Vehicle fault and timely warning method based on multi-source data analysis

Abstract

The present application relates to the field of vehicle fault detection, and more particularly to a vehicle fault timely warning method based on multi-source data analysis, which comprises: obtaining actual acceleration and theoretical acceleration benchmark by collecting multi-source driving data and performing dynamics calculation; obtaining transmission elastic hysteresis compensation factor by analyzing acceleration difference energy and relative deviation; obtaining power decoupling singularity check factor by checking singularity of theoretical and actual acceleration change direction; obtaining final fault discrimination index for fault threshold determination by weighting processing of base residual error through transmission elastic hysteresis compensation factor and power decoupling singularity check factor; obtaining vehicle drive system fault warning result by time window statistics and threshold determination of final fault discrimination index, thereby solving the problem that rigid dynamics model cannot distinguish transmission elastic hysteresis pseudo residual error from power decoupling real fault under high dynamic working condition, and is prone to false alarm and omission.

Description

Technical Field

[0001] This invention relates to the field of vehicle fault detection technology, and in particular to a method for timely early warning of vehicle faults based on multi-source data analysis. Background Technology

[0002] In the fields of modern vehicle engineering and driving safety monitoring, online fault monitoring and timely early warning of vehicle drive systems have become crucial components of vehicle control strategies and active safety system design. With the widespread application of engine electronic control units, transmission control units, electronic stability systems, and various onboard sensors, vehicles can acquire real-time multi-source driving data during operation, including engine or drive motor output torque, current transmission ratio, final drive ratio, vehicle speed, road gradient, and longitudinal acceleration. Based on this, existing technologies generally employ residual analysis methods based on physical models. These models construct a full-element rigid dynamic model using the principles of vehicle longitudinal dynamics as a reference. By combining calibration parameters such as the vehicle's air resistance coefficient, rolling resistance coefficient, frontal area, vehicle mass, mechanical efficiency, tire rolling radius, and rotational mass coefficient, the dynamic equations are used to calculate the theoretically expected acceleration of the vehicle under the current driving force. Simultaneously, the onboard inertial measurement unit collects the vehicle's actual longitudinal acceleration, and the absolute error between the actual acceleration and the theoretical acceleration prediction is used as the residual criterion. When the residual continuously exceeds a preset safety threshold within a certain period, it is determined that the vehicle drive system may have faults such as slippage, power loss, or sensor failure.

[0003] However, residual analysis methods based on full-element rigid dynamics models exhibit significant limitations when dealing with the nonlinear and time-varying dynamic characteristics of vehicle drive systems. On the one hand, existing reference models are generally based on rigid body assumptions, assuming that the generation of driving torque and the establishment of vehicle acceleration are synchronous on the time axis, that is, assuming that torque changes can be instantly converted into changes in the longitudinal acceleration of the whole vehicle. However, the actual vehicle transmission system consists of multi-body elastic damping components such as dual-mass flywheels, clutches, drive shafts, half shafts, and tires. Under high-dynamic excitation conditions such as rapid acceleration, frequent gear shifts, and hill starts, the driving energy must first overcome the gaps and elastic deformations in the transmission chain and accumulate as elastic potential energy. At the same time, it must also overcome the inertia of each rotating component before it can be gradually converted into the translational kinetic energy of the whole vehicle. This results in a significant phase lag in the actual acceleration response in the time domain compared to the theoretical acceleration prediction value of the rigid model. This time-domain mismatch caused by physical hysteresis in the transmission system can generate a large residual in vehicle acceleration even when there is no real fault in the system. This residual can be incorrectly identified as slippage or insufficient power by existing amplitude threshold-based judgment logic, leading to false alarms and frequent alarms. On the other hand, to suppress the spurious residuals caused by physical hysteresis, engineering practice often uses static amplitude suppression methods such as low-pass filtering, moving average, and fixed-weight attenuation on the residual signal to smooth and weaken the large short-term residuals. Although this kind of processing reduces false over-limits caused by physical hysteresis to a certain extent, it also introduces new risks: when the vehicle experiences typical power decoupling faults such as clutch slippage under high load, power interruption during gear shifting, or failure of drive motor torque output, the control system still generates a high theoretical acceleration prediction value based on the torque command, while the actual wheel acceleration shows a significant decrease or even the opposite trend to the theoretical acceleration. That is, the theoretical acceleration points to acceleration while the actual acceleration points to deceleration or remains unchanged. At this time, the residual not only increases significantly in amplitude but also shows a clear directional deviation in the direction of change. If existing technologies only perform amplitude filtering or static weight reduction on the residuals without distinguishing and strengthening the information on the direction of change, they are very likely to smooth out the characteristics of such decoupled faults, masking the power transmission anomalies that have a significant impact on driving safety, resulting in a decrease in fault detection rate and an increase in the risk of missed alarms. It is difficult to balance the suppression of false alarms and the sensitivity to real faults under dynamic operating conditions. Summary of the Invention

[0004] In view of this, the present invention aims to propose a timely early warning method for vehicle faults based on multi-source data analysis, so as to solve the problem that rigid dynamic models cannot distinguish between transmission elastic hysteresis pseudo residuals and real faults of dynamic decoupling under high dynamic conditions, and are prone to false alarms and missed alarms.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0006] A method for timely early warning of vehicle faults based on multi-source data analysis, the method comprising:

[0007] Step S1: Obtain the actual acceleration and theoretical acceleration benchmark by collecting and performing dynamic calculations on multi-source driving data, and obtain the basic residual of vehicle acceleration based on the difference between the actual acceleration and the theoretical acceleration benchmark;

[0008] Step S2: Obtain the transmission elastic hysteresis compensation factor by analyzing the acceleration differential energy and relative deviation;

[0009] Step S3: Obtain the dynamic decoupling singularity verification factor by performing singularity verification on the directionality of theoretical and actual acceleration changes;

[0010] Step S4: The basic residual is weighted by the transmission elastic hysteresis compensation factor and the power decoupling singularity verification factor to obtain the final fault discrimination index for fault threshold determination;

[0011] Step S5: Obtain vehicle drive system fault warning results by performing time window statistics and threshold determination on the final fault discrimination indicators.

[0012] Furthermore, the step of acquiring actual acceleration and theoretical acceleration benchmarks by collecting and performing dynamic calculations on multi-source driving data, and obtaining the basic residual of vehicle acceleration based on the difference between the actual acceleration and the theoretical acceleration benchmark, includes:

[0013] By installing an inertial measurement unit on the vehicle body near the center of gravity of the vehicle and setting a sampling frequency, the inertial measurement unit is continuously triggered to collect data during the vehicle's driving process according to the sampling frequency, thereby obtaining actual acceleration time series data that characterizes the longitudinal driving state of the vehicle, and using the actual acceleration time series data as the actual acceleration for the calculation of the basic residual.

[0014] The system collects actual output torque data of the engine or drive motor from the engine management system through the vehicle's controller area network, current gear ratio data and final drive ratio data from the transmission controller, and vehicle speed data and road slope data from the electronic stability system. The collected actual output torque data, current gear ratio data, final drive ratio data, vehicle speed data and road slope data are used as multi-source driving data for theoretical acceleration benchmark calculation.

[0015] The vehicle electronic control unit reads pre-calibrated vehicle calibration parameters from the on-board storage medium. These parameters include at least the rolling radius of the vehicle's drive wheels, vehicle mass, drag coefficient, frontal area, mechanical efficiency, rolling resistance coefficient, air density, and rotational mass coefficient. The vehicle calibration parameters are then combined with the multi-source driving data and input into the all-element vehicle longitudinal dynamics equation. Based on the all-element vehicle longitudinal dynamics equation, theoretical acceleration time-series data corresponding to the actual acceleration time-series data are calculated at each sampling time. This theoretical acceleration time-series data is then used as the theoretical acceleration reference.

[0016] For each target sampling time in the actual acceleration time series data, the absolute value of the difference between the actual acceleration at the target sampling time and the theoretical acceleration benchmark at the corresponding target sampling time is calculated, and the obtained absolute value of the difference is used as the basic residual corresponding to the target sampling time.

[0017] Furthermore, the step of obtaining the transmission elastic hysteresis compensation factor by analyzing the acceleration differential energy and relative deviation includes:

[0018] Acceleration differential energy reference data is obtained by performing first-order difference operations on the time series data of actual acceleration and theoretical acceleration; transmission elastic hysteresis compensation factor is obtained by performing relative deviation normalization and nonlinear penalty processing on the acceleration differential energy reference data.

[0019] Furthermore, the step of obtaining acceleration differential energy reference data by performing a first-order difference operation on the time-series data of actual acceleration and theoretical acceleration includes:

[0020] For any target sampling time in the actual acceleration time series data, excluding the first sampling time, the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time are extracted from the actual acceleration time series data. The difference between the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time is taken as the first-order differential change of the actual acceleration at the target sampling time. The square of the first-order differential change of the actual acceleration is taken as the actual acceleration differential energy assessment at the target sampling time.

[0021] For any target sampling time in the theoretical acceleration time series data, excluding the first sampling time, the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time are extracted from the theoretical acceleration time series data. The difference between the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time is taken as the first-order difference change of theoretical acceleration at the target sampling time. The square of the first-order difference change of theoretical acceleration is taken as the theoretical acceleration difference energy assessment at the target sampling time.

[0022] The set of actual acceleration differential energy assessments and theoretical acceleration differential energy assessments corresponding to all target sampling times is used as the acceleration differential energy benchmark data.

[0023] Furthermore, the step of obtaining the transmission elastic hysteresis compensation factor by performing relative deviation normalization and nonlinear penalty processing on the acceleration differential energy reference data includes:

[0024] For any target sampling time, extract the actual acceleration differential energy assessment and theoretical acceleration differential energy assessment corresponding to the target sampling time from the acceleration differential energy reference data. Extract the actual first-order differential change of acceleration, the theoretical first-order differential change of acceleration, and the theoretical acceleration value corresponding to the target sampling time from the actual acceleration time series data and the theoretical acceleration time series data. The calculation result of adding the preset robust constant to prevent the denominator from being zero to the actual acceleration differential energy assessment is used as the first energy correction assessment for the target sampling time.

[0025] The difference between the theoretical first-order difference change of acceleration and the actual first-order difference change of acceleration at the target sampling time is used as the acceleration difference deviation evaluation. The acceleration difference deviation evaluation is used as the numerator, and the sum of the absolute value of the theoretical acceleration value at the target sampling time and the robust constant is used as the denominator. The corresponding fractional acceleration difference relative deviation at the target sampling time is normalized and evaluated. The sum of the constant 1 and the square of the normalized evaluation of the acceleration difference relative deviation at the target sampling time is used as the nonlinear deviation correction term at the target sampling time.

[0026] The product of the theoretical acceleration differential energy assessment and the nonlinear deviation correction term is added to the actual acceleration differential energy assessment, and then added to the robust constant. The result is used as the second energy correction assessment at the target sampling time.

[0027] The result of dividing the first energy correction assessment by the second energy correction assessment is used as the transmission elastic hysteresis compensation factor corresponding to the target sampling time.

[0028] Furthermore, the method of obtaining the dynamic decoupling singularity verification factor by performing singularity verification on the directionality of theoretical and actual acceleration changes includes:

[0029] By performing first-order difference and directional combination processing on the time series data of actual acceleration and theoretical acceleration, basic data for dynamic decoupling directional discrimination are obtained.

[0030] By performing robustness processing on the sum-difference energy ratio and product of the basic data for dynamic decoupling directionality discrimination, the singularity characteristic data of dynamic decoupling is obtained.

[0031] By applying energy percentage gating and load gain processing to the dynamic decoupling singularity characteristic data, a dynamic decoupling singularity verification factor is obtained.

[0032] Furthermore, the process of obtaining basic data for dynamic decoupling directionality discrimination by performing first-order difference and directional combination processing on the time series data of actual acceleration and theoretical acceleration includes:

[0033] For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the sum of the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the combination and evaluation of the acceleration change direction at the target sampling time. The difference between the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the combination difference evaluation of the acceleration change direction at the target sampling time. The product of the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the product evaluation of the acceleration change direction at the target sampling time. The set of the combination and evaluation of the acceleration change direction, the combination difference evaluation of the acceleration change direction, and the product evaluation of the acceleration change direction corresponding to all target sampling times is used as the basic data for dynamic decoupling directionality discrimination.

[0034] Furthermore, the process of obtaining basic data for dynamic decoupling directionality discrimination by performing first-order difference and directional combination processing on the time series data of actual acceleration and theoretical acceleration includes:

[0035] For any target sampling moment in the actual acceleration time series data and theoretical acceleration time series data, extract the acceleration change direction combination and evaluation, acceleration change direction combination difference evaluation, and acceleration change direction product evaluation corresponding to the target sampling moment from the dynamic decoupling directionality discrimination basic data. Use the square of the acceleration change direction combination difference evaluation as the direction difference energy evaluation of the target sampling moment, use the square of the acceleration change direction combination and evaluation as the direction and energy evaluation of the target sampling moment, and use the absolute value of the acceleration change direction product evaluation as the absolute value evaluation of the direction product of the target sampling moment.

[0036] The calculation result of adding the direction and energy assessment, the absolute value of the product of directions, and a preset robust constant to prevent the denominator from being zero is used as the robust denominator assessment of the sum-difference energy ratio at the target sampling time. The direction difference energy assessment is used as the robust numerator assessment of the sum-difference energy ratio. The quotient of the robust numerator assessment and the robust denominator assessment of the sum-difference energy ratio is used as the robust assessment of the sum-difference energy ratio at the target sampling time. The square of the robust assessment of the sum-difference energy ratio is used as the dynamic decoupling singularity feature assessment corresponding to the target sampling time. The set of dynamic decoupling singularity feature assessments corresponding to all target sampling times is used as the dynamic decoupling singularity feature data.

[0037] Furthermore, the step of obtaining a dynamic decoupling singularity verification factor by performing energy proportion gating and load gain processing on the dynamic decoupling singularity characteristic data includes:

[0038] For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the dynamic decoupling singularity feature assessment corresponding to the target sampling time is extracted from the dynamic decoupling singularity feature data. The theoretical acceleration value and the first-order difference change of theoretical acceleration corresponding to the target sampling time are extracted from the theoretical acceleration time series data. The square of the theoretical acceleration value corresponding to the target sampling time is used as the theoretical acceleration energy assessment, and the square of the first-order difference change of theoretical acceleration corresponding to the target sampling time is used as the theoretical acceleration rate of change energy assessment. The theoretical acceleration energy assessment is used as the numerator, and the sum of the theoretical acceleration energy assessment, the theoretical acceleration rate of change energy assessment, and the preset robust constant to prevent the denominator from being zero is used as the denominator. The resulting fraction is used as the energy proportion gating coefficient assessment for the target sampling time. The absolute value of the theoretical acceleration value corresponding to the target sampling time is used as the load gain assessment. The product of the dynamic decoupling singularity feature assessment, the energy proportion gating coefficient assessment, and the load gain assessment is added to the constant 1, and the result is used as the dynamic decoupling singularity check factor corresponding to the target sampling time.

[0039] Furthermore, the step of obtaining the final fault discrimination index for fault threshold determination by weighting the basic residual through the transmission elastic hysteresis compensation factor and the dynamic decoupling singularity verification factor includes:

[0040] For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the calculation result of multiplying the basic residual evaluation, transmission elastic hysteresis compensation factor evaluation and dynamic decoupling singularity verification factor evaluation corresponding to the target sampling time is used as the single-point evaluation of the final fault discrimination index corresponding to the target sampling time. The set of single-point evaluations of the final fault discrimination index corresponding to all target sampling times is used as the final fault discrimination index for fault threshold determination.

[0041] Compared with the prior art, the present invention has the following advantages:

[0042] The vehicle fault timely early warning method based on multi-source data analysis described in this invention dynamically weights and adjusts the basic residual between the theoretical acceleration obtained from the full-element dynamic model and the actual acceleration collected by the on-board inertial sensor by constructing a transmission elastic hysteresis compensation factor and a power decoupling singularity verification factor. This prevents the residual amplitude from being mechanically constrained by a single fixed threshold, but rather adaptively adjusts according to the energy transfer state of the transmission system and the direction of acceleration change. Under high-dynamic conditions such as rapid acceleration, frequent gear shifting, and starting on long slopes, the vehicle transmission chain inevitably exhibits multibody elastic and inertial hysteresis. This invention utilizes the first-order differential energy of acceleration and its relative deviation to construct a nonlinear attenuation mechanism, which automatically compresses the non-faulty pseudo-residual caused by elastic hysteresis. This significantly reduces the false alarm rate of traditional rigid model residual diagnosis methods under complex conditions, and improves the alarm stability and driving experience of vehicles in scenarios such as urban congestion and mountain curves. Meanwhile, this invention verifies the singularity of the directional relationship between the theoretical and actual rates of acceleration change, introduces robust design based on the sum-difference energy ratio and product, and combines theoretical acceleration energy ratio gating and dynamic gain based on operating load. This significantly amplifies key characteristics that might otherwise be masked by hysteresis compensation in the final fault identification index when decoupling faults such as clutch slippage under high load, transmission power interruption, and drive motor torque output failure occur. Combined with a comprehensive judgment logic based on sliding time window integration and preset fault alarm thresholds, this invention can not only quickly and reliably identify power decoupling faults that significantly impact vehicle driving safety under high torque request conditions, triggering timely audible and visual warnings and issuing torque limit requests to the vehicle controller, but also exhibits good anti-interference capabilities under low load, idling vibration, and sensor noise conditions. It achieves an optimized balance between false alarm rate and false negative rate at the system level, significantly improving the safety and reliability of online fault monitoring in vehicle drive systems. Attached Figure Description

[0043] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0044] Figure 1 This is a flowchart of the method for timely early warning of vehicle faults based on multi-source data analysis according to an embodiment of the present invention. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0046] See Figure 1 This is a flowchart of a method for timely early warning of vehicle faults based on multi-source data analysis provided in Embodiment 1 of the present invention. Figure 1As shown, a vehicle fault timely warning method based on multi-source data analysis may include:

[0047] Step S1: Obtain the actual acceleration and theoretical acceleration benchmark by collecting and performing dynamic calculations on multi-source driving data, and obtain the basic residual of vehicle acceleration based on the difference between the actual acceleration and the theoretical acceleration benchmark.

[0048] First, an inertial measurement unit (IMU) is installed on the vehicle body near the vehicle's center of gravity, and a sampling frequency is set. In this embodiment, the sampling frequency is set to 100Hz, meaning data is collected every 10 milliseconds. During vehicle operation, the IMU is continuously triggered to collect data at the sampling frequency to obtain actual acceleration time-series data characterizing the vehicle's longitudinal driving state. This actual acceleration time-series data is used as the actual acceleration for basic residual calculation. Then, actual output torque data of the engine or drive motor is collected from the engine management system via the vehicle's controller area network, current gear ratio data and final drive ratio data are collected from the gearbox controller, and vehicle speed and road slope data are collected from the electronic stability control system. The collected actual output torque data, current gear ratio data, final drive ratio data, vehicle speed data, and road slope data are used as the theoretical acceleration reference. The calculation involves multi-source driving data; reading pre-calibrated vehicle calibration parameters from the vehicle's on-board storage medium via the vehicle's electronic control unit. These calibration parameters include at least the rolling radius of the vehicle's drive wheels, vehicle mass, drag coefficient, frontal area, mechanical efficiency, rolling resistance coefficient, air density, and rotational mass coefficient. The vehicle calibration parameters are then combined with the multi-source driving data and input into the all-element vehicle longitudinal dynamics equation. Based on the all-element vehicle longitudinal dynamics equation, theoretical acceleration time-series data corresponding to the actual acceleration time-series data are calculated at each sampling time. This theoretical acceleration time-series data is used as the theoretical acceleration benchmark. For each target sampling time in the actual acceleration time-series data, the absolute value of the difference between the actual acceleration at that target sampling time and the corresponding theoretical acceleration benchmark is calculated. This absolute value of the difference is used as the basic residual corresponding to that target sampling time.

[0049] Thus, the process of acquiring actual and theoretical acceleration benchmarks by collecting and performing dynamic calculations on multi-source driving data has been completed, and the basic residual of vehicle acceleration has been obtained based on the difference between the actual and theoretical acceleration benchmarks.

[0050] Step S2: Obtain the transmission elastic hysteresis compensation factor by analyzing the acceleration differential energy and relative deviation.

[0051] In transient conditions such as rapid vehicle acceleration or drastic changes in torque demand, an inherent time-domain mismatch exists between the predicted output of the rigid model and the actual response of the physical system. Specifically, while changes in engine output torque are instantly reflected in the calculated theoretical acceleration, the physical system must undergo a process of elastic potential energy accumulation. That is, the torque must first overcome the torsional clearances and elastic deformations of the driveshaft, half-shafts, and tires before actual acceleration can be established. This physical mechanism causes the rate of change of actual acceleration to lag significantly behind the rate of change of theoretical acceleration over time, and the numerical deviation between the two increases with the intensity of the torque change. Directly comparing the two will produce non-faulty spurious residuals. To solve this problem, a dynamic factor capable of sensing the energy transfer state of the system needs to be constructed. This factor should be based on the energy comparison between the theoretical and actual rates of change and introduce a nonlinear penalty mechanism sensitive to instantaneous deviations. When the theoretical energy changes drastically while the actual response is weak and accompanied by a large numerical deviation, the factor should be able to automatically identify that the system is in the elastic hysteresis region and output a decay coefficient close to zero, thereby strongly suppressing the basic residual and eliminating the risk of misjudgment caused by physical hysteresis.

[0052] In summary, this invention first obtains acceleration differential energy benchmark data by performing a first-order difference operation on the actual acceleration and theoretical acceleration time-series data. Specifically, for any target sampling time in the actual acceleration time-series data, excluding the first sampling time, the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time are extracted from the actual acceleration time-series data. The difference between the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time is taken as the first-order difference change of the actual acceleration at the target sampling time, and the square of the first-order difference change of the actual acceleration is taken as the actual acceleration differential energy benchmark at the target sampling time. For any target sampling time in the theoretical acceleration time series data, excluding the first sampling time, the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time are extracted from the theoretical acceleration time series data. The difference between the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time is taken as the first-order differential change of theoretical acceleration at the target sampling time. The square of the first-order differential change of theoretical acceleration is taken as the theoretical acceleration differential energy assessment at the target sampling time. The set of actual acceleration differential energy assessments and theoretical acceleration differential energy assessments corresponding to all target sampling times is taken as the acceleration differential energy benchmark data.

[0053] After obtaining the acceleration differential energy reference data, the transmission elastic hysteresis compensation factor is obtained by performing relative deviation normalization and nonlinear penalty processing on the acceleration differential energy reference data. Specifically, for any target sampling time, the actual acceleration differential energy assessment and theoretical acceleration differential energy assessment corresponding to the target sampling time are extracted from the acceleration differential energy reference data. The actual first-order differential change of acceleration, the theoretical first-order differential change of acceleration, and the theoretical acceleration value corresponding to the target sampling time are extracted from the actual acceleration time series data and the theoretical acceleration time series data. The calculation result of adding the preset robust constant to prevent the denominator from being zero to the actual acceleration differential energy assessment is used as the first energy correction assessment for the target sampling time. The theoretical first-order differential change of acceleration corresponding to the target sampling time is... The difference between the quantized quantity and the actual first-order differential change of acceleration is used as the acceleration differential deviation assessment. The acceleration differential deviation assessment is used as the numerator, and the sum of the absolute value of the theoretical acceleration value corresponding to the target sampling time and the robust constant is used as the denominator. The corresponding fractional acceleration differential relative deviation at the target sampling time is normalized and assessed. The sum of the constant 1 and the square of the normalized relative deviation of the acceleration differential at the target sampling time is used as the nonlinear deviation correction term corresponding to the target sampling time. The product of the theoretical acceleration differential energy assessment and the nonlinear deviation correction term is added to the actual acceleration differential energy assessment, and the result is added to the robust constant. This result is used as the second energy correction assessment at the target sampling time. The first energy correction assessment divided by the second energy correction assessment is used as the transmission elastic hysteresis compensation factor corresponding to the target sampling time.

[0054] In one implementation, assume the first The actual first-order difference change of acceleration at each moment is ;No. The theoretical first-order difference change of acceleration at each moment is ;No. The theoretical acceleration at that moment is The robustness constant is Then the first The formula for calculating the transmission elastic hysteresis compensation factor at time t is:

[0055]

[0056] in, Indicates the first The transmission elastic hysteresis compensation factor at each moment. Indicates the first The first-order difference change of the actual acceleration at each moment; Indicates the first The theoretical first-order difference change of acceleration at each moment; Indicates the first The theoretical acceleration at each moment; This represents a robustness constant, set in the embodiments of the present invention. .

[0057] It should be noted that, to address the spurious residual problem caused by the leading prediction of the rigid model, the aforementioned transmission elastic hysteresis compensation factor with nonlinear damping characteristics was constructed. Firstly, the formula uses the squared term of the difference. and These terms represent the instantaneous energy intensity of the actual response and the theoretical excitation, respectively. Using the energy term leverages the nonlinear amplification characteristic of square operations to more significantly highlight highly dynamic characteristic signals, thereby improving the signal-to-noise ratio of the algorithm in complex noise environments and ensuring the sensitivity of the factor to transient conditions.

[0058] Nonlinear deviation correction term in the denominator of the formula This is used to construct an adaptive hysteresis sensing switch. During steady-state vehicle driving or slow acceleration, due to the good tracking properties of the physical system, the difference between the theoretical and actual rates of change is small, and relative to the current acceleration benchmark... In this case, the relative error is low, causing the correction term value to approach 1. At this point, the numerator and denominator are of similar magnitude, keeping the transmission elastic hysteresis compensation factor around 1, thus ensuring the algorithm can completely retain residual information and maintain its ability to monitor minor faults. However, once the vehicle enters a rapid acceleration condition, the elastic hysteresis effect of the transmission system becomes apparent, leading to a change in the theoretical rate of change. Much greater than the actual rate of change The difference between the two increases significantly. At this point, the correction term uses squaring to drastically amplify this bias characteristic and acts as a gain multiplier on the theoretical energy term. This causes the denominator to grow explosively and become much larger than the numerator. This allows the transmission elastic hysteresis compensation factor to rapidly decay to near zero the instant hysteresis occurs, thus achieving precise suppression of spurious residuals. Through this adaptive adjustment mechanism based on energy deviation coupling, the false alarm problem caused by physical hysteresis in rigid models under transient conditions is solved.

[0059] Thus, the transmission elastic hysteresis compensation factor was obtained by analyzing the acceleration differential energy and relative deviation.

[0060] Step S3: Obtain the dynamic decoupling singularity verification factor by performing singularity verification on the directionality of theoretical and actual acceleration changes.

[0061] After suppressing hysteresis spurious residuals using a transmission elastic hysteresis compensation factor, the system faces a new secondary risk: underreporting of decoupling faults. When a vehicle experiences faults such as clutch slippage under high load, abnormal power interruption during gearbox shifting, or drive motor torque output failure, the control system issues a high torque command, causing a surge in the theoretical acceleration prediction. However, due to the failure of friction components in the power transmission path or abnormal actuator response, the actual wheel acceleration drops or fails to build up. At this point, the theoretical rate of change and the actual rate of change exhibit a directional divergence. Since the transmission elastic hysteresis compensation factor is designed to suppress large numerical differences, this significant deviation caused by the reverse divergence is easily misjudged as severe hysteresis and forcefully attenuated, thus masking the characteristics of such functional faults. To address this logical loophole, a verification factor capable of identifying directional consistency must be introduced. This factor needs to utilize the principle of algebraic singularity to generate an explosive gain when a reversal of the trend is detected, forcibly counteracting the suppressive effect of the transmission elastic hysteresis compensation factor and pulling the real fault signal back to the alarm range; at the same time, this factor needs to have the ability to adapt to working conditions, using the proportion of theoretical power request energy as a threshold to automatically shield noise interference under low load conditions.

[0062] In summary, this invention first obtains fundamental data for dynamic decoupling directionality discrimination by performing first-order difference and directional combination processing on the actual and theoretical acceleration time series data. Specifically, for any target sampling time in the actual and theoretical acceleration time series data, the sum of the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the acceleration change direction combination and evaluation for the target sampling time. The difference between the theoretical and actual first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the acceleration change direction combination difference evaluation for the target sampling time. The product of the theoretical and actual first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the acceleration change direction product evaluation for the target sampling time. The set of acceleration change direction combination and evaluation, acceleration change direction combination difference evaluation, and acceleration change direction product evaluation corresponding to all target sampling times is used as the fundamental data for dynamic decoupling directionality discrimination.

[0063] After obtaining the basic data for dynamic decoupling directionality discrimination, the data is further robustly processed by sum-difference energy ratio and product to obtain dynamic decoupling singularity feature data. Specifically, for any target sampling time in the actual acceleration time series data and theoretical acceleration time series data, the combination and evaluation of acceleration change directions, the difference evaluation of acceleration change direction combinations, and the evaluation of acceleration change direction product are extracted from the basic data for dynamic decoupling directionality discrimination. The square of the difference evaluation of acceleration change direction combinations is used as the direction difference energy evaluation of the target sampling time, and the square of the sum evaluation of acceleration change direction combinations is used as the direction and energy evaluation of the target sampling time. The absolute value of the product of the degree change direction is used as the absolute value of the product of the direction at the target sampling time. The result of adding the direction and energy assessments, the absolute value of the product of the direction, and the preset robust constant to prevent the denominator from being zero is used as the robust denominator assessment of the sum-difference energy ratio at the target sampling time. The direction difference energy assessment is used as the robust numerator assessment of the sum-difference energy ratio. The quotient of the robust numerator assessment of the sum-difference energy ratio and the robust denominator assessment of the sum-difference energy ratio is used as the robust assessment of the sum-difference energy ratio at the target sampling time. The square of the robust assessment of the sum-difference energy ratio is used as the dynamic decoupling singularity feature assessment corresponding to the target sampling time. The set of all dynamic decoupling singularity feature assessments corresponding to the target sampling time is used as the dynamic decoupling singularity feature data.

[0064] After obtaining the dynamic decoupling singularity characteristic data, the final step is to obtain the dynamic decoupling singularity verification factor by performing energy proportion gating and load gain processing on the dynamic decoupling singularity characteristic data. Specifically, for any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the dynamic decoupling singularity characteristic evaluation corresponding to the target sampling time is extracted from the dynamic decoupling singularity characteristic data, and the theoretical acceleration value and the first-order difference change of the theoretical acceleration corresponding to the target sampling time are extracted from the theoretical acceleration time series data. The square of the theoretical acceleration value corresponding to the target sampling time is used as the theoretical acceleration energy evaluation. The square of the first-order difference change of the theoretical acceleration is used as the energy assessment of the rate of change of theoretical acceleration. The energy assessment of theoretical acceleration is used as the numerator, and the sum of the energy assessment of theoretical acceleration, the energy assessment of the rate of change of theoretical acceleration, and the preset robust constant to prevent the denominator from being zero is used as the denominator. The resulting fraction is used as the energy proportion gating coefficient assessment at the target sampling time. The absolute value of the theoretical acceleration value corresponding to the target sampling time is used as the load gain assessment. The product of the dynamic decoupling singularity characteristic assessment, the energy proportion gating coefficient assessment, and the load gain assessment is added to the constant 1 to calculate the result as the dynamic decoupling singularity verification factor corresponding to the target sampling time.

[0065] In one embodiment, the first The expression for calculating the dynamic decoupling singularity check factor at time t is:

[0066]

[0067] in, Indicates the first The dynamic decoupling singularity check factor at each moment; Indicates the first The first-order difference change of the actual acceleration at each moment; Indicates the first The theoretical first-order difference change of acceleration at each moment; Indicates the first The theoretical acceleration at each moment; This represents a robustness constant, set in the embodiments of the present invention. .

[0068] The bracketed terms in the first part of the formula It is a decoupling fault feature extractor. When the vehicle is in normal driving or in a simple elastic hysteresis state, the theoretical rate of change is... Compared with the actual rate of change The directions are the same (i.e., both positive or both negative). In this case, the square of the sum in the denominator... The value is large, and the product term is large. This further increases the numerical stability of the denominator; while the squared difference in the numerator is relatively small. Therefore, the overall fraction value approaches zero, keeping the dynamic decoupling singularity check factor around 1, ensuring the algorithm does not interfere with normal hysteresis suppression logic. Here, an absolute value term for the product is introduced. The purpose of this term is to construct coupled energy damping. In numerical calculations, relying solely on the square of the sum as the denominator presents a risk: when the system is under high-frequency vibration, the theoretical and actual rates of change may exhibit transients with similar amplitudes but opposite signs, causing the square of the sum to approach zero, generating non-fault-related calculation spikes. By introducing a cross-correlation energy term, as long as either signal has a significant amplitude, this term ensures the denominator remains at a safe numerical level, preventing false triggering in the non-fault oscillation zone. However, once a power decoupling fault occurs, such as clutch slippage under high load or drive motor torque output failure, the theoretical end accelerates (positive), while the actual end decelerates (negative), with opposite signs. At this point, the numerator becomes the square of the sum of absolute values ​​(maximum); while in the denominator, the square of the sum becomes the square of the difference in absolute values ​​(minimum). This inverse characteristic causes a singular abrupt change in the value of this term, rapidly increasing. Further sharpening this characteristic through squaring operations allows for accurate identification of fault states with divergent directions. Secondly, the parenthesized terms in the second part of the formula... This constitutes an energy percentage gating mechanism. To prevent false triggering of reverse verification under vehicle idling vibration and sensor noise interference, this step utilizes theoretical acceleration energy. Its rate of change of energy The ratio is constructed as follows: Under steady-state cruise or idling conditions, the energy proportion of theoretical acceleration is low, and this value is small, automatically closing the fault verification channel; while under high-power-demand conditions such as rapid acceleration, the energy proportion of acceleration increases significantly, and this value approaches 1, automatically activating the fault verification function. This design achieves intelligent perception of the vehicle's operating status without needing to preset a fixed acceleration threshold. Finally, the third part of the formula... This constitutes an adaptive dynamic gain stage based on operating load. This stage directly defines the gain strength of the verification factor using the theoretical acceleration value, establishing a positive correlation between the fault signal amplification factor and the current power request intensity. The larger the acceleration value requested by the driver, the higher the current power demand; therefore, if a power decoupling fault occurs under these conditions, the impact on vehicle safety will be more significant. Thus, this stage linearly increases the signal gain with the increase in power request. When a reverse fault characteristic is detected, it provides signal amplification capability matching the current operating condition, effectively offsetting the attenuation effect of the transmission elastic hysteresis compensation factor under high load, ensuring that serious faults can be warned in a timely and accurate manner.

[0069] Step S4: The basic residual is weighted by the transmission elastic hysteresis compensation factor and the power decoupling singularity verification factor to obtain the final fault discrimination index for fault threshold determination.

[0070] This step aims to apply the dynamic optimization factors constructed in steps S2 and S3 to the existing residual analysis framework to generate the final discrimination criteria. Specifically, for any target sampling time in the actual acceleration time series data and theoretical acceleration time series data, the calculation result of multiplying the basic residual evaluation, transmission elastic hysteresis compensation factor evaluation, and dynamic decoupling singularity verification factor evaluation corresponding to the target sampling time is used as the single-point evaluation of the final fault discrimination index corresponding to the target sampling time. The set of single-point evaluations of the final fault discrimination index corresponding to all target sampling times is used as the final fault discrimination index for fault threshold determination.

[0071] Thus, the final fault discrimination index for fault threshold determination is obtained by weighting the basic residuals using the transmission elastic hysteresis compensation factor and the power decoupling singularity verification factor.

[0072] Step S5: Obtain vehicle drive system fault warning results by performing time window statistics and threshold determination on the final fault discrimination indicators.

[0073] After obtaining the final fault identification index of the vehicle, the final fault identification index is input to the fault determination logic module. This module adopts a sliding time window integration strategy to calculate the cumulative amplitude of the final fault identification index within a set time window. In this embodiment of the invention, the time window length is set to 1 second of sampling data, corresponding to a sampling frequency of 100Hz. The time window contains 100 sampling points. This statistical value is compared with a preset system fault alarm threshold. In this embodiment of the invention, the fault threshold is set to 0.3. This fault threshold can be adjusted according to the actual scenario and is not required.

[0074] If the statistical value is below the threshold, it indicates that the current index fluctuation is mainly caused by the physical hysteresis residue after suppression by the transmission elastic hysteresis compensation factor, which is within the normal driving range, and the system remains silent.

[0075] If the statistical value continues to be higher than the threshold, it indicates that the indicator contains abnormal components that cannot be suppressed and may be amplified by the power decoupling singularity check factor, thus determining that there is a real fault in the vehicle drive system.

[0076] At this point, the system further confirms the nature of the fault based on the magnitude of the dynamic decoupling singularity check factor: if the final fault discrimination index exceeds the limit and is accompanied by a high gain state of the dynamic decoupling singularity check factor (i.e. If the fault is detected, it is clearly determined to be a decoupling fault such as clutch slippage under high load or failure of drive motor torque output. An audible and visual warning signal will be issued to the driver immediately, and a torque limit request will be sent to the vehicle control unit (VCU) to protect the transmission system.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for timely early warning of vehicle faults based on multi-source data analysis, characterized in that, The method includes: Step S1: Obtain the actual acceleration and theoretical acceleration benchmark by collecting and performing dynamic calculations on multi-source driving data, and obtain the basic residual of vehicle acceleration based on the difference between the actual acceleration and the theoretical acceleration benchmark; Step S2: Obtain the transmission elastic hysteresis compensation factor by analyzing the acceleration differential energy and relative deviation; Step S3: Obtain the dynamic decoupling singularity verification factor by performing singularity verification on the directionality of theoretical and actual acceleration changes; Step S4: The basic residual is weighted by the transmission elastic hysteresis compensation factor and the power decoupling singularity verification factor to obtain the final fault discrimination index for fault threshold determination; Step S5: Obtain vehicle drive system fault warning results by performing time window statistics and threshold determination on the final fault discrimination indicators; The method of obtaining the transmission elastic hysteresis compensation factor by analyzing the acceleration differential energy and relative deviation includes: obtaining acceleration differential energy reference data by performing a first-order difference operation on the actual acceleration and theoretical acceleration time series data; and obtaining the transmission elastic hysteresis compensation factor by performing relative deviation normalization and nonlinear penalty processing on the acceleration differential energy reference data. The method of obtaining the dynamic decoupling singularity verification factor by verifying the singularity of the directionality of theoretical and actual acceleration changes includes: obtaining basic data for dynamic decoupling directionality discrimination by performing first-order difference and directionality combination processing on the time series data of actual and theoretical acceleration; obtaining dynamic decoupling singularity feature data by performing sum-difference energy ratio and product robustness processing on the basic data for dynamic decoupling directionality discrimination; and obtaining the dynamic decoupling singularity verification factor by performing energy proportion gating and load gain processing on the dynamic decoupling singularity feature data. The process of obtaining the transmission elastic hysteresis compensation factor by performing relative deviation normalization and nonlinear penalty processing on the acceleration differential energy reference data includes: for any target sampling time, extracting the actual acceleration differential energy assessment and theoretical acceleration differential energy assessment corresponding to the target sampling time from the acceleration differential energy reference data; extracting the actual first-order differential change of acceleration, the theoretical first-order differential change of acceleration, and the theoretical acceleration value corresponding to the target sampling time from the actual acceleration time series data and the theoretical acceleration time series data; adding the calculated result of the preset robust constant to prevent the denominator from being zero to the actual acceleration differential energy assessment as the first energy correction assessment for the target sampling time; and adding the theoretical first-order differential change of acceleration to the actual acceleration differential energy assessment at the target sampling time. The difference in the order of differential changes is used as the acceleration differential deviation assessment. The acceleration differential deviation assessment is used as the numerator, and the sum of the absolute value of the theoretical acceleration value corresponding to the target sampling time and the robust constant is used as the denominator. The corresponding fractional acceleration differential relative deviation at the target sampling time is normalized and assessed. The sum of the constant 1 and the square of the normalized acceleration differential relative deviation assessment at the target sampling time is used as the nonlinear deviation correction term corresponding to the target sampling time. The product of the theoretical acceleration differential energy assessment and the nonlinear deviation correction term is added to the actual acceleration differential energy assessment, and the result is added to the robust constant as the second energy correction assessment at the target sampling time. The first energy correction assessment divided by the second energy correction assessment is used as the transmission elastic hysteresis compensation factor corresponding to the target sampling time.

2. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process of acquiring actual acceleration and theoretical acceleration benchmarks by collecting and performing dynamic calculations on multi-source driving data, and obtaining the basic residual of vehicle acceleration based on the difference between the actual acceleration and the theoretical acceleration benchmark, includes: By installing an inertial measurement unit on the vehicle body near the center of gravity of the vehicle and setting a sampling frequency, the inertial measurement unit is continuously triggered to collect data during the vehicle's driving process according to the sampling frequency, thereby obtaining actual acceleration time series data that characterizes the longitudinal driving state of the vehicle, and using the actual acceleration time series data as the actual acceleration for the calculation of the basic residual. The system collects actual output torque data of the engine or drive motor from the engine management system through the vehicle's controller area network, current gear ratio data and final drive ratio data from the transmission controller, and vehicle speed data and road slope data from the electronic stability system. The collected actual output torque data, current gear ratio data, final drive ratio data, vehicle speed data and road slope data are used as multi-source driving data for theoretical acceleration benchmark calculation. The vehicle electronic control unit reads pre-calibrated vehicle calibration parameters from the on-board storage medium. These parameters include at least the rolling radius of the vehicle's drive wheels, vehicle mass, drag coefficient, frontal area, mechanical efficiency, rolling resistance coefficient, air density, and rotational mass coefficient. The vehicle calibration parameters are then combined with the multi-source driving data and input into the all-element vehicle longitudinal dynamics equation. Based on the all-element vehicle longitudinal dynamics equation, theoretical acceleration time-series data corresponding to the actual acceleration time-series data are calculated at each sampling time. This theoretical acceleration time-series data is then used as the theoretical acceleration reference. For each target sampling time in the actual acceleration time series data, the absolute value of the difference between the actual acceleration at the target sampling time and the theoretical acceleration benchmark at the corresponding target sampling time is calculated, and the obtained absolute value of the difference is used as the basic residual corresponding to the target sampling time.

3. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process of obtaining acceleration differential energy reference data by performing a first-order difference operation on the time-series data of actual acceleration and theoretical acceleration includes: For any target sampling time in the actual acceleration time series data, excluding the first sampling time, the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time are extracted from the actual acceleration time series data. The difference between the actual acceleration value corresponding to the target sampling time and the actual acceleration value corresponding to the previous sampling time is taken as the first-order differential change of the actual acceleration at the target sampling time. The square of the first-order differential change of the actual acceleration is taken as the actual acceleration differential energy assessment at the target sampling time. For any target sampling time in the theoretical acceleration time series data, excluding the first sampling time, the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time are extracted from the theoretical acceleration time series data. The difference between the theoretical acceleration value corresponding to the target sampling time and the theoretical acceleration value corresponding to the previous sampling time is taken as the first-order difference change of theoretical acceleration at the target sampling time. The square of the first-order difference change of theoretical acceleration is taken as the theoretical acceleration difference energy assessment at the target sampling time. The set of actual acceleration differential energy assessments and theoretical acceleration differential energy assessments corresponding to all target sampling times is used as the acceleration differential energy benchmark data.

4. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process involves performing first-order difference and directional combination processing on the time-series data of actual and theoretical accelerations to obtain basic data for dynamic decoupling directionality discrimination, including: For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the sum of the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the combination and evaluation of the acceleration change direction at the target sampling time. The difference between the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the combination difference evaluation of the acceleration change direction at the target sampling time. The product of the theoretical first-order difference change and the actual first-order difference change corresponding to the target sampling time is used as the product evaluation of the acceleration change direction at the target sampling time. The set of the combination and evaluation of the acceleration change direction, the combination difference evaluation of the acceleration change direction, and the product evaluation of the acceleration change direction corresponding to all target sampling times is used as the basic data for dynamic decoupling directionality discrimination.

5. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process involves performing first-order difference and directional combination processing on the time-series data of actual and theoretical accelerations to obtain basic data for dynamic decoupling directionality discrimination, including: For any target sampling moment in the actual acceleration time series data and theoretical acceleration time series data, extract the acceleration change direction combination and evaluation, acceleration change direction combination difference evaluation, and acceleration change direction product evaluation corresponding to the target sampling moment from the dynamic decoupling directionality discrimination basic data. Use the square of the acceleration change direction combination difference evaluation as the direction difference energy evaluation of the target sampling moment, use the square of the acceleration change direction combination and evaluation as the direction and energy evaluation of the target sampling moment, and use the absolute value of the acceleration change direction product evaluation as the absolute value evaluation of the direction product of the target sampling moment. The calculation result of adding the direction and energy assessment, the absolute value of the product of directions, and a preset robust constant to prevent the denominator from being zero is used as the robust denominator assessment of the sum-difference energy ratio at the target sampling time. The direction difference energy assessment is used as the robust numerator assessment of the sum-difference energy ratio. The quotient of the robust numerator assessment and the robust denominator assessment of the sum-difference energy ratio is used as the robust assessment of the sum-difference energy ratio at the target sampling time. The square of the robust assessment of the sum-difference energy ratio is used as the dynamic decoupling singularity feature assessment corresponding to the target sampling time. The set of dynamic decoupling singularity feature assessments corresponding to all target sampling times is used as the dynamic decoupling singularity feature data.

6. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process of obtaining a dynamic decoupling singularity verification factor by performing energy proportion gating and load gain processing on the dynamic decoupling singularity characteristic data includes: For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the dynamic decoupling singularity feature assessment corresponding to the target sampling time is extracted from the dynamic decoupling singularity feature data. The theoretical acceleration value and the first-order difference change of theoretical acceleration corresponding to the target sampling time are extracted from the theoretical acceleration time series data. The square of the theoretical acceleration value corresponding to the target sampling time is used as the theoretical acceleration energy assessment, and the square of the first-order difference change of theoretical acceleration corresponding to the target sampling time is used as the theoretical acceleration rate of change energy assessment. The theoretical acceleration energy assessment is used as the numerator, and the sum of the theoretical acceleration energy assessment, the theoretical acceleration rate of change energy assessment, and the preset robust constant to prevent the denominator from being zero is used as the denominator. The resulting fraction is used as the energy proportion gating coefficient assessment for the target sampling time. The absolute value of the theoretical acceleration value corresponding to the target sampling time is used as the load gain assessment. The product of the dynamic decoupling singularity feature assessment, the energy proportion gating coefficient assessment, and the load gain assessment is added to the constant 1, and the result is used as the dynamic decoupling singularity check factor corresponding to the target sampling time.

7. The method for timely early warning of vehicle faults based on multi-source data analysis according to claim 1, characterized in that, The process of weighting the basic residuals using a transmission elastic hysteresis compensation factor and a dynamic decoupling singularity verification factor to obtain the final fault discrimination index for fault threshold determination includes: For any target sampling time in the actual acceleration time series data and the theoretical acceleration time series data, the calculation result of multiplying the basic residual evaluation, transmission elastic hysteresis compensation factor evaluation and dynamic decoupling singularity verification factor evaluation corresponding to the target sampling time is used as the single-point evaluation of the final fault discrimination index corresponding to the target sampling time. The set of single-point evaluations of the final fault discrimination index corresponding to all target sampling times is used as the final fault discrimination index for fault threshold determination.

Images