match the decomposed sub-signals with the preset resistance film wear characteristic signals, and select a group of sub-signals C k (t) with a matching degree greater than the matching degree threshold, where k = {1, 2,..., K}, and K is the total number of selected sub-signals;

[0016] For each selected sub-signal C k (t), calculate its energy E k , and the calculation formula is: E k = ∑ t |C k (t)| 2 , where Ek Represents sub-signal C k (t) energy;

[0017] Calculate sub-signal C k (t) and sub-signal C l (t) energy ratio, the calculation formula is as follows: Where Rn kl Represents sub-signal C k (t) and sub-signal C l (t) energy ratio, E k Represents sub-signal C k (t) energy, E l Represents sub-signal C l (t) energy, k, l ∈ {1, 2,..., K}.

[0018] In a preferred embodiment, compare the energy ratio between different sub-signals with a preset energy ratio threshold. If the energy ratio is greater than the energy ratio threshold, accumulate the energy ratios greater than the energy ratio threshold to obtain the wear energy fluctuation coefficient;

[0019] Compare the wear energy fluctuation coefficient with a preset wear energy fluctuation coefficient threshold. If the wear energy fluctuation coefficient is greater than the wear energy fluctuation coefficient threshold, mark the current resistance film state as the initial wear state.

[0020] In a preferred embodiment, the acquisition logic of the sensing signal joint entropy fluctuation coefficient is as follows:

[0021] The other vehicle-mounted sensors described include a mass air flow sensor and a manifold absolute pressure sensor;

[0022] Obtain the output signal Scl(t) of the throttle position sensor, the output signal Zlk(t) of the mass air flow sensor, and the output signal Qgy(t) of the manifold absolute pressure sensor, and construct a multi-dimensional signal sequence: SS(t) = {Scl(t), Zlk(t), Qgy(t)};

[0023] Preset multiple sliding window scales {τ1, τ2,..., τ i} Each scale represents a different time granularity, and perform the following operations for each scale:

[0024] Operation A1: In each sliding window, construct a joint probability density distribution using kernel density estimation: Represents the joint probability density calculated for any sliding window scale τ i ; Represents the joint probability density of the sliding window scale τ i ; Among them, YB represents the sliding window scale τ i The total number of sampling points, DK represents the bandwidth parameter, HS represents the Gaussian kernel function, XL j represents the multi-dimensional sensor output signal vector of the j-th sampling point, represents the sliding window scale τ i The mean value of the multi-dimensional sensor output signal vector within;

[0025] Operation A2: Calculate the local joint entropy: Among them represents the sliding window scale τ i The local joint entropy of;

[0026] Calculate the sliding window scale τ i And the sliding window scale τ n The joint entropy difference of: Among them represents the sliding window scale τ i And the sliding window scale τ n The joint entropy difference value of, represents the sliding window scale τ n The local joint entropy of;

[0027] All pairwise combinations of the sliding window scales are obtained to get the perturbation spectrum DT: Where i, n ∈ {1, 2,..., N}, and N is the total number of sliding window scales;

[0028] Extract the perturbation mean from the perturbation spectrum DT: Where μΔLS represents the perturbation mean;

[0029] Extract the perturbation standard deviation from the perturbation spectrum DT: Where σΔLS represents the perturbation standard deviation;

[0030] Calculate the joint entropy fluctuation coefficient of the sensing signal: Where cxs represents the joint entropy fluctuation coefficient of the sensing signal.

[0031] In a preferred embodiment, the acquisition logic of the signal time series similarity anomaly coefficient is as follows:

[0032] Synchronously collect the output voltage signal of the throttle position sensor and the accelerator pedal signal from the on-board diagnostic system to form a time series, and extract the output voltage signal of the throttle position sensor and the accelerator pedal signal to construct a D×F distance matrix DD: DD(d,f) = |Scl d -Ymt f |, where DD(d,f) is the element in the d-th row and f-th column of the distance matrix DD, Scld Denote the output voltage signal of the d-th throttle position sensor extracted as Ymt f Denote the f-th accelerator pedal signal extracted, where d = {1, 2, ..., D}, D is the total number of sampling points of the output voltage signal of the throttle position sensor, and f = {1, 2, ..., F}, F is the total number of sampling points of the accelerator pedal signal;

[0033] Calculate the minimum cumulative distance from the starting point (1, 1) to the ending point (D, F) through dynamic programming: Among them, C(d, f) represents the minimum cumulative distance from the starting point to the point (d, f);

[0034] Backtrack from the ending point (D, F) to the starting point (1, 1) in reverse. During the backtracking process, select the minimum forward path to obtain the dynamic time warping distance DT between the output voltage signal of the throttle position sensor and the accelerator pedal signal: DT = C(D, F), where C(D, F) is the total minimum cumulative distance;

[0035] Calculate the derivative sequences of the output voltage signal of the throttle position sensor and the accelerator pedal signal: Among them, ΔScl represents the derivative of the output voltage signal of the throttle position sensor with respect to the time variable t, and ΔYmt represents the derivative of the accelerator pedal signal with respect to the time variable t;

[0036] Calculate the phase matching error DP: DP = |ΔScl - ΔYmt|;

[0037] Calculate the correlation coefficient DX between the output voltage signal of the throttle position sensor and the accelerator pedal signal: Among them, Cov(Scl, Ymt) represents the covariance between the output voltage signal of the throttle position sensor and the accelerator pedal signal, σScl represents the standard deviation of the output voltage signal of the throttle position sensor, and σYmt represents the standard deviation of the accelerator pedal signal;

[0038] Calculate the signal time series similarity anomaly coefficient: xsy = a1 * DT + a2 * DP + a3 * (1 - DX), where xsy represents the signal time series similarity anomaly coefficient, and a1, a2, a3 respectively represent the preset proportional coefficients of the dynamic time warping distance, phase matching error, and correlation coefficient, and a1, a2, a3 are all greater than 0.

[0039] In a preferred embodiment, a latent gradual fault evaluation model is constructed based on the combined entropy fluctuation coefficient and the signal time series similarity anomaly coefficient of the sensing signal, and a latent gradual fault evaluation index yxj is output. The formula on which the model is based is as follows: yxj = b1 * cxs + b2 * xsy, where b1 and b2 respectively represent the preset proportionality coefficients of the combined entropy fluctuation coefficient and the signal time series similarity anomaly coefficient of the sensing signal, and both b1 and b2 are greater than 0.

[0040] In a preferred embodiment, the latent gradual fault evaluation index is compared with a preset latent gradual fault evaluation index threshold to classify the degree of latent gradual fault in the initial stage of resistor film wear, as follows:

[0041] If the latent gradual fault evaluation index is greater than the latent gradual fault evaluation index threshold, a high latent gradual fault degree is generated; if the latent gradual fault evaluation index is less than or equal to the latent gradual fault evaluation index threshold, a low latent gradual fault degree is generated.

[0042] In a preferred embodiment, whenever a high latent gradual fault degree is generated, a time series of the latent gradual fault evaluation index output by the latent gradual fault evaluation model is obtained, and cluster analysis is performed on the time series of the latent gradual fault evaluation index, and early warning of the latent gradual fault in the initial stage of resistor film wear is carried out according to the results of the cluster analysis, as follows:

[0043] Step S51, determine the number R of cluster centers of the K-means algorithm according to the silhouette coefficient method, and randomly select R latent gradual fault evaluation indexes from the time series of the latent gradual fault evaluation index as the initial cluster centers;

[0044] Step S52, calculate the Euclidean distance between each latent gradual fault evaluation index in the time series of the latent gradual fault evaluation index and the cluster center, and assign each latent gradual fault evaluation index to the cluster center with the closest Euclidean distance;

[0045] Step S53, calculate the average value of the latent gradual fault evaluation indexes in each cluster center and use it as the new cluster center coordinates;

[0046] Step S54, repeat Step S52 and Step S53 to iteratively update the cluster center until the cluster center no longer changes;

[0047] According to the finally obtained multiple cluster centers, calculate the difference between the coordinates of different cluster centers, and accumulate the differences between the coordinates of different cluster centers to obtain a difference cumulative value.

[0048] In a preferred embodiment, the difference cumulative value is compared with a preset difference cumulative value threshold to give an early warning of the latent gradual fault in the initial stage of resistor film wear, as follows:

[0049] If the cumulative difference value is greater than the cumulative difference value threshold, an immediate warning signal is generated;

[0050] If the cumulative difference value is less than or equal to the cumulative difference value threshold, no warning is given temporarily, and the latent gradual change state is continuously monitored.

[0051] In a preferred embodiment, an automotive fault remote diagnosis system includes a wear recognition module, an information coupling perception module, a timing similarity analysis module, a latent gradual change fault evaluation module, and a latent gradual change warning module;

[0052] The wear recognition module is used to obtain the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system, and perform joint time-frequency domain analysis on it to identify the initial wear state of the resistance film;

[0053] The information coupling perception module is used to obtain the joint entropy information of the throttle position sensor and other on-vehicle sensors when the resistance film is in the initial wear state, and calculate the joint entropy fluctuation coefficient of the sensing signals according to the joint entropy information of the throttle position sensor and other on-vehicle sensors;

[0054] The timing similarity analysis module is used to obtain the timing similarity information of the throttle position sensor and the accelerator pedal signal when the resistance film is in the initial wear state, and calculate the signal timing similarity anomaly coefficient according to the timing similarity information of the throttle position sensor signal and the accelerator pedal signal;

[0055] The latent gradual change fault evaluation module constructs a latent gradual change fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signals and the signal timing similarity anomaly coefficient, outputs a latent gradual change fault evaluation index, and evaluates the latent gradual change fault degree in the initial wear stage of the resistance film;

[0056] The latent gradual change warning module performs clustering analysis on the latent gradual change fault degrees at different time points, and gives a warning about the latent gradual change fault in the initial wear stage of the resistance film.

[0057] The technical effects and advantages of the present invention:

[0058] 1. The present invention obtains the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system. Through time-frequency domain analysis, it can effectively identify the tiny voltage drift or fluctuation caused by the wear of the resistance film in the initial stage of wear, avoiding the problem that traditional diagnostic methods cannot detect these subtle changes, ensuring the effectiveness of early warning. By calculating the joint entropy fluctuation coefficient and the signal time-series similarity anomaly coefficient, it can more comprehensively evaluate the health status of the throttle sensor, improve the accuracy of diagnosing latent gradual faults, avoid the limitations of single-signal analysis, better capture the early signs of sensor degradation or failure. By constructing a latent gradual fault evaluation model and outputting an evaluation index, it can effectively quantify and evaluate the latent gradual faults in the initial stage of resistance film wear. Compared with the traditional diagnostic method triggered by fault codes, this method can identify even the tiny changes that do not exceed the ECU diagnostic threshold, and thus discover potential latent faults in advance. This plays an important role in improving the fault tolerance, reliability, and early warning ability of the vehicle remote fault diagnosis system. By performing cluster analysis on the degree of latent gradual faults at different time points, it can dynamically monitor the evolution trend of faults and give early warnings in a timely manner. This early warning mechanism is not only effective in the initial stage of resistance film wear, but also can continuously track the progress of faults, ensure that faults are dealt with in a timely manner before they occur, avoid the detection lag of traditional fault-code-triggered methods, and provide a more efficient and sensitive means of fault identification and prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings;

[0060] Figure 1 is a flowchart of the method of Embodiment 1 of the present invention;

[0061] Figure 2 is a flowchart of the system of Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1: Figure 1 A remote vehicle fault diagnosis method of the present invention is given, including the following steps:

[0064] Step S1, obtain the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system, and perform time-frequency domain joint analysis on it to identify the initial wear state of the resistance film;

[0065] Step S2: When the resistance film is in the initial wear state, obtain the joint entropy information of the throttle position sensor and other vehicle sensors, and calculate the joint entropy fluctuation coefficient of the sensing signals based on the joint entropy information of the throttle position sensor and other vehicle sensors;

[0066] Step S3: When the resistance film is in the initial wear state, obtain the timing similarity information of the throttle position sensor and the accelerator pedal signal, and calculate the signal timing similarity anomaly coefficient based on the timing similarity information of the throttle position sensor signal and the accelerator pedal signal;

[0067] Step S4: Construct a hidden gradual fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signals and the signal timing similarity anomaly coefficient, output the hidden gradual fault evaluation index, and evaluate the degree of hidden gradual fault in the initial wear stage of the resistance film;

[0068] Step S5: Conduct cluster analysis on the degrees of hidden gradual faults at different time points to give an early warning of the hidden gradual faults in the initial wear stage of the resistance film;

[0069] In step S1, the output voltage signal of the throttle position sensor is collected in real time through a vehicle diagnostic system (such as an OBD-II interface), and the output voltage signal is filtered and denoised to ensure the quality and accuracy of the signal;

[0070] It should be noted that common filtering processes include low-pass filtering, high-pass filtering, band-pass filtering, median filtering, Kalman filtering, etc., which will not be elaborated here;

[0071] Perform a joint time-frequency domain analysis on the output voltage signal of the throttle position sensor after filtering and denoising to identify the initial wear state of the resistance film, specifically as follows:

[0072] Select a wavelet basis function (such as Daubechies wavelet, Symlet wavelet, etc., and the Daubechies wavelet function is selected as the wavelet basis function in this embodiment) to perform wavelet packet transform on the output voltage signal Scl(t). After the transform, multiple decomposed sub-signals are obtained. Each sub-signal corresponds to a different frequency band. Mark the decomposed sub-signals as C j (t), where j = {1, 2,..., 2 J} represents different frequency bands, J is the number of transform layers, and each C j (t) contains signal components with different frequency ranges;

[0073] Match the decomposed sub-signals with the preset resistance film wear characteristic signals, and select a group of sub-signals C k (t) with a matching degree greater than the matching degree threshold, where k = {1, 2,..., K}, and K is the total number of selected sub-signals;

[0074] It should be noted that the common methods for feature matching between the decomposed sub-signals and the preset resistance film wear characteristic signals include correlation coefficient matching, energy feature matching, and machine learning model matching. The feature matching method adopted in this embodiment is correlation coefficient matching, that is, the Pearson correlation coefficient between the decomposed sub-signal and the preset resistance film wear characteristic signal is calculated as the matching degree, and the group of sub-signals with the largest Pearson correlation coefficient is selected as the final choice. The feature matching method can be selected according to the actual situation and will not be elaborated here;

[0075] For each selected sub-signal C k (t), calculate its energy E k , and the calculation formula is: E k =∑ t |C k (t)| 2 , where E k represents the energy of the sub-signal C k (t), reflecting the signal strength of this frequency band;

[0076] Calculate the energy ratio of the sub-signal C k (t) and the sub-signal C l (t), which reflects the relative change of energy between different sub-signals. The calculation formula is as follows: Where Rn kl represents the energy ratio of the sub-signal C k (t) and the sub-signal C l (t), E k represents the energy of the sub-signal C k (t), E l represents the energy of the sub-signal C l (t), k, l ∈ {1, 2,..., K};

[0077] It should be noted that the above formulas are all calculated by taking the numerical value after removing the dimension. The common methods for removing the dimension include Min-Max normalization, Z-Score standardization, etc., which will not be elaborated here;

[0078] Compare the energy ratio between different sub-signals with the preset energy ratio threshold. If the energy ratio is greater than the energy ratio threshold, it indicates that there may be frequency drift or energy fluctuation caused by wear. Accumulate the energy ratios greater than the energy ratio threshold to obtain the wear energy fluctuation coefficient;

[0079] Compare the wear energy fluctuation coefficient with the preset wear energy fluctuation coefficient threshold. If the wear energy fluctuation coefficient is greater than the wear energy fluctuation coefficient threshold, mark the current resistance film state as the initial wear state;

[0080] Step S2, when the resistance film is in the initial wear state, obtain the joint entropy information of the throttle position sensor and other vehicle sensors, and calculate the joint entropy fluctuation coefficient of the sensing signals based on the joint entropy information of the throttle position sensor and other vehicle sensors.

[0081] In the present invention, the joint entropy fluctuation coefficient of the sensing signals is a quantitative index for measuring the dynamic information coupling relationship between the throttle position sensor and other vehicle sensors. The core lies in measuring the correlation between different sensor signals and the stability of information interaction. This coefficient not only reflects the stability of the throttle position sensor signal itself, but also reveals the degree of fluctuation in the multi-source information fusion relationship between it and multiple key variables in the vehicle control system during the time-series evolution process. It is one of the important characteristics for judging the potential performance degradation of the throttle position sensor, especially in the initial wear state of the resistance film.

[0082] Under normal conditions, the joint entropy between the throttle position sensor and other vehicle sensors has a certain stability, that is, its multi-variable joint information structure shows strong consistency and regularity. At this time, the joint entropy fluctuation coefficient of the sensing signals is small, indicating that the signal linkage between various sensors is stable and the coupling is good, and the overall operation of the system is in a healthy state. Once the resistance film shows initial wear, its output voltage signal will produce a slight drift in some position segments where the throttle opening is frequently used. This drift may not be sufficient to trigger a fault code, but it will quietly affect the coordination between the throttle and other control signals. For example, the coordination between the throttle opening and the throttle input may be weakened, and the matching with the air-fuel ratio adjustment may also shift, thus triggering a microscopic disturbance in the information coupling mode.

[0083] This disturbance will cause irregular fluctuations in the joint entropy between the sensors, that is, the multi-source information synergy of the system shows a small but continuous "loosening", resulting in an increase in the joint entropy fluctuation coefficient of the sensing signals. Therefore, a larger joint entropy fluctuation coefficient of the sensing signals reflects an abnormal local information decoupling trend in the system, indicating that the logical consistency between the output behavior of the throttle position sensor and other subsystems is weakening. This change is often one of the important manifestations of the latent gradual change fault in the initial wear of the resistance film. On the contrary, a smaller joint entropy fluctuation coefficient indicates that the information structure between the sensors is maintained well, indicating that the throttle position sensor has not shown obvious signs of performance degradation.

[0084] By constructing the joint entropy fluctuation coefficient of sensing signals, the present invention can identify the potential performance degradation risk of the throttle position sensor in the early wear stage of the resistance film in advance without relying on the trigger of fault codes. Its advantages are reflected in the following aspects: First, it can reveal the hidden signal abnormality rules from the level of system information coupling, breaking through the limitations of traditional single-sensor abnormality detection; second, it can quantitatively express the dynamic uncertainty changes of complex sensing networks, enhancing the sensitivity and accuracy of the diagnostic model. Therefore, evaluating the latent gradual fault in the initial stage of resistance film wear based on the joint entropy fluctuation coefficient of sensing signals not only improves the perception ability of the remote diagnosis system for latent faults without fault codes, but also significantly enhances the intelligent fault tolerance and forward warning performance of the system, having significant engineering application value and industrial promotion prospects.

[0085] The acquisition logic of the joint entropy fluctuation coefficient of sensing signals is as follows:

[0086] The other vehicle-mounted sensors mentioned include the mass air flow sensor and the manifold absolute pressure sensor;

[0087] It should be noted that the mass air flow sensor is used to measure the mass of air entering the engine per unit time, and the manifold absolute pressure sensor is used to measure the absolute pressure in the intake manifold;

[0088] Obtain the output signal Scl(t) of the throttle position sensor, the output signal Zlk(t) of the mass air flow sensor, and the output signal Qgy(t) of the manifold absolute pressure sensor, and construct a multi-dimensional signal sequence: SS(t) = {Scl(t), Zlk(t), Qgy(t)};

[0089] Preset multiple sliding window scales {τ1, τ2,..., τ i}, each scale representing a different time granularity (such as 1s, 5s, 10s...), and perform the following operations for each scale:

[0090] Operation A1: In each sliding window, construct the joint probability density distribution using kernel density estimation: Represents the joint probability density calculated for any one sliding window scale τ i ; Represents the joint probability density of the sliding window scale τ i ; where YB represents the total number of sampling points of the sliding window scale τ i , DK represents the bandwidth parameter, HS represents the Gaussian kernel function, XL j represents the multi-dimensional sensor output signal vector of the jth sampling point, represents the mean value of the multi-dimensional sensor output signal vectors within the sliding window scale τ i ;

[0091] It should be noted that the bandwidth parameter refers to a smoothing control parameter used to determine the "influence range" or "diffusion degree" of the kernel function around the data points. The Gaussian kernel function is one of the most common kernel function forms and is used for weighting near the sample points;

[0092] Operation A2: Calculate the local joint entropy: where represents the local joint entropy of the sliding window scale τ i ;

[0093] Calculate the joint entropy difference between the sliding window scale τ i and the sliding window scale τ n : where represents the joint entropy difference value between the sliding window scale τ i and the sliding window scale τ n , represents the local joint entropy of the sliding window scale τ n ;

[0094] Combine all pairs of sliding window scales to obtain the perturbation spectrum DT: where i, n ∈ {1, 2,..., N}, and N is the total number of sliding window scales;

[0095] The perturbation spectrum reflects the "stability / instability" of the coupling between the throttle position signal and other signal information under multiple time granularities;

[0096] Extract the perturbation mean from the perturbation spectrum DT: where μΔLS represents the perturbation mean;

[0097] Extract the perturbation standard deviation from the perturbation spectrum DT: where σΔLS represents the perturbation standard deviation;

[0098] Calculate the joint entropy fluctuation coefficient of the sensing signal: where cxs represents the joint entropy fluctuation coefficient of the sensing signal;

[0099] It should be noted that the above formulas are all dimensionless and take their numerical values for calculation. Common methods for removing dimensions include Min - Max normalization, Z - Score standardization, etc., which will not be elaborated here;

[0100] In the present invention, the calculation of the combined entropy fluctuation coefficient of the sensing signals is mainly used to measure the signal coupling degree and dynamic changes between the throttle position sensor and other vehicle-mounted sensors (such as the mass air flow sensor and the manifold absolute pressure sensor). This process identifies the potential non-linear dependence relationships and the fluctuation characteristics of information coupling between the signals through multi-dimensional signal sequence construction and multi-time granularity analysis at the sliding window scale. First, the output signals of three sensors are obtained from the on-board diagnostic system: the signal of the throttle position sensor, the signal of the mass air flow sensor, and the signal of the manifold absolute pressure sensor. By synchronously collecting the time series, a multi-dimensional signal sequence is formed, where each signal can be regarded as a feature dimension, reflecting the working states and environmental parameters of different vehicle-mounted sensors. Next, multiple sliding window scales are set for the multi-dimensional sequence of each sensor signal. These scales represent different time granularities (such as 1 second, 5 seconds, 10 seconds, etc.), and each scale reflects different dynamic information. Specifically, shorter time granularities capture the rapid fluctuations of the signals, while longer time granularities can capture the chronic changes of the signals. At each sliding window scale, first, the joint probability density of the signals within each window needs to be calculated. The joint probability density reflects the joint behavior of the output signals of each sensor within a specific time window, that is, the common change pattern between the signals within this time range. For this purpose, kernel density estimation is used to calculate the joint probability density distribution of the signals within each window. Kernel density estimation is a non-parametric method used to estimate the probability density distribution of multi-dimensional signals by smoothing each sampling point. At each sliding window scale, the goal of kernel density estimation is to construct a smooth joint probability distribution, which can reveal the correlation and dependence structure between the signals. The bandwidth parameter is used to control the degree of smoothing, and the Gaussian kernel function acts on each sampling point through its distribution function to calculate its contribution to the overall probability distribution. After obtaining the joint probability density distribution, the local joint entropy at each window scale is further calculated. The local joint entropy measures the information uncertainty between the signals at this scale. The higher the entropy value, the greater the randomness and complexity between the signals; while a lower entropy value means a stronger dependence between the signals and a more stable dynamic behavior of the system. In this way, we can quantify the complexity of the signals at different time scales and reveal the potential non-linear coupling patterns between the signals. Next, the difference in joint entropy between different sliding window scales is calculated. Since the signal changes at different scales have different time granularities, calculating the difference in joint entropy can compare the stability and variability of signal coupling at each scale. By calculating the difference in joint entropy between every two sliding window scales, a set of perturbation maps is obtained. This perturbation map shows the instability of signal coupling and reveals the variability of the signals at different time scales. In the perturbation map, we extract the perturbation mean and the perturbation standard deviation.The perturbation mean reflects the average fluctuation level of signal coupling, while the perturbation standard deviation represents the amplitude of the fluctuation. A larger perturbation standard deviation indicates that the signal fluctuates greatly at different scales, and there may be an unstable coupling relationship in the system, which may further indicate system faults or potential fault risks. By comprehensively considering the mean and standard deviation in the perturbation spectrum, the joint entropy fluctuation coefficient of the sensing signal is calculated. This coefficient quantifies the intensity of the coupling fluctuation between signals and can reflect the stability and instability of signals at multiple time granularities. A higher joint entropy fluctuation coefficient usually indicates that the coupling relationship between signals is relatively complex and there may be signs of hidden faults. While a lower fluctuation coefficient may mean that the dependence between signals is stronger and the system operates more smoothly. Through this method, the coupling state between the throttle position sensor and other vehicle sensors can be dynamically monitored and evaluated, providing strong data support for the early warning of system faults, especially for the early detection and diagnosis of hidden gradual faults such as resistance film wear.

[0101] Step S3, when the resistance film is in the initial wear state, obtain the timing similarity information between the throttle position sensor and the accelerator pedal signal, and calculate the signal timing similarity anomaly coefficient according to the timing similarity information between the throttle position sensor signal and the accelerator pedal signal;

[0102] In the present invention, the signal timing similarity anomaly coefficient is an index used to measure the consistency of the dynamic response relationship between the throttle position sensor signal and the accelerator pedal signal, and can reflect the matching degree and abnormal deviation of the two at the time series level. Under normal vehicle operating conditions, as the input source of the driving intention, the signal change of the accelerator pedal should be highly synchronized with the throttle position change, forming a stable causal coupling relationship. That is, when the driver steps on the accelerator, the throttle should respond precisely, and the timing behaviors of the two should maintain high similarity in terms of amplitude trend, response delay, and fluctuation pattern. However, when the resistance film is in the initial wear stage, the output voltage of the sensor will experience slight drift and response jitter, resulting in a decrease in the followability of the throttle position signal to the throttle input, thereby causing a deviation in the timing similarity between the two. Therefore, the signal timing similarity anomaly coefficient proposed in the present invention is used to evaluate the deviation degree of its time matching mode from the normal reference mode. A larger signal timing similarity anomaly coefficient indicates that abnormal behaviors such as non-linear deviation, response lag, or weak decoupling have occurred in the response between the throttle position signal and the throttle input, which may reflect a decrease in the signal modulation ability caused by the wear of the resistance film; while a smaller anomaly coefficient indicates that the mapping relationship between the two on the time axis is still stable, not significantly disturbed, and the working state of the resistance film is normal. By calculating this coefficient, the transition stage from normal to abnormal during the wear process of the resistance film can be effectively captured, and then early identification and quantitative evaluation of latent gradual faults can be realized. This method can expose potential risks in advance based on the similarity degradation characteristics of the response behavior without triggering the traditional ECU fault threshold, avoiding sudden failures of the vehicle.

[0103] The acquisition logic of the signal timing similarity anomaly coefficient is as follows:

[0104] Synchronously collect the output voltage signal of the throttle position sensor and the accelerator pedal signal from the on-vehicle diagnostic system to form a time series, and extract the output voltage signal of the throttle position sensor and the accelerator pedal signal to construct a D×F distance matrix DD: DD(d,f) = |Scl d -Ymt f |, where DD(d,f) is the element in the d-th row and f-th column of the distance matrix DD, Scl d represents the d-th extracted output voltage signal of the throttle position sensor, and Ymt f represents the f-th extracted accelerator pedal signal, d = {1,2,...,D}, D is the total number of sampling points of the output voltage signal of the throttle position sensor, f = {1,2,...,F}, and F is the total number of sampling points of the accelerator pedal signal;

[0105] Calculate the minimum cumulative distance from the starting point (1, 1) to the ending point (D,F) through dynamic programming: where C(d,f) represents the minimum cumulative distance from the starting point to the point (d,f);

[0106] Backtrack from the end point (D,F) to the starting point (1,1) in reverse. During the backtracking process, select the minimum forward path to obtain the dynamic time warping distance DT between the throttle position sensor output voltage signal and the accelerator pedal signal: DT = C(D,F), where C(D,F) is the total minimum cumulative distance;

[0107] Calculate the derivative sequences of the throttle position sensor output voltage signal and the accelerator pedal signal: where ΔScl represents the derivative of the throttle position sensor output voltage signal with respect to the time variable t, and ΔYmt represents the derivative of the accelerator pedal signal with respect to the time variable t;

[0108] Calculate the phase matching error DP: DP = |ΔScl - ΔYmt|;

[0109] Calculate the correlation coefficient DX between the throttle position sensor output voltage signal and the accelerator pedal signal: where Cov(Scl,Ymt) represents the covariance between the throttle position sensor output voltage signal and the accelerator pedal signal, σScl represents the standard deviation of the throttle position sensor output voltage signal, and σYmt represents the standard deviation of the accelerator pedal signal;

[0110] Calculate the signal time series similarity anomaly coefficient: xsy = a1*DT + a2*DP + a3*(1 - DX), where xsy represents the signal time series similarity anomaly coefficient, and a1, a2, and a3 respectively represent the preset proportionality coefficients of the dynamic time warping distance, the phase matching error, and the correlation coefficient, and a1, a2, and a3 are all greater than 0;

[0111] It should be noted that the above formulas are all calculated by taking the numerical values after removing the dimensions. Common methods for removing dimensions include Min - Max normalization, Z - Score standardization, etc., which will not be elaborated here; a1, a2, and a3 are set according to the actual situation. For example, the expert weighting method is adopted, that is, experts in the relevant field are invited to determine the preset proportionality coefficients of each index through professional opinion surveys and comprehensive evaluations. For example, a1, a2, and a3 can be 0.3, 0.4, 0.3;

[0112] In the present invention, the signal timing similarity anomaly coefficient is used to measure the degree of anomaly in the timing similarity between the throttle position sensor and the accelerator pedal signal. The aim is to evaluate whether there are potential anomalies or latent faults by analyzing the timing of the two signals. This process combines information from aspects such as dynamic time warping distance, signal derivative, phase matching error, and signal correlation, providing a more refined anomaly detection ability. First, the voltage signal output by the throttle position sensor and the accelerator pedal signal are synchronously collected from the on-vehicle diagnostic system. These signals form time series, respectively reflecting the real-time changes of the throttle and the accelerator pedal. Based on these signals, a distance matrix is constructed to measure the difference between the throttle position sensor and the accelerator pedal signal. Each element of the matrix represents the gap between the throttle position and the accelerator pedal signal at a certain time point. Next, the minimum cumulative distance between the two signal sequences is calculated. Specifically, starting from the starting point (1,1), the minimum cumulative path from each point to the end point (D,F) is calculated step by step, and the optimal alignment path is determined through the backtracking process. The minimum cumulative distance reflects the overall time alignment between the signals. Further, the derivative sequences of the throttle position sensor and the accelerator pedal signal are calculated. The derivative of a signal represents the rate of change of the signal, which is crucial in analyzing the dynamic characteristics of the signal. By calculating the derivative sequences of the two signals, we can obtain the matching degree of the signals in instantaneous changes. The calculation of the derivative helps to more accurately capture the time deviation between the signals, especially in the minor fluctuations of the signals. Next, the phase matching error is calculated. The phase matching error is an important indicator for measuring the matching degree of signals on the time axis. By comparing the derivative sequences and calculating the phase error between the two signals, it can be reflected whether there is a timing inconsistency between them. If the change rates of the signals are inconsistent, it indicates that there may be anomalies in the coordinated operation between the throttle and the accelerator pedal within a certain time period, which may lead to problems such as vehicle response lag or insensitivity. The correlation coefficient between the throttle position sensor signal and the accelerator pedal signal is calculated. The correlation coefficient measures the linear relationship between the two signals, and the covariance reflects the common degree of change of the two signals. A higher correlation coefficient indicates that the change trends of the two signals are highly consistent, otherwise, there may be inconsistent situations. In analyzing the similarity of timing signals, the correlation coefficient is an important indicator to help judge whether the coordinated action between the signals is normal. Finally, based on the dynamic time warping distance, phase matching error, and correlation coefficient, the signal timing similarity anomaly coefficient is calculated. Considering the non-linear alignment between the signals, the matching situation of instantaneous changes, and the overall coordinated action, it provides an accurate quantitative tool for judging the timing anomaly of the throttle and accelerator pedal signals. The larger the value of this coefficient, the stronger the inconsistency between the signal timings, and there may be anomalies.On the contrary, a smaller coefficient indicates that the signal timings are basically consistent and the system is operating normally. Through this series of calculations, the abnormal degree of the timing similarity between the throttle position sensor and the accelerator pedal signal can be accurately evaluated, thereby identifying potential faults or performance degradation at an early stage. The introduction of the signal timing similarity abnormal coefficient provides a more accurate means for fault diagnosis, especially for those fault types that do not trigger traditional fault codes but have begun to show gradual performance changes, which can effectively improve the intelligence and fault tolerance of the vehicle fault remote diagnosis system.

[0113] Step S4, construct a latent gradual fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signal and the signal timing similarity abnormal coefficient, and output a latent gradual fault evaluation index to evaluate the latent gradual fault degree at the initial stage of resistor film wear.

[0114] Construct a latent gradual fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signal and the signal timing similarity abnormal coefficient, and output a latent gradual fault evaluation index yxj. The formula on which the model is based is as follows: yxj = b1 * cxs + b2 * xsy, where b1 and b2 respectively represent the preset proportional coefficients of the joint entropy fluctuation coefficient of the sensing signal and the signal timing similarity abnormal coefficient, and both b1 and b2 are greater than 0.

[0115] It should be noted that the above formulas are all calculated by taking the numerical values after dimensionless processing. Common methods for removing dimensions include Min - Max normalization, Z - Score standardization, etc., which will not be elaborated here; b1 and b2 are set according to the actual situation. For example, the expert weighting method is adopted, that is, experts in the relevant field are invited to determine the preset proportional coefficients of each index through professional opinion surveys and comprehensive evaluations. For example, b1 and b2 can be 0.5 and 0.5.

[0116] From the above calculation expressions, the larger the joint entropy fluctuation coefficient of the sensing signal and the larger the signal timing similarity abnormal coefficient, the larger the latent gradual fault evaluation index, indicating that the latent gradual fault of the throttle position sensor is more serious, indicating a higher degree of wear of the resistor film, and there may be more performance degradation or potential fault hazards in the system. On the contrary, the smaller the joint entropy fluctuation coefficient of the sensing signal and the smaller the signal timing similarity abnormal coefficient, the smaller the latent gradual fault evaluation index, indicating that the correlation between sensor signals is stable, the timing similarity is high, the working state of the system is good, the wear of the resistor film is at a low level, and no serious performance deterioration has occurred.

[0117] Compare the latent gradual fault evaluation index with the preset latent gradual fault evaluation index threshold to classify the latent gradual fault degree at the initial stage of resistor film wear, as follows:

[0118] If the implicit gradual fault evaluation index is greater than the implicit gradual fault evaluation index threshold, it indicates that the degree of implicit gradual fault in the initial stage of resistor film wear is relatively serious, and a high degree of implicit gradual fault is generated; if the implicit gradual fault evaluation index is less than or equal to the implicit gradual fault evaluation index threshold, it indicates that the degree of implicit gradual fault in the initial stage of resistor film wear is within the controllable range, and a low degree of implicit gradual fault is generated;

[0119] Step S5, perform cluster analysis on the degrees of implicit gradual faults at different time points to give early warnings for the implicit gradual faults in the initial stage of resistor film wear;

[0120] Whenever a high degree of implicit gradual fault is generated, obtain the time series constructed from the implicit gradual fault evaluation indices output by the implicit gradual fault evaluation model, perform cluster analysis on the time series of the implicit gradual fault evaluation indices, and give early warnings for the implicit gradual faults in the initial stage of resistor film wear according to the results of the cluster analysis, specifically as follows:

[0121] Step S51, determine the number R of cluster centers of the K-means algorithm according to the silhouette coefficient method, and randomly select R implicit gradual fault evaluation indices from the time series of the implicit gradual fault evaluation indices as the initial cluster centers;

[0122] Step S52, calculate the Euclidean distance between each implicit gradual fault evaluation index in the time series of the implicit gradual fault evaluation indices and the cluster centers, and assign each implicit gradual fault evaluation index to the cluster center with the closest Euclidean distance;

[0123] Step S53, calculate the average value of the implicit gradual fault evaluation indices in each cluster center and use it as the new cluster center coordinates;

[0124] Step S54, repeat Step S52 and Step S53 to iteratively update the cluster centers until the cluster centers no longer change;

[0125] According to the finally obtained multiple cluster centers, calculate the differences between the coordinates of different cluster centers, and accumulate the differences between the coordinates of different cluster centers to obtain the difference cumulative value;

[0126] Compare the difference cumulative value with the preset difference cumulative value threshold to give early warnings for the implicit gradual faults in the initial stage of resistor film wear, specifically as follows:

[0127] If the difference cumulative value is greater than the difference cumulative value threshold, it indicates that the degree of implicit gradual fault in the initial stage of resistor film wear is continuously deepening, and the early warning mechanism should be triggered to generate an immediate early warning signal;

[0128] If the cumulative difference value is less than or equal to the cumulative difference value threshold, it indicates that the difference between the clustering centers is small, meaning that the degree of latent gradual change fault has not increased significantly, and the latent gradual change fault in the initial stage of resistor film wear may be in a relatively stable state. At this time, the fault change trend is relatively slow, and the system can temporarily not give an early warning and continue to monitor the latent gradual change state to ensure timely response when the fault intensifies;

[0129] The present invention obtains the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system, and through time-frequency domain analysis, effectively identifies the tiny voltage drift or fluctuation caused by the resistor film in the initial stage of wear, avoiding the problem that traditional diagnostic methods cannot detect these subtle changes, ensuring the effectiveness of early warning. By calculating the joint entropy fluctuation coefficient and the signal time series similarity anomaly coefficient, it more comprehensively evaluates the health status of the throttle sensor, improves the accuracy of latent gradual change fault diagnosis, avoids the limitation of single signal analysis, better captures the early signs of sensor degradation or failure. By constructing a latent gradual change fault evaluation model and outputting an evaluation index, it can effectively quantify and evaluate the latent gradual change fault in the initial stage of resistor film wear. Compared with the traditional diagnostic method triggered by fault codes, this method can identify even the tiny changes that do not exceed the ECU diagnostic threshold, and thus discover potential latent faults in advance, which plays an important role in improving the fault tolerance, reliability and early warning ability of the vehicle remote fault diagnosis system. By performing clustering analysis on the degree of latent gradual change fault at different time points, it can dynamically monitor the evolution trend of the fault and give an early warning in time. This early warning mechanism is not only effective in the initial stage of resistor film wear, but also can continuously track the progress of the fault to ensure that the fault is processed in time before it occurs, avoiding the detection lag of the traditional fault code-triggered method, and providing a more efficient and sensitive means of fault identification and prevention.

[0130] Embodiment 2: This embodiment introduces an automotive fault remote diagnosis system, as Figure 2 shown, which includes a wear recognition module, an information coupling perception module, a time series similarity analysis module, a latent gradual change fault evaluation module, and a latent gradual change early warning module;

[0131] The wear recognition module is used to obtain the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system and perform time-frequency domain joint analysis on it to identify the initial wear state of the resistor film;

[0132] The information coupling perception module is used to obtain the joint entropy information of the signals of the throttle position sensor and other on-vehicle sensors when the resistor film is in the initial wear state, and calculate the sensing signal joint entropy fluctuation coefficient according to the joint entropy information of the throttle position sensor and other on-vehicle sensors;

[0133] The timing similarity analysis module, when the resistive film is in the initial wear state, obtains the timing similarity information of the throttle position sensor and the accelerator pedal signal, and calculates the signal timing similarity anomaly coefficient according to the timing similarity information of the throttle position sensor signal and the accelerator pedal signal;

[0134] The latent gradual change fault evaluation module constructs a latent gradual change fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signal and the signal timing similarity anomaly coefficient, outputs the latent gradual change fault evaluation index, and evaluates the degree of the latent gradual change fault in the initial wear stage of the resistive film;

[0135] The latent gradual change warning module performs clustering analysis on the degrees of the latent gradual change faults at different time points, and warns of the latent gradual change faults in the initial wear stage of the resistive film;

[0136] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0137] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that contains one or more collections of available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0138] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution, and the order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0139] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system and method can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0140] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways.

[0141] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application, and all should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A remote diagnosis method for vehicle faults, characterized in that: It includes the following steps: Step S1: Obtain the output voltage signal of the throttle position sensor through the on-vehicle diagnostic system, and perform a joint time-frequency domain analysis on it to identify the initial wear state of the resistance film; Step S2: When the resistance film is in the initial wear state, obtain the joint entropy information of the signals of the throttle position sensor and other on-vehicle sensors, and calculate the joint entropy fluctuation coefficient of the sensing signals according to the joint entropy information of the throttle position sensor and other on-vehicle sensors; Step S3: When the resistance film is in the initial wear state, obtain the temporal similarity information of the signals of the throttle position sensor and the accelerator pedal signal, and calculate the signal temporal similarity anomaly coefficient according to the temporal similarity information of the throttle position sensor signal and the accelerator pedal signal; Step S4: Construct a hidden gradual fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signals and the signal temporal similarity anomaly coefficient, output a hidden gradual fault evaluation index, and evaluate the degree of the hidden gradual fault in the initial wear stage of the resistance film; Step S5: Perform cluster analysis on the degrees of the hidden gradual faults at different time points to give an early warning of the hidden gradual faults in the initial wear stage of the resistance film.

2. The remote diagnostic method for vehicle faults according to claim 1, characterized in that: In Step S1, the output voltage signal of the throttle position sensor is collected in real time through the on-vehicle diagnostic system, and the output voltage signal is subjected to filtering and denoising processing; Perform a joint time-frequency domain analysis on the output voltage signal of the throttle position sensor after filtering and denoising to identify the initial wear state of the resistance film, specifically as follows: Select a wavelet basis function to perform wavelet packet transform on the output voltage signal Scl(t). After the transform, multiple decomposed sub-signals are obtained. Each sub-signal corresponds to a different frequency band. Mark the decomposed sub-signals as C j (t), where j = {1, 2,..., 2 J} represents different frequency bands, J is the number of transform levels, and each C j (t) contains signal components with different frequency ranges; Perform feature matching on the decomposed sub-signals and the preset characteristic signals of the resistor film wear, and select a group of sub-signals \(C_{k}(t)\) where the matching degree is greater than the matching degree threshold, where \(k = \{1, 2, \cdots, K\}\) and \(K\) is the total number of selected sub-signals; k (t), where \(k=\{1,2,\cdots,K\}\), and \(K\) is the total number of selected sub-signals; For each selected sub-signal C k (t), calculate its energy E k , and the calculation formula is: E k = ∑ t |C k (t)| 2 , where E k represents the energy of the sub-signal C k (t); Calculated sub-signal C k (t) and the energy ratio of sub-signal C l (t) is calculated as follows: where Rn kl represents the energy ratio of sub-signal C k (t) and sub-signal C l (t), E k represents the energy of sub-signal C k (t), E l represents the energy of sub-signal C l (t), k, l ∈ {1, 2,..., K}.

3. The remote diagnosis method for vehicle faults according to claim 2, characterized in that: Compare the energy ratio between different sub-signals with a preset energy ratio threshold. If the energy ratio is greater than the energy ratio threshold, accumulate the energy ratios greater than the energy ratio threshold to obtain a wear energy fluctuation coefficient; Compare the wear energy fluctuation coefficient with a preset wear energy fluctuation coefficient threshold. If the wear energy fluctuation coefficient is greater than the wear energy fluctuation coefficient threshold, mark the current state of the resistance film as the initial wear state.

4. A remote diagnosis method for vehicle faults according to claim 1, characterized in that: The acquisition logic of the joint entropy fluctuation coefficient of the sensing signals is as follows: The other on-vehicle sensors mentioned above include a mass air flow sensor and a manifold absolute pressure sensor; Obtain the output signal Scl(t) of the throttle position sensor, the output signal Zlk(t) of the mass air flow sensor, and the output signal Qgy(t) of the manifold absolute pressure sensor, and construct a multi-dimensional signal sequence: SS(t) = {Scl(t), Zlk(t), Qgy(t)}; Preset multiple sliding window scales {τ1, τ2,..., τ i}, each scale representing a different time granularity, and perform the following operations for each scale: Operation A1: Within each sliding window, construct a joint probability density distribution using kernel density estimation: denotes the joint probability density calculated for any sliding window scale τ i ; denotes the joint probability density of the sliding window scale τ i ; where YB represents the total number of sampling points of the sliding window scale τ i , DK represents the bandwidth parameter, HS represents the Gaussian kernel function, and XL j represents the multi-dimensional sensor output signal vector of the j-th sampling point denotes the mean of the multi-dimensional sensor output signal vectors within the sliding window scale τ i ; Operation A2: Calculate the local joint entropy: where represents the local joint entropy of the sliding window scale τ i ; Calculate the sliding window scale τ i and the sliding window scale τ n of the joint entropy difference: where represents the joint entropy difference value of the sliding window scale τ i and the sliding window scale τ n ; represents the local joint entropy of the sliding window scale τ n ; All pairs of sliding window scales are combined to obtain the perturbation spectrum DT: where i, n ∈ {1, 2,..., N}, and N is the total number of sliding window scales; Extract the disturbance mean from the disturbance spectrum DT: where μΔLS represents the disturbance mean; Extract the disturbance standard deviation from the disturbance map DT: where σΔLS represents the disturbance standard deviation; Calculate the joint entropy fluctuation coefficient of the sensing signal: where cxs represents the joint entropy fluctuation coefficient of the sensing signal.

5. A remote diagnosis method for vehicle faults according to claim 1, characterized in that: The acquisition logic of the signal temporal similarity anomaly coefficient is as follows: Synchronously collect the output voltage signal of the throttle position sensor and the accelerator pedal signal from the on-vehicle diagnostic system to form a time series, and extract the output voltage signal of the throttle position sensor and the accelerator pedal signal to construct a distance matrix DD of D×F: DD(d,f) = |Scl d -Ymt f |, where DD(d,f) is the element in the d-th row and f-th column of the distance matrix DD, Scl d represents the d-th output voltage signal of the throttle position sensor extracted, Ymt f represents the f-th accelerator pedal signal extracted, d = {1, 2,..., D}, D is the total number of sampling points of the output voltage signal of the throttle position sensor, f = {1, 2,..., F}, F is the total number of sampling points of the accelerator pedal signal; Calculate the minimum cumulative distance from the starting point (1, 1) to the ending point (D, F) through dynamic programming: where C(d, f) represents the minimum cumulative distance from the starting point to the point (d, f); Retrace from the end point (D, F) to the starting point (1, 1) in the reverse direction, and select the smallest forward path during the retracing process to obtain the dynamic time warping distance DT between the output voltage signal of the throttle position sensor and the accelerator pedal signal: DT = C(D, F), where C(D, F) is the total minimum cumulative distance; Calculate the derivative sequences of the output voltage signal of the throttle position sensor and the accelerator pedal signal: where ΔScl represents the derivative of the output voltage signal of the throttle position sensor with respect to the time variable t, and ΔYmt represents the derivative of the accelerator pedal signal with respect to the time variable t; Calculate the phase matching error DP: DP = |ΔScl - ΔYmt|; Calculate the correlation coefficient DX between the output voltage signal of the throttle position sensor and the accelerator pedal signal: where Cov(Scl, Ymt) represents the covariance between the output voltage signal of the throttle position sensor and the accelerator pedal signal, σScl represents the standard deviation of the output voltage signal of the throttle position sensor, and σYmt represents the standard deviation of the accelerator pedal signal; Calculate the signal temporal similarity anomaly coefficient: xsy = a1 * DT + a2 * DP + a3 * (1 - DX), where xsy represents the signal temporal similarity anomaly coefficient, and a1, a2, and a3 respectively represent the preset proportional coefficients of the dynamic time warping distance, the phase matching error, and the correlation coefficient, and a1, a2, and a3 are all greater than 0.

6. The remote diagnosis method for vehicle faults according to claim 1, characterized in that: Construct a hidden gradual fault evaluation model based on the joint entropy fluctuation coefficient and signal time series similarity anomaly coefficient of the sensing signal, and output the hidden gradual fault evaluation index yxj. The formula on which the model is based is as follows: yxj = b1 * cxs + b2 * xsy, where b1 and b2 respectively represent the preset proportionality coefficients of the joint entropy fluctuation coefficient and signal time series similarity anomaly coefficient of the sensing signal, and both b1 and b2 are greater than 0.

7. A remote diagnosis method for vehicle faults according to claim 6, characterized in that: Compare the hidden gradual fault evaluation index with the preset hidden gradual fault evaluation index threshold to classify the degree of hidden gradual fault in the initial stage of resistor film wear, as follows: If the hidden gradual fault evaluation index is greater than the hidden gradual fault evaluation index threshold, a high hidden gradual fault degree is generated; if the hidden gradual fault evaluation index is less than or equal to the hidden gradual fault evaluation index threshold, a low hidden gradual fault degree is generated.

8. A remote diagnosis method for vehicle faults according to claim 7, characterized in that: Whenever a high hidden gradual fault degree is generated, obtain the time series of the hidden gradual fault evaluation index output by the hidden gradual fault evaluation model, perform cluster analysis on the time series of the hidden gradual fault evaluation index, and give an early warning of the hidden gradual fault in the initial stage of resistor film wear according to the results of the cluster analysis, as follows: Step S51, determine the number of K-means algorithm cluster centers R according to the silhouette coefficient method, and randomly select R hidden gradual fault evaluation indexes from the time series of the hidden gradual fault evaluation index as the initial cluster centers; Step S52, calculate the Euclidean distance between each hidden gradual fault evaluation index in the time series of the hidden gradual fault evaluation index and the cluster center, and assign each hidden gradual fault evaluation index to the cluster center with the closest Euclidean distance; Step S53, calculate the average value of the hidden gradual fault evaluation indexes in each cluster center and use it as the new cluster center coordinates; Step S54, repeat steps S52 and S53 to iteratively update the cluster centers until the cluster centers no longer change; According to the finally obtained multiple cluster centers, calculate the difference between the coordinates of different cluster centers, and accumulate the differences between the coordinates of different cluster centers to obtain the difference cumulative value.

9. A method for remote diagnosis of vehicle faults according to claim 8, characterized in that: Compare the difference cumulative value with the preset difference cumulative value threshold to give an early warning of the hidden gradual fault in the initial stage of resistor film wear, as follows: If the difference cumulative value is greater than the difference cumulative value threshold, an immediate warning signal is generated; If the difference cumulative value is less than or equal to the difference cumulative value threshold, no warning is given temporarily, and the hidden gradual state is continuously monitored.

10. A remote vehicle fault diagnosis system for implementing a remote vehicle fault diagnosis method according to any one of claims 1-9, characterized in that: It includes a wear recognition module, an information coupling perception module, a time series similarity analysis module, a hidden gradual fault evaluation module, and a hidden gradual warning module; The wear recognition module is used to obtain the output voltage signal of the throttle position sensor through the on-board diagnostic system and perform joint time-frequency domain analysis on it to identify the initial wear state of the resistor film; The information coupling perception module is used to obtain the joint entropy information of the throttle position sensor and other on-board sensors when the resistor film is in the initial wear state, and calculate the joint entropy fluctuation coefficient of the sensing signal according to the joint entropy information of the throttle position sensor and other on-board sensors; The timing similarity analysis module, when the resistive film is in the initial wear state, obtains the timing similarity information of the throttle position sensor and the accelerator pedal signal, and calculates the signal timing similarity anomaly coefficient according to the timing similarity information of the throttle position sensor signal and the accelerator pedal signal; The latent gradual change fault evaluation module constructs a latent gradual change fault evaluation model based on the joint entropy fluctuation coefficient of the sensing signal and the signal timing similarity anomaly coefficient, outputs the latent gradual change fault evaluation index, and evaluates the latent gradual change fault degree in the initial wear stage of the resistive film; The latent gradual change warning module performs cluster analysis on the latent gradual change fault degrees at different time points, and warns of the latent gradual change faults in the initial wear stage of the resistive film.

Images