Method for testing durability of window glass lifter of new energy automobile

Abstract

The invention relates to the technical field of durability testing, in particular to a durability testing method for a new energy automobile glass lifter, which comprises the following steps of: acquiring a ratio of a current change rate to a sliding speed, judging a trend by differentiating a sliding window, generating a trend instruction, calculating current peak offset, and generating delay pulses by fuzzy classification. And acquiring symmetry offset and screening abnormal sequences, outputting an early warning level through a two-parameter clustering model, and correcting an abnormal label level through a Drools engine. According to the method, the ratio of the motor load current change rate to the glass sliding speed is dynamically monitored, the ratio of the current change rate to the sliding speed is monitored, abnormal fluctuation is analyzed and captured through a sliding window, a fuzzy logic classifier is combined to classify the current peak offset to generate delay pulses, the symmetry offset is calculated, and the trend is evaluated. The section positions are compared, the grade labels are output, multiple algorithms are fused to form closed-loop feedback, and the early warning precision and the testing efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of durability testing, and in particular to a durability testing method for a window regulator of a new energy vehicle. Background Art

[0002] The technical field of durability testing of new energy vehicles includes methods for evaluating and verifying the performance changes of multiple key components of new energy vehicles during long-term use. By setting simulated use environments and operating conditions, automotive components are tested for a long time and repeatedly to determine failures and performance degradation that occur in actual applications. The key aspects of durability testing include the electric drive system, power battery system, vehicle structural parts, and vehicle internal control and transmission systems. Through scientific and systematic testing methods, the performance stability and structural integrity data of components under long-term operation are obtained. In the process of development in the field, a complete testing system supported by environmental simulation test benches, load cycle systems and automated testing platforms has gradually been formed, with the characteristics of long test cycles, complex operating parameters and fine data analysis.

[0003] Among them, a durability test method for window lifters of new energy vehicles refers to a test scheme used to evaluate the structural and functional changes of window lifters of new energy vehicles under long-term operating conditions. The key to the patent subject includes the continuous and repeated opening and closing control operations of the window lifter in a way of setting the operating cycle and lifting stroke, and simulating the working state under the condition of vehicle voltage changes through an external control system; at the same time, a mechanical loading device is used to simulate the influence of glass quality and external resistance on the operation process of the lifter; and the response parameters of the lifter in multiple operating stages are synchronously collected through a time recording device, a current and voltage monitor and a position sensor to form a systematic data record; during the test, the motor temperature rise and the degree of gear wear are regularly sampled and analyzed. This method uses continuous operation control, voltage disturbance simulation, mechanical resistance loading and multi-dimensional synchronous monitoring as key means to complete the entire test process.

[0004] Existing technologies use fixed-cycle operation control and static threshold determination mechanisms, making it difficult to identify the progressive performance degradation characteristics of lifters under complex operating conditions such as voltage fluctuations and dynamic changes in mechanical resistance. Environmental simulation test benches rely on preset load cycle patterns and lack the ability to analyze the correlation between motor load current and motion parameters in real time, resulting in delayed detection of abnormal operating conditions. While mechanical loading devices can simulate constant resistance, they lack a dynamic mapping model between resistance changes and electrical signal responses, making them unable to capture nonlinear current fluctuations caused by structural damage due to increased gear clearance. Data acquisition systems employ a discrete monitoring approach and fail to perform simultaneous correlation analysis of multi-parameter time series features, resulting in the smoothing and filtering out of key fault signatures during data preprocessing. Traditional testing methods rely on empirical threshold setting and manual visual inspection, lacking quantitative analysis of weak signals generated by early wear, and are prone to misjudgment or missed detections. During the development stage of microcracks on gear tooth surfaces, existing technologies, lacking a correlation model between current peak offset and motion symmetry, make it difficult to distinguish between normal operating fluctuations and early fault signals, causing the warning threshold setting to deviate from the degradation trajectory. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a durability testing method for a new energy vehicle window regulator.

[0006] In order to achieve the above-mentioned object, the present invention adopts the following technical solution: a durability testing method for a new energy vehicle window regulator, comprising the following steps:

[0007] S1: Obtain the lifter motor load current change rate and the glass sliding speed and calculate the ratio. Calculate the difference of the ratio data series of consecutive cycles and substitute it into the sliding window analysis method for judgment. If both are greater than the differential trend judgment threshold and show an increasing trend, a trend judgment instruction is generated.

[0008] S2: Based on the trend judgment instruction, calculate the current peak offset between the peak amplitude of the motor load current in the current cycle and the peak amplitude of the previous cycle, substitute it into the fuzzy logic classifier to divide the amplitude level, and generate a delay adjustment pulse according to the delay mapping rule corresponding to the level;

[0009] S3: Using the delay adjustment pulse to collect the time from glass start-up to current steep rise and the time from current zero return to displacement stop during the lifting and lowering cycle, calculate the symmetric time offset by directional segment, apply the sliding window proportion function to evaluate, and use the current peak offset as a benchmark to select data with an offset increase of more than twice to form a symmetric offset sequence;

[0010] S4: Based on the symmetrical offset sequence and the current peak offset, the offset segment positions in the cycle are compared, and a dual-parameter K-means clustering model is input for clustering, and a wear warning level label is output;

[0011] S5: Input the wear warning level label data, trend judgment instructions and symmetry offset sequence into the Drools rule engine and perform Z-score normalization processing. If the differential trend and symmetry offset trend continue to rise and the peak fluctuation is abnormal, the label level is automatically corrected.

[0012] As a further solution of the present invention, the trend judgment instruction specifically includes a differential trend threshold, an increasing trend state, and a sliding window analysis result; the delay adjustment pulse includes a current peak offset, an amplitude level classification result, and a delay mapping rule; the symmetry offset sequence specifically refers to a time symmetry offset, a sliding window proportion evaluation result, and a two-fold growth screening condition; the wear warning level label includes a segment position comparison parameter, a two-parameter clustering model output, and a wear level classification standard; the automatically corrected label level specifically includes Z-score standardized data, a continuous trend judgment indicator, and a peak fluctuation anomaly.

[0013] As a further solution of the present invention, the lifter motor load current change rate and the glass sliding speed are obtained and the ratio is calculated. The difference of the ratio data series of consecutive cycles is calculated and substituted into the sliding window analysis method for judgment. If both are greater than the difference trend judgment threshold and show an increasing trend, the specific steps of generating a trend judgment instruction are as follows:

[0014] S101: Obtaining the rate of change of the lifter motor load current and the glass sliding speed, respectively calculating the change value of the motor load current and the change value of the glass sliding distance, and calculating the ratio within the same cycle to generate a ratio sequence;

[0015] S102: Based on the ratio sequence, call data segments within three consecutive periods, construct a differential data group and a continuous window based on the difference between the ratios of two adjacent periods, arrange and analyze multiple differential data within the window, and generate a differential trend interval;

[0016] S103: For the differential trend interval, make a judgment according to the differential trend determination threshold, compare the data trend change direction and the incrementality in multiple windows one by one, adjust the differential trend determination threshold according to the window size, lower the threshold for large windows and increase it for small windows, and judge whether it satisfies both the differential trend determination threshold and the continuous incremental direction, and generate a trend judgment instruction.

[0017] As a further solution of the present invention, based on the trend judgment instruction, the current peak offset between the peak amplitude of the motor load current in the current cycle and the peak amplitude of the previous cycle is calculated, and the current peak offset is substituted into the fuzzy logic classifier to divide the amplitude level, and the specific steps of generating the delay adjustment pulse according to the delay mapping rule corresponding to the level are as follows:

[0018] S201: Based on the trend judgment instruction, calling the maximum instantaneous current value in the motor load current signal of the current cycle and the previous cycle, calculating the difference between the maximum instantaneous current values ​​in the two cycles, and comparing it with the current amplitude change reference value corresponding to the current cycle acquisition period to obtain the current peak offset;

[0019] S202: Based on the current peak offset, calling the amplitude boundary value in the fuzzy logic classifier classification rule, classifying the corresponding position of the current peak offset into the target amplitude level interval, and performing a mapping operation with the hierarchical structure set by the fuzzy logic classifier to generate an amplitude level identifier;

[0020] S203: calling the amplitude level identifier, and loading the corresponding pulse control instruction within a period according to the pulse adjustment coefficient corresponding to the identifier in the delay mapping rule table preset by the fuzzy logic classifier to generate a delay adjustment pulse.

[0021] As a further solution of the present invention, the time from glass start-up to current steep rise and the time from current zero return to displacement cessation during the rise and fall cycle are collected by using the delay adjustment pulse, the time symmetry offset is calculated by directional segment, and a sliding window proportion function is applied for evaluation. Using the current peak offset as a benchmark, data with an offset increase of more than twice are screened to form a symmetry offset sequence. The specific steps are as follows:

[0022] S301: Obtaining the timing data of the rising and falling periods in the delay adjustment pulse, extracting the start and end time from the glass starting to the current steep rise stage and the start and end time from the current returning to zero to the displacement stop stage, calculating the duration difference between the two time periods, and constructing the rising and falling response time difference according to the action direction;

[0023] S302: Based on the lifting and lowering response time difference, the time difference between the lifting and lowering stages is used as the basis for determining the offset direction, with the positive and negative signs set in the order of the lifting stage first and the lowering stage last. The resulting ratio is then used as a basis for symmetry determination and classified into the lifting and lowering action category structure to obtain a symmetry offset.

[0024] S303: Call the symmetry offset, calculate the offset change rate through the sliding window proportion function, compare the offset difference between the current cycle and the previous cycle, the synchronous growth of the symmetry offset and the current peak offset indicates a mechanical unilateral abnormality, and independent changes distinguish electrical interference, determine whether it exceeds twice the current peak offset, and obtain the symmetry offset sequence.

[0025] As a further solution of the present invention, based on the symmetrical offset sequence and the current peak offset, the offset segment positions in the cycle are compared, and a dual-parameter K-means clustering model is input for clustering division. The specific steps of outputting the wear warning level label are as follows:

[0026] S401: Based on the symmetrical offset sequence and the current peak offset, construct an offset time segment by marking the ratio of the start and end times to the total cycle duration, form a time axis index table of the offset segments by sorting them by cycle number, and generate a cycle segment offset ratio based on the time proportion distribution of the offset segments in the cycle;

[0027] S402: Calling the cycle segment offset ratio and the current peak offset, constructing the two numerical values ​​into a data set with dual-variable attributes, inputting them into a dual-parameter K-means clustering model, initializing the cluster centers, and generating cycle offset cluster labels based on the spatial distribution of the two variables;

[0028] The dual-parameter K-means clustering model includes input variable dimension, cluster number setting, cluster center initialization, distance measurement method, iterative convergence criterion, and output structure;

[0029] S403: Based on the periodic offset clustering labels, extract the mean of the offset ratio and peak offset of each sample in multiple label categories as the label feature benchmark and construct a feature distribution interval. Perform interval positioning judgment based on the ratio of the mean offset difference between multiple categories of samples and the corresponding interval to obtain the wear warning level label.

[0030] As a further solution of the present invention, the input is input into a two-parameter K-means clustering model and the cluster center is initialized using the formula:

[0031]

[0032] Among them, d p Represents the dynamic weighted distance index of the p-th sample point, R q represents the period segment deviation ratio measurement, represents the coordinates of the rth initial cluster center, P s represents the current peak offset measurement value, α represents the bivariate weight adjustment factor (0≤α≤1), β represents the variance normalization coefficient of the cycle segment offset ratio, and γ represents the range normalization coefficient of the current peak offset.

[0033] As a further solution of the present invention, the wear warning level label data, trend judgment instructions, and symmetry offset sequence are input into the Drools rule engine and Z-score normalization is performed. If the differential trend and symmetry offset trend continuously increase and the peak fluctuation is abnormal, the label level is automatically corrected in the following specific steps:

[0034] S501: Based on the wear warning level label data, friction trend judgment instruction and symmetry offset sequence input items, the data is loaded into the Drools rule engine, field values ​​are read and pre-matched, variable binding and rule activation initialization processes are performed on the structural format of the input data, and a rule triggering basic variable set is generated;

[0035] The Drools rule engine includes a knowledge base, a working memory, a rule structure, a rule execution process, and a variable binding mechanism;

[0036] S502: Call the rule to trigger multiple numerical variables in the basic variable set, calculate the mean and difference respectively according to the field category, complete the discretization mapping of all variables through the Z-score normalization method, and generate a normalized offset fluctuation series;

[0037] S503: Based on the normalized offset fluctuation sequence, the continuous time periods related to the differential trend and the fluctuation trajectory of the symmetrical offset are screened to determine whether the offset increment directions of the two data in multiple continuous segments continue to change in a positive direction at the same time. If the condition is met, the label level field is updated.

[0038] As a further solution of the present invention, the discretization mapping of all variables is completed by Z-score standardization, using the formula:

[0039]

[0040] Among them, θ mn represents the weighted normalized feature value of the nth sample in the mth field, X mn Represents the original numerical variable measurement value, μ m represents the sliding window mean of the mth field, σ m Represents the dynamic standard deviation of the mth field, δ m represents the difference between adjacent data windows of the mth field, θ n represents the cross-field correlation factor of the nth sample, and ξ represents the global normalization reference constant.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are:

[0042] In the present invention, by dynamically monitoring the ratio of the motor load current change rate to the glass sliding speed, the differential trend within the continuous cycle is judged in real time in combination with the sliding window analysis method, thereby capturing the abnormal fluctuation characteristics during the operation of the lifter. Based on the hierarchical processing of the current peak offset by the fuzzy logic classifier, the generation of delayed adjustment pulses is performed, which significantly improves the response sensitivity to the sudden change of the motor load. By calculating the symmetry offset and evaluating the trend of the sliding window proportion function, a multi-dimensional time series association model is established to enhance the ability to identify nonlinear degradation characteristics during the wear of mechanical components. A two-parameter clustering model is used to compare and analyze the segment positions of the current peak offset and the symmetry offset sequence, breaking through the limitations of the traditional single-dimensional threshold judgment and performing multi-level division of wear levels. The technical path deeply integrates time series data analysis, fuzzy logic reasoning and machine learning algorithms to form a closed-loop feedback mechanism. While reducing the dependence on manual intervention, it significantly improves the accuracy and test efficiency of early wear warning, providing high-resolution data support for the quantitative evaluation of the durability performance of the lifter under working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the main steps of the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0045] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0046] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0047] S1: Obtain the lifter motor load current change rate and the glass sliding speed and calculate the ratio. Calculate the difference of the ratio data series of consecutive cycles and substitute it into the sliding window analysis method for judgment. If both are greater than the differential trend judgment threshold and show an increasing trend, a trend judgment instruction is generated.

[0048] S2: Based on the trend judgment instruction, the current peak offset between the peak amplitude of the motor load current in the current cycle and the previous cycle is calculated, and the current is substituted into the fuzzy logic classifier to divide the amplitude level, and the delay adjustment pulse is generated according to the delay mapping rule corresponding to the level;

[0049] S3: By delaying and adjusting the pulse acquisition period, the time from glass start-up to current steep rise and the time from current zero return to displacement cessation are calculated. The symmetric time offset is calculated by directional segment and evaluated using a sliding window proportion function. Using the current peak offset as a benchmark, data with an offset increase of more than twice are selected to form a symmetric offset sequence.

[0050] S4: Based on the symmetrical offset sequence and the current peak offset, the offset segment position in the cycle is compared, and a two-parameter K-means clustering model is input for clustering, and the wear warning level label is output;

[0051] S5: Input the wear warning level label data, trend judgment instructions, and symmetry offset sequence into the Drools rule engine and perform Z-score normalization. If the differential trend and symmetry offset trend continue to rise and the peak fluctuation is abnormal, the label level is automatically corrected.

[0052] The trend judgment instructions specifically include the differential trend threshold, the increasing trend status, and the sliding window analysis results. The delay adjustment pulse includes the current peak offset, the amplitude level classification results, and the delay mapping rules. The symmetry offset sequence specifically refers to the time symmetry offset, the sliding window proportion assessment results, and the two-fold growth screening conditions. The wear warning level label includes the segment position comparison parameter, the two-parameter clustering model output, and the wear level classification standard. The automatically corrected label level specifically includes Z-score standardized data, continuous trend judgment indicators, and peak fluctuation anomalies.

[0053] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0054] S101: Obtaining the rate of change of the lifter motor load current and the glass sliding speed, respectively calculating the change value of the motor load current and the change value of the glass sliding distance, and calculating the ratio within the same cycle to generate a ratio sequence;

[0055] Based on the operation data collection of the lifter motor, within the time window t = 0s to t = 2s, the current value sequence [2.3A, 2.5A, 2.8A, 3.1A, 3.4A] is obtained with a sampling interval of 100ms. The glass displacement sensor data is synchronously collected to obtain the position sequence [0mm, 12mm, 25mm, 39mm, 54mm]. The current difference between adjacent sampling points is calculated to obtain the change rate sequence [0.2A, 0.3A, 0.3A, 0.3A]. The displacement difference generates the speed sequence [12mm / s, 13mm / s, 14m / s]. m / s, 15 mm / s], take the average current change rate of 0.25 A and the average speed of 12.5 mm / s in the period from t = 0.1s to 0.3s, perform the ratio operation 0.25A / 12.5mm / s = 0.02A·s / mm, repeat the process to obtain five periodic ratio sequences [0.0167, 0.0200, 0.0214, 0.0200, 0.0227], as shown in Table 1, forming the ratio sequence [0.0167, 0.0200, 0.0214, 0.0200, 0.0227].

[0056] Table 1 Motor operating parameter collection table

[0057]

[0058]

[0059] S102: Based on the ratio sequence, call data segments within three consecutive periods, construct a differential data group based on the difference between the ratios of two adjacent periods and construct a continuous window, arrange and analyze multiple differential data within the window, and generate a differential trend interval;

[0060] Select the ratio sequence index 1-3 data [0.0200, 0.0214, 0.0200], calculate the adjacent index difference 0.0214-0.0200=0.0014 and 0.0200-0.0214=-0.0014, and construct the difference group [0.0014,

[0061] -0.0014], set the window length to 2, sort the difference values ​​by size to get [-0.0014, 0.0014], determine the interval lower limit -0.0014 and upper limit 0.0014, take the second window index 2-4 data [0.0214,

[0062] 0.0200,0.0227], calculate the difference -0.0014 and 0.0027, sort them to get [-0.0014, 0.0027], generate the second interval [-0.0014, 0.0027], and the continuous window data forms a trend interval set {[-0.0014, 0.0014], [-0.0014, 0.0027]}.

[0063] S103: For the differential trend interval, judge according to the differential trend judgment threshold, compare the data trend change direction and the incrementality in the multiple windows one by one, determine whether it satisfies both the differential trend judgment threshold and the continuous incremental direction, and generate a trend judgment instruction.

[0064] The trend judgment threshold is set to 0.0020, which is derived from the maximum fluctuation value of normal working conditions in data statistics. The upper limit of the first window 0.0014 is less than the threshold 0.0020 and is marked as an invalid interval. The upper limit of the second window 0.0027 is greater than the threshold 0.0020. The data direction is verified: the last item 0.0027 in the difference sequence [-0.0014, 0.0027] is greater than the previous item -0.0014, but no continuous increase is formed. Only the threshold condition is met within the judgment window, but the direction condition is not met. The output trend judgment instruction is 0 (invalid instruction). When three consecutive windows simultaneously meet the upper limit > 0.0020 and the difference value is monotonically increasing, the output instruction is 1 (valid trend).

[0065] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0066] S201: Based on the trend judgment instruction, call the maximum instantaneous current value in the motor load current signal of the current cycle and the previous cycle, calculate the difference between the maximum instantaneous current values ​​in the two cycles, and compare it with the current amplitude change reference value corresponding to the current cycle acquisition period to obtain the current peak offset;

[0067] Based on the trend judgment, the command value is 1, and the maximum current value of the nth cycle is extracted. With the n-1th cycle Calculate the difference ΔI = 3.5-3.2 = 0.3A and obtain the reference value B of the current amplitude change in the current cycle c = 0.25A (calculated by the moving average of the maximum current value of the first 10 cycles), perform offset calculation O p =|0.3-0.25|=0.05A, as shown in Table 2, when O p >B c The peak current deviation is 0.05A.

[0068] Table 2 Current peak comparison table

[0069]

[0070] S202: Based on the current peak offset, the amplitude boundary value in the fuzzy logic classifier classification rule is called, the corresponding position of the current peak offset is classified into the target amplitude level interval, and a mapping operation is performed with the hierarchical structure set by the fuzzy logic classifier to generate an amplitude level identifier;

[0071] Set the fuzzy logic amplitude boundary value to [0.1, 0.2] A (based on the 2%-4% safety fluctuation range of the motor rated current of 5A), compare the offset 0.05A with the boundary value, 0.05<0.1 is judged as the low amplitude level L1, and the classification mapping table is queried to obtain the identification code L1. When the offset is 0.15A, 0.1≤0.15<0.2 is classified as the medium amplitude level M2, and the identification code M2 ​​is generated. If the offset is 0.25A, ≥0.2 is classified as the high amplitude level H3. Through the table lookup operation, the level identification is matched with the preset 9-level classification structure [L1, L2, L3, M1, M2, M3, H1, H2, H3], and the amplitude level identification M2 is output.

[0072] S203: calling the amplitude level identifier, and loading the corresponding pulse control instruction within the cycle according to the pulse adjustment coefficient corresponding to the identifier in the delay mapping rule table preset by the fuzzy logic classifier, to generate a delay adjustment pulse;

[0073] The adjustment coefficient k corresponding to the M2 identifier in the delay mapping rule table is called = 0.8 (derived from the regression analysis of 500 sets of data: k = 1-0.1×level number, M2 number is 2), and the reference pulse width T b =10ms as a reference, calculate the pulse width T after adjustment a =10×0.8=8ms, generating a PWM waveform with a duty cycle of 45% (frequency 1kHz, amplitude 12V), loading it into the motor driver module, and outputting a delay-adjusted pulse sequence [0,1,0,1,0] (level duration 8ms, interval time 8ms).

[0074] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0075] S301: Obtaining the timing data of the rising and falling cycles in the delayed adjustment pulse, extracting the start and end time from the glass starting to the current steep rise phase and the start and end time from the current returning to zero to the displacement stop phase, calculating the duration difference between the two time periods, and constructing the rising and falling response time difference according to the action direction;

[0076] Based on the delay-adjusted pulse waveform data (frequency 1kHz, sampling rate 10kHz), during the rise and fall cycle k, the current sensor collects the rising phase data points [0.5A@1250ms, 1.8A@1255ms,

[0077] The starting point of the current steep rise is defined as the first sampling point (1256ms) where the current exceeds the baseline by two standard deviations (0.2A) of 0.5A, and the end point is defined as three consecutive sampling points (1280ms) where the current stabilizes within the ±5% fluctuation range (3.5A±0.175A). The synchronous displacement sensor records the glass movement starting point (1256ms) (position 0mm→2mm) and the stopping point (1376ms) (position 54mm→54mm), and calculates the rise time Δt. up =1280-1256=24ms, drop time Δt down =1376-1352=24ms (1352ms is the turning point when the current drops from 3.5A to 1.0A), constructing the response time difference Δt diff =24-24=0ms, marking direction coefficient s=+1 (according to ISO13849-1 standard: the rising segment starts earlier than the falling segment and is marked as +1). As shown in Table 3, when the rising segment time of cycle k+1 is detected to be 28ms (1290-1262) and the falling segment time is 20ms (1380-1360), the difference Δt is calculated. diff =28-20=8ms, the direction mark remains +1, and the rising and falling response time difference sequence [0ms, 8ms] is generated.

[0078] Table 3 Timing parameter analysis table

[0079] Parameter name Rising stage Descending phase Start time (ms) 1256 (current>0.7A) 1352 (current < 3.2A) End time (ms) 1280 (current stability ±5%) 1376 (displacement change <0.1mm) Baseline duration (ms) 24 (design specification value) 24 (design specification value) Allowable deviation range (ms) ±2 (manufacturer tolerance standard) ±2 (manufacturer tolerance standard)

[0080] S302: Based on the lifting and lowering response time difference, the time difference between the lifting and lowering stages is used as the basis for determining the offset direction. Positive and negative signs are set in the order of the lifting stage first and the lowering stage last. The resulting ratio is then used as a basis for symmetry determination and classified into the lifting action category structure to obtain a symmetry offset.

[0081] Set the symmetry judgment threshold θ = 5ms (according to Article 9.2.3 of GB / T5226.1-2019 Mechanical and Electrical Safety Standard, take 20% of the design time), when the absolute value of the time difference |Δt diff When |=0ms, it is classified as C1 (completely symmetrical). When the period k+2 is detected, the rising time is 26ms and the falling time is 32ms, and the difference Δt is calculated. diff = 26-32 = -6ms, direction mark s = -1 (the descending segment starts earlier than the ascending segment), symmetry offset S o=-6×(-1)=+6ms. According to the three-level classification standard: 0-5ms is C1 (normal), 5-10ms is C2 (warning), and >10ms is C3 (fault). +6ms is classified as C2. When constructing the offset sequence, take three consecutive cycle data [0ms, +8ms, +6ms] and calculate the moving average μ=(0+8+6) / 3=4.67ms. The standard deviation When μ>5ms or σ>2ms, an asymmetry alarm is triggered.

[0082] S303: Calling the symmetry offset, calculating the offset change rate using a sliding window ratio function, and comparing the offset difference between the current cycle and the previous cycle. Synchronous growth of the symmetry offset and the current peak offset indicates a unilateral mechanical abnormality, while independent changes identify electrical interference. Determine whether the symmetry offset exceeds twice the current peak offset to obtain a symmetry offset sequence.

[0083] Define the sliding window length W = 3 cycles, calculate the change rate r = (10-(8+6) / 2) / 10 = 0.3 of the current cycle k+3 offset +10ms relative to the previous two cycles [+8ms, +6ms], and synchronously obtain the current peak offset sequence [0.05A, 0.07A, 0.18A, 0.22A] and calculate its change rate r c =(0.22-0.05) / 0.05=3.4, when r>0.2 and r c >2 (0.3>0.2 and 3.4>2), it is determined to be a mechanical unilateral abnormality. If the symmetrical offset sequence [+6ms, -8ms, +5ms] change rate r=(5-(-8+6) / 2) / 5=0.7 is detected, and the current offset is stable at 0.15A±0.03A (r c =0.03 / 0.15=0.2), it is determined to be electrical interference (PWM controller noise), and the symmetry offset is compared Offset from current peak The multiple relationship 10 / 0.22=45.45 (far exceeding the preset alarm threshold by 20 times), triggering the secondary fault code F202, generating a symmetrical offset sequence with status marks [0(+0), 8(C2), 6(C2), 10(F202)].

[0084] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0085] S401: Based on the symmetrical offset sequence and the current peak offset, construct an offset time segment by marking the ratio of the start and end times to the total cycle duration, sort the offset segments by cycle number to form a time axis index table, and generate a cycle segment offset ratio based on the time proportion distribution of the offset segments in the cycle;

[0086] Based on the symmetrical offset sequence [0ms, +8ms, +6ms] and the current peak offset [0.05A, 0.18A, 0.22A], the starting time of cycle k is extracted and the end moment Calculate the total cycle time T (k) =1376-1256=120ms, constructing the offset time segment τ (k) =8 / 120=0.0667, generate the time axis index table (Table 4) according to the cycle number k=1 to k=3, when k=2 cycle offset +8ms corresponding segment τ is detected (2) =8 / 118=0.0678, k=3 period offset + 6ms corresponding to τ (3) =6 / 122=0.0492, take the mean of three periods μ when calculating the proportional distribution τ =(0.0667+0.0678+0.0492) / 3=0.0612, standard deviation σ τ =0.0098, generating the periodic segment offset ratio sequence [0.0667, 0.0678, 0.0492].

[0087] Table 4 Time axis index table

[0088] Cycle Number Start time (ms) End time (ms) Offset Ratio 1 1256 1376 0.0667 2 1320 1438 0.0678 3 1465 1587 0.0492

[0089] S402: Calling the cycle segment offset ratio and the current peak offset, constructing the two values ​​into a data set with dual-variable attributes, inputting them into a dual-parameter K-means clustering model, initializing the cluster centers, and generating cycle offset cluster labels based on the spatial distribution of the two variables;

[0090] The two-parameter K-means clustering model includes input variable dimension, cluster number setting, cluster center initialization strategy, distance measurement method, iterative convergence criterion, and output structure;

[0091] Set the number of clusters K = 3 (according to the elbow rule SSE decline inflection point), initialize the cluster center Take the sample point p1 = (0.0667, 0.05), Among them, d p Represents the dynamic weighted distance index of the p-th sample point, R q represents the period segment deviation ratio measurement, represents the coordinates of the rth initial cluster center, P s represents the current peak offset measurement value, α represents the bivariate weight adjustment factor (0≤α≤1), β represents the variance normalization coefficient of the period segment offset ratio, and γ represents the range normalization coefficient of the current peak offset. The weighted distance to C1 is calculated as: α = 0.6, β = 0.0098 2 =0.000096, Similarly, the distance to C2 is calculated to be 2.14, and the distance to C3 is 5.62. p1 is classified into C2, and the center point is iteratively updated: C1 = (0.049, 0.05), C2 = (0.067, 0.18), and C3 = (0.072, 0.23), generating the cluster label sequence [2, 2, 1].

[0092] S403: Based on the periodic shift cluster labels, the mean of the shift ratio and peak shift of each sample in multiple label categories is extracted as the label feature benchmark and the feature distribution interval is constructed. The interval positioning is determined based on the shift mean difference ratio of multiple categories of samples and the corresponding interval to obtain the wear warning level label;

[0093] Extract the mean μ of the 3 sample offset ratios of cluster 1 R1 =0.049, peak shift mean μ P1 =0.05A, set the benchmark interval R1∈[0.04,0.06], P1∈[0.04,0.06], detect the sample point (0.0492,0.05) with the mean deviation ratio δ=|0.0492-0.049| / 0.049+|0.05-0.05| / 0.05=0, when the new sample (0.068,0.19) falls into the cluster 2 interval R2∈[0.06,0.08], P When 2∈[0.15, 0.20], the mean difference ratio δ is calculated as |0.068-0.067| / 0.067+|0.19-0.18| / 0.18=0.016+0.056=0.072, and the warning level is divided according to the threshold of 0.1: 0-0.05 is L1 (normal), 0.05-0.1 is L2 (observation), and >0.1 is L3 (warning), generating the wear label sequence [L1, L2].

[0094] See also Figure 1 The present invention provides a technical solution: a durability test method for a new energy vehicle window regulator, comprising the following steps:

[0095] S501: Based on the wear warning level label data, friction trend judgment instructions, and symmetry offset sequence input items, the data is loaded into the Drools rule engine, field values ​​are read and pre-matched, and variable binding and rule activation initialization processes are performed on the structural format of the input data to generate a basic variable set for rule triggering;

[0096] The Drools rule engine includes a knowledge base, working memory, rule structure, rule execution process, and variable binding mechanism;

[0097] Based on the wear warning label sequence [L1, L2, L3], trend instruction [0, 1, 1], and symmetric offset sequence [0ms, +8ms, +6ms], create a Fact object fact1 = {label: L2, trend: 1, offset: +8} in the Drools rule engine, bind variables $var1 = fact1.label and $var2 = fact1.trend, activate rule R001 in the rule base, and trigger when $var1 = L2 and $var2 = 1 ,Initialize the working memory to load the sample data set [{L2,1,8},{L3,1,6}], set the sliding window size N=5, and when the number of times the rule is triggered within the window exceeds the threshold 3 times, generate the basic variable set {label-count: 2, trend-sum: 2, offset-avg: 7}, as shown in Table 5. When the fourth sample {L3,1,10} is detected, the variable set is updated to {label-count: 3, trend-sum: 3, offset-avg: 8}.

[0098] Table 5 Rule trigger variable table

[0099] variable name Numerical Calculation rules label-count 2 Count the number of L2 / L3 label occurrences trend-sum 2 The trend instruction is the accumulated value of 1 offset-avg 7 Arithmetic mean of offset

[0100] S502: Invoke the rule to trigger multiple numerical variables in the basic variable set, calculate the mean and difference by field category, complete the discretization mapping of all variables through Z-score normalization, and generate a normalized offset fluctuation series;

[0101] Select the basic variable set {label-count: 3, trend-sum: 3, offset-avg: 8}, calculate the sliding window mean of field m=1, (label-count) Standard deviation The adjacent window difference δ1 = 3-2 = 1, the correlation factor θ1 = 0.8, (based on the Pearson correlation coefficient between fields), take ξ = 2.0, (empirical constant), and calculate the standardized eigenvalue For the field m=2, (trend-sum) calculates μ2=2.5, σ2=0.5, and generates the normalized sequence [1.447, 1.0, 1.414].

[0102] S503: Based on the normalized offset fluctuation sequence, continuous time periods related to the differential trend and fluctuation trajectories of the symmetrical offset are screened to determine whether the offset increment directions of the two data items in multiple consecutive segments simultaneously and continuously change in a positive direction. If the conditions are met, the tag level field is updated.

[0103] Analyze the normalized sequence [1.447, 1.0, 1.414] and its trend over three consecutive cycles. Calculate the increments for cycles 1-3: Δ1 = 1.0-1.447 = -0.447 and Δ2 = 1.414-1.0 = 0.414. When the number of consecutive increments (Δ > 0) reaches 2, check the symmetrical offset sequence [+8ms, +6ms, +10ms] for increments of +6-8 = -2 and +10-6 = +4. Check whether both data points meet the requirement of 2 positive changes over the past three cycles. Currently, only the offset sequence meets the requirement, and the label level is delayed to L2. If both the normalized increment and the offset increment meet the requirement of +0.5 and +3, the label level is upgraded to L3.

[0104] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A durability test method for a new energy vehicle window regulator, characterized in that: The following steps are involved: S1: Obtain the lifter motor load current change rate and the glass sliding speed and calculate the ratio. Calculate the difference of the ratio of consecutive cycles and substitute it into the sliding window analysis method. The larger the window, the smaller the differential trend judgment threshold. Generate a trend judgment instruction. S2: Based on the trend judgment instruction, the current peak offset between the peak amplitude of the motor load current in the current cycle and the previous cycle is calculated, and the current peak offset is substituted into the fuzzy logic classifier in combination with the environmental disturbance compensation mechanism to perform dynamic matching and divide the amplitude into levels, and generate a delay adjustment pulse according to the delay mapping rule corresponding to the level; S3: The delay adjustment pulse is used to collect the time from glass start-up to current steep rise and the time from current zero return to displacement stop in the lifting and lowering cycle, and the symmetric offset of the time is calculated according to the direction segment. Based on the phase locking phenomenon between the symmetric offset and the current peak offset, the data with the offset growth exceeding two times are selected to form a symmetric offset sequence; S4: Based on the symmetric offset sequence and the current peak offset, the offset segment position in the cycle is compared, and a two-parameter K-means clustering model is input to perform clustering according to the variable dimension and the number of clusters, and the wear warning level label is output.

2. The durability testing method for new energy vehicle window regulator according to claim 1, characterized in that: The trend judgment instruction specifically includes a differential trend threshold, an increasing trend state, and a sliding window analysis result. The delay adjustment pulse includes a current peak offset, an amplitude level classification result, and a delay mapping rule. The symmetry offset sequence specifically refers to a time symmetry offset, a sliding window proportion assessment result, and a two-fold growth screening condition. The wear warning level label includes a segment position comparison parameter, a dual-parameter clustering model output, and a wear level classification standard.

3. The durability testing method for new energy vehicle window regulator according to claim 2, characterized in that: Obtain the lifter motor load current change rate and the glass sliding speed and calculate the ratio. Calculate the difference of the ratio data series of consecutive cycles and substitute it into the sliding window analysis method for judgment. If both are greater than the differential trend judgment threshold and show an increasing trend, the specific steps to generate the trend judgment instruction are as follows: S101: Obtaining the rate of change of the lifter motor load current and the glass sliding speed, respectively calculating the change value of the motor load current and the change value of the glass sliding distance, and calculating the ratio within the same cycle to generate a ratio sequence; S102: Based on the ratio sequence, call data segments within three consecutive periods, construct a differential data group and a continuous window based on the difference between the ratios of two adjacent periods, arrange and analyze multiple differential data within the window, and generate a differential trend interval; S103: For the differential trend interval, make a judgment according to the differential trend determination threshold, compare the data trend change direction and the incrementality in multiple windows one by one, adjust the differential trend determination threshold according to the window size, lower the threshold for large windows and increase it for small windows, and judge whether it satisfies both the differential trend determination threshold and the continuous incremental direction, and generate a trend judgment instruction.

4. The durability testing method for new energy vehicle window regulator according to claim 3, characterized in that: Based on the trend judgment instruction, the current peak offset between the peak amplitude of the motor load current in the current cycle and the peak amplitude of the previous cycle is calculated, and the current peak offset is substituted into the fuzzy logic classifier to divide the amplitude level, and the specific steps of generating the delay adjustment pulse according to the delay mapping rule corresponding to the level are as follows: S201: Based on the trend judgment instruction, calling the maximum instantaneous current value in the motor load current signal of the current cycle and the previous cycle, calculating the difference between the maximum instantaneous current values ​​in the two cycles, and comparing it with the current amplitude change reference value corresponding to the current cycle acquisition period to obtain the current peak offset; S202: Based on the current peak offset, calling the amplitude boundary value in the fuzzy logic classifier classification rule, classifying the corresponding position of the current peak offset into the target amplitude level interval, and performing a mapping operation with the hierarchical structure set by the fuzzy logic classifier to generate an amplitude level identifier; S203: calling the amplitude level identifier, and loading the corresponding pulse control instruction within a period according to the pulse adjustment coefficient corresponding to the identifier in the delay mapping rule table preset by the fuzzy logic classifier to generate a delay adjustment pulse.

5. The durability testing method for new energy vehicle window regulator according to claim 4, characterized in that: The delay adjustment pulse is used to collect the time from glass start-up to current steep rise and the time from current zero return to displacement stop in the lifting and lowering cycle, and the symmetric offset of the time is calculated according to the direction segment. The sliding window proportion function is applied to evaluate, and the current peak offset is used as a benchmark to screen the data with an offset increase of more than twice to form a symmetric offset sequence. The specific steps are as follows: S301: Obtaining the timing data of the rising and falling periods in the delay adjustment pulse, extracting the start and end time from the glass starting to the current steep rise stage and the start and end time from the current returning to zero to the displacement stop stage, calculating the duration difference between the two time periods, and constructing the rising and falling response time difference according to the action direction; S302: Based on the lifting and lowering response time difference, the time difference between the lifting and lowering stages is used as the basis for determining the offset direction, with the positive and negative signs set in the order of the lifting stage first and the lowering stage last. The resulting ratio is then used as a basis for symmetry determination and classified into the lifting and lowering action category structure to obtain a symmetry offset. S303: Call the symmetry offset, calculate the offset change rate through the sliding window proportion function, compare the offset difference between the current cycle and the previous cycle, the synchronous growth of the symmetry offset and the current peak offset indicates a mechanical unilateral abnormality, and independent changes distinguish electrical interference, determine whether it exceeds twice the current peak offset, and obtain the symmetry offset sequence.

6. The durability testing method for new energy vehicle window regulator according to claim 5, characterized in that: Based on the symmetrical offset sequence and the current peak offset, the offset segment positions in the cycle are compared, and a dual-parameter K-means clustering model is input for clustering. The specific steps for outputting the wear warning level label are as follows: S401: Based on the symmetrical offset sequence and the current peak offset, construct an offset time segment by marking the ratio of the start and end times to the total cycle duration, form a time axis index table of the offset segments by sorting them by cycle number, and generate a cycle segment offset ratio based on the time proportion distribution of the offset segments in the cycle; S402: Calling the cycle segment offset ratio and the current peak offset, constructing the two numerical values ​​into a data set with dual-variable attributes, inputting them into a dual-parameter K-means clustering model, initializing the cluster centers, and generating cycle offset cluster labels based on the spatial distribution of the two variables; The dual-parameter K-means clustering model includes input variable dimension, cluster number setting, cluster center initialization, distance measurement method, iterative convergence criterion, and output structure; S403: Based on the periodic offset clustering labels, extract the mean of the offset ratio and peak offset of each sample in multiple label categories as the label feature benchmark and construct a feature distribution interval. Perform interval positioning judgment based on the ratio of the mean offset difference between multiple categories of samples and the corresponding interval to obtain the wear warning level label.

7. The durability testing method for new energy vehicle window regulator according to claim 6, characterized in that: The input is put into the two-parameter K-means clustering model and the cluster center is initialized using the formula: Among them, d p Represents the dynamic weighted distance index of the p-th sample point, R q represents the period segment deviation ratio measurement, represents the coordinates of the rth initial cluster center, P s represents the current peak offset measurement value, α represents the bivariate weight adjustment factor (0≤α≤1), β represents the variance normalization coefficient of the cycle segment offset ratio, and γ represents the range normalization coefficient of the current peak offset.

8. The durability testing method for new energy vehicle window regulators according to claim 6, characterized in that: The method further comprises: S5: Input the wear warning level label data, trend judgment instruction, and symmetry offset sequence into the Drools rule engine and perform Z-score normalization processing. If the difference trend and symmetry offset trend continue to rise and the peak fluctuation is abnormal, the label level is automatically corrected; The automatically corrected label levels are specifically Z-score standardized data, continuous trend judgment indicators, and peak fluctuation anomalies.

9. The durability testing method for new energy vehicle window regulators according to claim 8, characterized in that: The wear warning level label data, trend judgment instructions, and symmetry offset sequence are input into the Drools rule engine and Z-score normalization is performed. If the differential trend and symmetry offset trend continue to rise and the peak fluctuation is abnormal, the label level is automatically corrected as follows: S501: Based on the wear warning level label data, friction trend judgment instruction and symmetry offset sequence input items, the data is loaded into the Drools rule engine, field values ​​are read and pre-matched, variable binding and rule activation initialization processes are performed on the structural format of the input data, and a rule triggering basic variable set is generated; The Drools rule engine includes a knowledge base, a working memory, a rule structure, a rule execution process, and a variable binding mechanism; S502: Call the rule to trigger multiple numerical variables in the basic variable set, calculate the mean and difference respectively according to the field category, complete the discretization mapping of all variables through the Z-score normalization method, and generate a normalized offset fluctuation series; S503: Based on the normalized offset fluctuation sequence, the continuous time periods related to the differential trend and the fluctuation trajectory of the symmetrical offset are screened to determine whether the offset increment directions of the two data in multiple continuous segments continue to change in a positive direction at the same time. If the condition is met, the label level field is updated.

10. The durability testing method for new energy vehicle window regulators according to claim 9, characterized in that: The discretization mapping of all variables is completed by Z-score standardization, using the formula: Among them, θ mn represents the weighted normalized feature value of the nth sample in the mth field, X mn Represents the original numerical variable measurement value, μ m represents the sliding window mean of the mth field, σ m Represents the dynamic standard deviation of the mth field, δ m represents the difference between adjacent data windows of the mth field, θ n represents the cross-field correlation factor of the nth sample, and ξ represents the global normalization reference constant.

Images