New energy automobile damper state intelligent early warning method based on attitude monitoring

Abstract

The invention discloses a new energy automobile shock absorber state intelligent early warning method based on attitude monitoring, and relates to the field of state intelligent early warning, and the method comprises the steps: collecting multi-dimensional attitude data such as the vertical acceleration, the pitch angle speed and the roll angle speed of an automobile in real time, and carrying out the self-adaptive detection of an impact event through combining with a dynamic threshold algorithm; and normal driving vibration and abnormal impact working conditions can be accurately distinguished. In the feature extraction stage, multi-dimensional physical features such as a peak-to-peak value, a root-mean-square value, a damping ratio and the like are combined and coded, so that the dynamic attenuation characteristics of the shock absorber in an impact event can be comprehensively represented. By establishing a similar event feature database in a healthy state and performing multi-feature space deviation calculation by adopting a mahalanobis distance, the robustness of anomaly detection is remarkably improved. Therefore, the technical problems of poor environmental adaptability, low diagnosis precision and the like of the traditional method can be effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of state intelligent early warning, and more specifically, to a new energy vehicle shock absorber state intelligent early warning method based on posture monitoring. BACKGROUND

[0002] As an important direction of industrial transformation, the driving comfort and safety of new energy vehicles are of great concern, and the shock absorber as a key component of the suspension system directly affects these core performances. Due to the particularity of the battery layout and the weight distribution of the whole vehicle, higher requirements are put forward for the performance and durability of the shock absorber. The shock absorber will gradually age due to vibration and impact in long-term use, and its performance degradation not only affects the comfort, but also may bring safety hazards such as decreased handling and increased braking distance. The traditional detection method relies on periodic manual inspection or maintenance after failure, which has a lag and is difficult to achieve early warning and preventive maintenance, which easily leads to increased maintenance costs and even safety accidents. Therefore, it is urgent to develop a method that can monitor the state of the shock absorber in real time and intelligently and has fault early warning capability.

[0003] In the existing research on shock absorber state evaluation based on vehicle posture monitoring, there are still many technical challenges. First, in the actual driving process, the road conditions are complex and changeable, such as flat, bumpy, and potholed, which will cause significant changes in posture. These changes caused by road conditions are difficult to distinguish from posture abnormalities caused by performance degradation of the shock absorber. Second, the individualized driving behavior of the driver, such as sudden acceleration, sudden braking, high-speed cornering, etc., will also have a strong impact on the vehicle posture. In addition, changes in vehicle parameters such as load and tire pressure will also introduce interference factors, making the features extracted from the posture data mixed with a large amount of information unrelated to the state of the shock absorber. Especially when the shock absorber has early and slight performance degradation, the changes it causes are often weak and easily masked by the above-mentioned interference, making it difficult to extract fault features, and the accuracy and robustness of the early warning model are insufficient, which is prone to false positives or false negatives. More importantly, in the real driving environment, there is a lack of standardized and "pure" road excitation signals as test benchmarks, making it difficult for methods based on fixed thresholds or simple models to adapt to diverse working conditions, and it is difficult to establish a stable and reliable health state reference system, thereby affecting the timeliness and accuracy of the shock absorber state judgment.

[0004] Therefore, how to effectively strip the influence of road conditions and driving behavior, accurately identify and quantify the feature changes caused by performance degradation of the shock absorber from the vehicle posture data, is a key problem to be solved for realizing intelligent early warning of the state of the shock absorber of new energy vehicles. SUMMARY

[0005] In order to solve the above technical problems, the present application is proposed.

[0006] According to an aspect of the present application, a new energy vehicle shock absorber state intelligent early warning method based on posture monitoring is provided, which comprises:

[0007] Obtaining a vehicle posture data time series data stream, the vehicle posture data comprising vehicle vertical acceleration, pitch angular velocity and roll angular velocity;

[0008] Extracting a vehicle vertical acceleration time series data stream from the vehicle posture data time series data stream, and performing impact event detection on the vehicle vertical acceleration time series data stream based on a preset rule;

[0009] In response to detecting an impact event, intercepting vehicle posture data time series data streams of a preset time period before and after the impact event to obtain a vehicle posture data event segment;

[0010] Extracting event vertical response features, event pitch response features and event roll response features from the vehicle posture data event segment, and combining the event vertical response features, the event pitch response features and the event roll response features to obtain an event vehicle posture response feature encoding vector;

[0011] Based on the average speed value of the impact event and the event type of the impact event, querying a corresponding health feature benchmark encoding vector from a same event feature database in a healthy state;

[0012] Calculating the deviation value between the event vehicle posture response feature encoding vector and the health feature benchmark encoding vector;

[0013] Based on the deviation value, determining whether to generate a new energy vehicle shock absorber state abnormality early warning prompt.

[0014] Compared with the prior art, the new energy vehicle shock absorber state intelligent early warning method based on posture monitoring provided by the present application can accurately distinguish normal driving vibration from abnormal impact working conditions by real-time collection of multi-dimensional posture data such as vehicle vertical acceleration, pitch angular velocity and roll angular velocity, and adaptive detection of impact events by combining a dynamic threshold algorithm. In the feature extraction stage, a combination encoding of multi-dimensional physical features such as peak-to-peak value, root mean square value and damping ratio is used to comprehensively represent the dynamic attenuation characteristics of the shock absorber in the impact event, overcoming the defect of insufficient sensitivity of single feature. By establishing a same event feature database in a healthy state and using Mahalanobis distance to calculate the deviation in the multi-feature space, not only the statistical correlation between features is considered, but also the benchmark response mode under different speeds and impact types is adaptively matched, significantly improving the robustness of abnormality detection. The overall scheme forms a closed-loop diagnosis logic through multi-dimensional data fusion, dynamic event detection and intelligent feature comparison, and can effectively solve the technical pain points of poor environmental adaptability and low diagnosis accuracy of traditional methods. BRIEF DESCRIPTION OF DRAWINGS

[0015] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description thereof taken in conjunction with the accompanying drawings, in which:

[0016] Figure 1 Flow chart of the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application.

[0017] Figure 2 Data flow chart of the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application.

[0018] Figure 3 Flow chart of step S2 in the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application.

[0019] Figure 4 Flow chart of step S4 in the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] Embodiments of the present disclosure will be described in more detail with reference to the drawings. While the present disclosure is shown and described in connection with certain embodiments thereof, it is to be understood that the present disclosure is capable of further minor modifications, and that this application is intended to cover any and all such variations and modifications as fall within the scope of the present disclosure. Accordingly, it is to be understood that the drawings are for illustrative purposes and are not to scale.

[0021] To solve the problems in the background art, the present application provides a method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring. Figure 1 Flow chart of the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application. Figure 2 Data flow chart of the method for intelligent early warning of the state of a new energy vehicle shock absorber based on posture monitoring according to an embodiment of the present application. As Figure 1 and Figure 2As shown, the new energy vehicle shock absorber state intelligent early warning method based on posture monitoring according to the embodiment of the application comprises: S1, acquiring a vehicle posture data time series data stream, the vehicle posture data comprising vehicle vertical acceleration, pitch angular velocity and roll angular velocity; S2, extracting a vehicle vertical acceleration time series data stream from the vehicle posture data time series data stream, and performing impact event detection on the vehicle vertical acceleration time series data stream based on a preset rule; S3, in response to detecting an impact event, intercepting vehicle posture data time series data streams of preset time periods before and after the impact event to obtain a vehicle posture data event segment; S4, extracting an event vertical response feature, an event pitch response feature and an event roll response feature from the vehicle posture data event segment, and combining the event vertical response feature, the event pitch response feature and the event roll response feature to obtain an event vehicle posture response feature coding vector; S5, based on an average vehicle speed value of the impact event and an event type of the impact event, querying a corresponding health feature benchmark coding vector from a same-class event feature database in a healthy state; S6, calculating a deviation value between the event vehicle posture response feature coding vector and the health feature benchmark coding vector; and S7, based on the deviation value, determining whether to generate a new energy vehicle shock absorber state abnormal early warning prompt.

[0022] In step S1, a vehicle posture data time series data stream is acquired, and the vehicle posture data comprises vehicle vertical acceleration, pitch angular velocity and roll angular velocity. In particular, the vehicle posture data time series data stream is collected by an IMU sensor. It can be understood that the performance state of the vehicle shock absorber is directly reflected in the dynamic response of the vehicle body during vehicle driving. The vehicle vertical acceleration can directly represent the vibration of the vehicle body in the vertical direction and is the most direct parameter for feeling the road impact and the filtering effect of the shock absorber. The pitch angular velocity describes the speed of rotation of the vehicle around the horizontal axis and reflects the response characteristics of the front and rear suspensions and shock absorbers under longitudinal impact such as passing through a speed bump, braking and accelerating. The roll angular velocity describes the speed of rotation of the vehicle around the vertical axis and reflects the response characteristics of the left and right suspensions and shock absorbers under lateral imbalance such as turning and passing through a single-sided bump. The three parameters together constitute the key information for describing the dynamic response of the vehicle posture under impact events. Therefore, by collecting these data through the IMU sensor, three-axis acceleration and three-axis angular velocity can be output in real time at a high frequency and high accuracy, thereby providing high-quality original data streams for subsequent impact event detection and feature extraction.

[0023] In detail, step S1 can be acquired by the following way: first, an IMU sensor needs to be planned and installed on the new energy vehicle. The IMU sensor is preferably installed at a position close to the vehicle centroid on the vehicle chassis, for example, the central floor under the B pillar of the vehicle, to ensure that its measurement results can better reflect the main motion posture of the vehicle. When installing, the three orthogonal sensitive axes of the IMU need to be aligned with the coordinate system of the vehicle, or its installation posture is known and fixed relative to the vehicle coordinate system, so as to carry out accurate coordinate conversion subsequently. For example, the X axis of the IMU can be set along the vehicle longitudinal direction (the forward direction is positive), the Y axis along the vehicle transverse direction (the left side is positive), and the Z axis along the vehicle vertical direction (upward is positive).

[0024] After completing the physical installation, the IMU sensor is configured with parameters. A key preset parameter is the data acquisition frequency. In order to ensure that the rapid response characteristics of the shock absorber to the road impact can be captured, the sampling frequency should not be too low. For example, the sampling frequency of the IMU can be set to 100 Hz, which means that every 10 milliseconds, the IMU will output a complete set of measurement data. After the IMU is started, the accelerometer and gyroscope inside the IMU will measure the three-axis linear acceleration and three-axis angular velocity raw data of the vehicle in the IMU's own coordinate system in real time.

[0025] Next, the acquired IMU raw data is processed to obtain the required vehicle posture data. The vehicle vertical acceleration is directly taken from the acceleration reading of the IMU Z axis, if the Z axis of the IMU has been accurately aligned with the vertical direction of the vehicle. In some cases, compensation for the gravity component may be required, especially when the analysis focuses on dynamic acceleration changes, which can be achieved by high-pass filtering or subtracting the Z axis reading in the initial static state (about 9.8 m / s2). The pitch angular velocity refers to the angular velocity of the vehicle rotating around its transverse axis, for example, the Y axis, which can be obtained from the angular velocity reading of the corresponding axis output by the IMU. Similarly, the roll angular velocity refers to the angular velocity of the vehicle rotating around its longitudinal axis, for example, the X axis, which can be obtained from the angular velocity reading of the corresponding axis output by the IMU. Through such processing, at each sampling time point t, a data triplet can be obtained, i.e., the vehicle vertical acceleration value at that time, the pitch angular velocity value at that time, and the roll angular velocity value at that time. The sequential data triplet obtained by sequentially collecting and processing these data can be serialized to form the final output vehicle posture data time series data stream, which serves as the input for subsequent impact event detection and feature analysis.

[0026] In step S2, a vehicle vertical acceleration time series data stream is extracted from the vehicle pose data time series data stream, and shock event detection is performed on the vehicle vertical acceleration time series data stream based on preset rules. Accordingly, since the vehicle vertical acceleration is the most direct and sensitive physical indicator for monitoring the vehicle body response when the vehicle encounters road surface shocks such as driving over speed bumps or potholes. Therefore, in order to accurately identify specific road excitation events that are of important diagnostic value for evaluating the performance of shock absorbers, the vehicle vertical acceleration time series data stream needs to be extracted from the vehicle pose data time series data stream and shock event detection needs to be performed. That is, by focusing the analysis on the response of these shock events, the complex and continuous driving conditions can be effectively simplified into relatively standardized and concentrated excitation segments, so that the dynamic behavior of the shock absorber under specific shocks can be more clearly observed and quantified, providing a key data basis and analysis window for subsequent accurate feature extraction and state judgment.

[0027] In detail, the extraction of the vehicle vertical acceleration time series data stream from the vehicle pose data time series data stream can be achieved by accessing each data record in the vehicle pose data time series data stream one by one. For each data record containing three pose components, only the value representing the vehicle vertical acceleration is selected and taken out. All the selected vehicle vertical acceleration values at different times are arranged in strict accordance with their original time sequence, thereby forming a target output vehicle vertical acceleration time series data stream with a single dimension. For example, if the data record in the input stream at a certain sampling time t is (vehicle vertical acceleration value at this time, pitch angular velocity value at this time, roll angular velocity value at this time), after the extraction operation, the value corresponding to time t in the output vehicle vertical acceleration time series data stream is the vehicle vertical acceleration value at this time. The output data stream will be directly used for subsequent shock event detection.

[0028] Specifically, in one specific example of the present application, Figure 3 The flowchart of step S2 in the new energy vehicle shock absorber state intelligent early warning method based on pose monitoring according to the embodiment of the present application. As shown in Figure 3As shown, in step S2, the impact event detection is performed on the vehicle vertical acceleration time series data stream based on a preset rule, including: S21, inputting the current vehicle speed value into a dynamic threshold function to obtain a current dynamic threshold; S22, determining a data point in the vehicle vertical acceleration time series data stream that first exceeds the current dynamic threshold, and marking the data point as a start point of a potential event; S23, performing backward traversal along the start point of the potential event on the vehicle vertical acceleration time series data stream, and determining whether the absolute value of a positive peak value and the absolute value of a negative peak value are both greater than the current dynamic threshold; S24, in response to the absolute value of the positive peak value and the absolute value of the negative peak value being both greater than the current dynamic threshold, calculating a pulse width based on a time stamp of the positive peak value and a time stamp of the negative peak value; S25, when the pulse width meets a pulse width condition, marking the potential event as an impact event.

[0029] It can be understood that when a vehicle passes through the same road unevenness such as a deceleration strip or a pothole at different speeds, the amplitude of the vertical acceleration response generated thereby is significantly different. Normally, the higher the vehicle speed, the greater the vertical acceleration peak value caused by the impact. If a fixed threshold is used to detect impact events, the threshold may be too high when driving at low speed, resulting in real impact events being missed; and the threshold may be too low when driving at high speed, causing some road disturbances that are not typical impacts to be misjudged as impact events. Therefore, by introducing a dynamically adjusted threshold with the vehicle speed, the detection standard of impact events can be adaptively changed according to the instantaneous driving state of the vehicle, thereby improving the accuracy and robustness of impact event identification, and effectively distinguishing real impact events that are meaningful for shock absorber performance evaluation from normal driving vibrations or slight road unevenness.

[0030] In detail, step S21 can be implemented in the following manner: first, the current vehicle speed value needs to be acquired in real time from a vehicle data bus such as a CAN bus or other sensors such as a GPS module, which is used as the input of the dynamic threshold function. After the current vehicle speed value is acquired, it is substituted into a pre-set dynamic threshold function. This function defines the quantitative relationship between the vertical acceleration threshold for impact detection and the vehicle driving speed. The specific form and internal parameters of this function, such as the reference threshold, the speed sensitivity coefficient, and the speed segmentation points, are pre-calibrated and determined based on a large amount of actual road test data, vehicle dynamics simulation analysis, or in combination with the characteristics of a specific vehicle model.

[0031] For example, the dynamic threshold function processes as follows according to the preset dynamic additional function rule: when the input current vehicle speed value is less than 20 kilometers per hour, the current dynamic threshold is directly set to a fixed value, for example, the current dynamic threshold is equal to 0.4 times the gravitational acceleration (note: 1 times the gravitational acceleration is approximately equal to 9.8 meters per second squared). When the input current vehicle speed value is greater than or equal to 20 kilometers per hour, the current dynamic threshold (in units of times the gravitational acceleration) is calculated by the following formula: current dynamic threshold = 0.4 + 0.01 x (current vehicle speed value - 20). For example, the current vehicle speed value is 40 kilometers per hour, then the calculation of the current dynamic threshold is: 0.4 + 0.01 x (40 - 20) finally gets 0.6 times the gravitational acceleration, that is, the current dynamic threshold.

[0032] Correspondingly, the vehicle will continue to generate vibration during driving, but not all vibrations are related to significant road impact such as speed bumps, potholes. By setting a current dynamic threshold that dynamically adjusts with vehicle speed, general vibrations in daily driving can be effectively filtered out, and only those impacts with large enough acceleration amplitude that may test the shock absorber are concerned. For this purpose, to lock a clear starting point for subsequent accurate impact event analysis and shock absorber state evaluation, the present application needs to mark the starting point of the potential event, that is, to select the data point that first exceeds the threshold as the starting point of the potential event, in order to capture the initial moment of the impact event, which is crucial for subsequent analysis of the complete waveform of the impact response, calculation of pulse width, extraction of transient response features, etc. This marking point frames the starting boundary of the time window for subsequent refined analysis.

[0033] In detail, step S22 can be implemented in the following way: Firstly, the time series data stream of vehicle vertical acceleration is checked point by point in time sequence. For each value of vertical acceleration in the data stream, it is compared with the current dynamic threshold value, which is 0.6 times of gravity acceleration, which has been calculated. When the first data point in the sequence is monitored, whose value is greater than 0.6 times of gravity acceleration, the data point is determined. The time series data stream of vehicle vertical acceleration (unit has been converted to times of gravity acceleration) presents the following values in a certain time period, covering the typical positive and negative parts of the impact response: time stamp T1 : vehicle vertical acceleration = 0.1 times of gravity acceleration; time stamp T2: vehicle vertical acceleration = 0.3 times of gravity acceleration; time stamp T3: vehicle vertical acceleration = 0.5 times of gravity acceleration (acceleration starts to rise significantly); time stamp T4: vehicle vertical acceleration = 0.7 times of gravity acceleration (this value 0.7 times of gravity acceleration has exceeded the current dynamic threshold value 0.6 times of gravity acceleration); time stamp T5: vehicle vertical acceleration = 1.1 times of gravity acceleration (reaches the positive peak); time stamp T6: vehicle vertical acceleration = 0.4 times of gravity acceleration; time stamp T7: vehicle vertical acceleration = -0.2 times of gravity acceleration; time stamp T8: vehicle vertical acceleration = -0.8 times of gravity acceleration (reaches the negative peak); time stamp T9: vehicle vertical acceleration = -0.3 times of gravity acceleration. In the above sequence, when the algorithm checks time stamp T4, its corresponding vertical acceleration value 0.7 times of gravity acceleration is the first value exceeding 0.6 times of gravity acceleration. Therefore, this data point containing time stamp T4 and its corresponding vertical acceleration value 0.7 times of gravity acceleration will be marked as the starting point of the potential event. Even though the value 1.1 times of gravity acceleration of subsequent time stamp T5 is far beyond the threshold, but since the data point of time stamp T4 is the first one to meet the condition, T4 is selected. The starting point of this potential event (containing its time information and acceleration value) is the output of this step, which is used for subsequent impact event confirmation and detailed feature analysis.

[0034] It should be appreciated that a typical impact caused by a significant road unevenness will cause the vehicle to experience a complete oscillation process, i.e. a rapid compression of the suspension followed by a rebound, which is reflected in the vertical acceleration signal as a sequence of positive and negative peaks of significant magnitude. By setting this double peak validation condition, which requires that both the positive and negative peaks must exceed the current dynamic threshold, which is dynamically adjusted according to the vehicle speed, it is possible to effectively exclude pseudo events or minor disturbances, which, although having an initial amplitude exceeding the threshold, are not complete or significant impacts, e.g. only a one-way impact or a subsequent vibration quickly decaying. This significantly improves the quality and reliability of the identified impact events, ensuring that the subsequent analysis, such as the shock absorber status evaluation, is based on real and representative impact data. Then, considering the pulse width, defined as the duration of the main energy concentration during the impact process, e.g. the time interval from the appearance of the positive peak to the appearance of the negative peak, is an important parameter for describing the impact event morphology and judging the response characteristics of the shock absorber. The degradation of the shock absorber performance can cause the vibration duration after the impact to be longer. Therefore, accurately calculating the pulse width provides a key quantitative indicator for subsequent analysis of shock absorber damping characteristics, vibration decay rate, etc., and is an important part of intelligent early warning.

[0035] In detail, steps S23 and S24 can be implemented by first performing S23: inputting the start point of the potential event determined in the previous step, which is time stamp T4 and its corresponding vertical acceleration value of 0.7g, the complete vehicle vertical acceleration time series data stream, i.e. time stamps T1 to T9 and their corresponding vertical acceleration values, and the current dynamic threshold, for example, the above-mentioned setting of 0.6g.

[0036] The processing process starts from the start point (T4) of the potential event, and analyzes the data points in the (later in time) vehicle vertical acceleration time series data stream. The algorithm looks for the first local maximum after T4 as the positive peak, and looks for the first local minimum after the positive peak as the negative peak. According to the data output above: traverse the data stream from T4 backward (forward in time): time stamp T5: vehicle vertical acceleration equals 1.1g. This is a local maximum after T4, recorded as a positive peak, with a value of +1.1g. Continue to traverse to find the negative peak: time stamp T6: vehicle vertical acceleration equals 0.4g. Time stamp T7: vehicle vertical acceleration equals -0.2g. Time stamp T8: vehicle vertical acceleration equals -0.8g. Time stamp T8 is a local minimum, recorded as a negative peak, with a value of -0.8g. Time stamp T9: vehicle vertical acceleration = -0.3g.

[0037] That is, after T4, a positive peak occurs at timestamp T5 with a value of +1.1 times gravity. After this positive peak, a negative peak occurs at timestamp T8 with a value of -0.8 times gravity. Then, a decision is made that the absolute value of the positive peak, i.e., |+1.1 times gravity| = 1.1 times gravity, is greater than the current dynamic threshold of 0.6 times gravity. The absolute value of the negative peak, i.e., |-0.8 times gravity| = 0.8 times gravity, is also greater than the current dynamic threshold of 0.6 times gravity. Since both conditions are met, the output of S23 is a logical true decision.

[0038] Subsequently, if the decision of S23 is true, S24 is executed: the input includes the true decision of S23, the timestamp of the recognized positive peak (timestamp T5), and the timestamp of the negative peak (timestamp T8). The process is to calculate the time difference between these two timestamps. That is, the pulse width = timestamp T8 - timestamp T5. For example, if T5 is 10.15 seconds and T8 is 10.25 seconds, the pulse width is 0.10 seconds. This pulse width will be used for subsequent feature analysis.

[0039] Accordingly, the pulse width reflects the main duration of the impact excitation. Thus, in order to further confirm the potential impact event that has passed the preliminary amplitude and completeness (two-way peak) screening, the present application makes a conditional decision on the above pulse width. That is, the pulse width condition is to further distinguish the real impact caused by typical road unevenness such as speed bump, standard pothole, etc. from those vibration events that, although significant in amplitude, are abnormal in duration (too short can mean sharp transient noise or very irregular impact, too long can mean sustained jolt rather than single impact). Only when the pulse width falls within a pre-defined reasonable range, the potential event is considered as a typical impact event that is meaningful for shock absorber status evaluation.

[0040] In detail, step S25 can be implemented in the following way: Firstly, a pre-set pulse width condition is needed. This pulse width condition is defined as a time range, for example, a minimum pulse width and a maximum pulse width, which are needed to be calibrated and pre-set according to the vehicle type, typical road impact characteristics and experimental data. For example, through a large number of driving experiments, it can be found that for the target vehicle type, the pulse width (for example, defined as the time difference between the positive and negative peaks) of the impact event generated by the standard deceleration strip is between 0.05 seconds and 0.3 seconds. Therefore, the pre-set pulse width condition can be: the pulse width is greater than or equal to 0.05 seconds and the pulse width is less than or equal to 0.3 seconds. The calculated pulse width is compared with this pre-set pulse width condition. If the pulse width meets the condition, the potential event of the occurrence of the event is formally marked as an impact event. Following the example in the foregoing, the pulse width calculated by step S24 is 0.10 seconds. The pre-set pulse width condition is: the pulse width is greater than or equal to 0.08 seconds and the pulse width is less than or equal to 0.2 seconds. The judgment process is as follows: compare the pulse width 0.10 seconds with the pulse width condition: 0.10 seconds is greater than or equal to 0.08 seconds and 0.10 seconds is less than or equal to 0.20 seconds. Since both sub-conditions are met, the pulse width 0.10 seconds meets the pre-set pulse width condition. Therefore, the initially identified potential event, i.e. the event represented by the data segment containing time stamps T4 to T9, is formally marked as a valid impact event. This impact event and its related parameters, such as peak value, pulse width, occurrence time, etc. will be used as the output of this step for subsequent shock absorber monitoring and early warning analysis. If the calculated pulse width is not within this range, the potential event will be discarded and not marked as an impact event.

[0041] In step S3, in response to detecting the impact event, the vehicle attitude data time series data stream of a pre-set time period before and after the impact event is intercepted to obtain a vehicle attitude data event segment. It should be understood that only a few characteristic points of the impact event itself, such as the peak value, the pulse width, is not enough to fully depict the performance of the shock absorber. The attitude change of the vehicle before and after the impact, especially the vibration attenuation characteristics after the impact, contains key information about the damping, support capacity, etc. of the shock absorber. Therefore, it is necessary to intercept a continuous data segment containing a long enough time before the impact as a reference or transition, at the core response when the impact occurs and the decay process after the impact. This segment, i.e. the vehicle attitude data event segment, provides a complete data basis for subsequent accurate calculation of multi-dimensional vibration response characteristics such as peak-to-peak value, root mean square, decay time, etc.

[0042] In detail, step S3 can be implemented by inputting the impact event confirmed as valid in the previous step, which contains at least one explicit time reference point, such as its start time stamp, e.g. T4 in the above steps, and the continuously collected vehicle attitude data time series data stream. In addition, two key preset parameters are also needed: the pre-set time period before the impact event and the pre-set time period after the impact event. The length of these two pre-set time periods needs to be pre-set and optimized through a large number of experimental tests and empirical analysis according to the characteristics of the vehicle model, the typical impact response time and the analysis requirements. For example, the pre-set time period before the impact event can be set to 0.5 seconds to capture the stable state or transition state before the impact; the pre-set time period after the impact event can be set to 1.5 to 2.5 seconds to ensure that the vibration caused by the impact and its subsequent decay process can be recorded completely.

[0043] First, a certain key time point of the identified impact event is taken as a reference, for example, the start time stamp of the event, T-start. Then, according to the pre-set time period before the impact (e.g. T-before = 0.5 seconds) and the pre-set time period after the impact (e.g. T-after = 2.0 seconds), the start time and the end time of the extracted data segment are calculated. The start time of the extraction T1-start = T-start minus T-before; the end time of the extraction T1-end = T-start + T-after. Finally, from the complete vehicle attitude data time series data stream, all vehicle vertical acceleration data, pitch angular velocity data and roll angular velocity data within the time range from T1-start to T1-end are extracted. For example, the start time stamp of an impact event T-start (the time corresponding to T4) is 10.10 seconds in the record. The pre-set time period before the impact is 0.5 seconds, and the pre-set time period after the impact is 2.0 seconds. Then, the start time of the extracted vehicle attitude data event segment will be 10.10 seconds - 0.5 seconds = 9.60 seconds, and the end time will be 10.10 seconds + 2.0 seconds = 12.10 seconds. The algorithm will extract all vertical acceleration, pitch angular velocity and roll angular velocity readings with time stamps between 9.60 seconds and 12.10 seconds from the total data stream to form this multi-component data segment, which is the vehicle attitude data event segment.

[0044] In step S4, the event vertical response feature, event pitch response feature and event roll response feature are extracted from the vehicle posture data event segment, and the event vertical response feature, the event pitch response feature and the event roll response feature are combined to obtain the event vehicle posture response feature coding vector. Accordingly, since the response of a single dimension is not enough to fully characterize the complex dynamic behavior of the shock absorber under a real impact. By extracting the key response features in the three directions of vertical, pitch and roll respectively, the multi-degree-of-freedom vibration characteristics of the vehicle in the impact event can be more completely captured. Combining these independently extracted features into a unified event vehicle posture response feature coding vector not only achieves data dimensionality reduction and structuring, but more importantly, forms a comprehensive fingerprint that can fully reflect the overall posture change of the vehicle under the impact event, providing standardized and information-rich input for the subsequent intelligent early warning of the shock absorber status based on pattern recognition.

[0045] Specifically, in a specific example of this application, Figure 4 FIG4 is a flow chart of step S4 in the intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to an embodiment of the present application. Figure 4 As shown, step S4, extracting event vertical response features, event pitch response features and event roll response features from the vehicle posture data event segment, and combining the event vertical response features, the event pitch response features and the event roll response features to obtain an event vehicle posture response feature encoding vector, including: S41, extracting a vehicle vertical acceleration local time series, a pitch angular velocity local time series and a roll angular velocity local time series from the vehicle posture data event segment; S42, extracting the event vertical response feature from the vehicle vertical acceleration local time series; S43, extracting the event pitch response feature from the pitch angular velocity local time series; S44, extracting the event roll response feature from the roll angular velocity local time series; S45, combining the event vertical response feature, the event pitch response feature and the event roll response feature to obtain the event vehicle posture response feature encoding vector. In particular, in a specific example of the present application, the event vertical response characteristics include peak-to-peak value, root mean square value, overshoot, number of oscillations, main vibration frequency and damping ratio estimation; the event pitch response characteristics include peak-to-peak value, root mean square value, decay time, number of overshoots and logarithmic decay rate; the event roll response characteristics include peak-to-peak value, root mean square value, decay time, number of overshoots and logarithmic decay rate.

[0046] In detail, step S4 can be implemented by first performing step S41, whose input is the vehicle posture data event segment intercepted in the previous step. This segment is a multi-dimensional time series containing time-aligned vehicle vertical acceleration, pitch rate and roll rate data. This multi-dimensional data segment is decomposed into three independent single-dimensional time series, i.e. the vehicle vertical acceleration local time series, the pitch rate local time series and the roll rate local time series are extracted from the vehicle posture data event segment as the output of this sub-step. For example, if the vehicle posture data event segment records the vehicle's posture data from 9.60s to 12.10s, then the vehicle vertical acceleration local time series is all the vertical acceleration sensor readings in time sequence within this 2.5s time window, and similarly, the pitch rate local time series and the roll rate local time series are obtained in the same way.

[0047] Next, step S42 is performed. The event vertical response features are extracted from the vehicle vertical acceleration local time series, and the specific calculation methods are as follows: 1. Peak-to-peak value: the absolute value of the difference between the maximum and minimum values in the sequence is calculated. For example, if the maximum acceleration in the sequence is 1.3 times the gravitational acceleration and the minimum is -0.9 times the gravitational acceleration, then the peak-to-peak value is equal to 1.3-(-0.9) = 2.2 times the gravitational acceleration. 2. Root mean square value: the values of each data point in the sequence are squared, the average of these squared values is calculated, and then the square root is taken to obtain. 3. Overshoot: refers to the maximum deviation of the impact response from its final steady-state value, or in this simplified form, the absolute amplitude of the first main peak after the impact. For example, the first peak value after the impact is 1.3 times the gravitational acceleration, then the overshoot is 1.3 times the gravitational acceleration. 4. Number of oscillations: the number of times the signal crosses a pre-set threshold (e.g. 0.05 times the gravitational acceleration, which is pre-set according to sensor noise and analysis requirements) around its mean line (or zero line for dynamic acceleration that has been compensated for gravity) during the decay process is calculated. 5. Main vibration frequency: the frequency component with the most concentrated energy is determined by performing a fast Fourier transform (FFT) analysis on the time series. 6. Damping ratio estimate: it can be estimated based on the amplitude ratio of two consecutive peaks in the same direction (e.g. A1 and A2) by methods such as the logarithmic decrement method, for example, the damping ratio is approximately equal to (the natural logarithm of A1 divided by A2) divided by (2 times pi). The six values obtained by these calculations form the output of the event vertical response features.

[0048] Then, step S43 is executed. The event pitch response feature is extracted from the local time series of the pitch angular velocity, including: 1. Peak-to-peak value: the calculation method is the same as the event vertical response feature. 2. Root mean square value: the calculation method is the same as the event vertical response feature. 3. Decay time: refers to the time it takes for the oscillation amplitude of the pitch angular velocity to decay from its first peak to a preset percentage of the peak (for example, 10%, this percentage needs to be pre-calibrated based on vehicle characteristics and experimental data). 4. Number of overshoots: calculate the number of times the amplitude of the pitch angular velocity exceeds a set threshold (for example, 20% of the initial peak) during the decay process. 5. Logarithmic decay rate: take the natural logarithm of the ratio of two consecutive unidirectional peak amplitudes (P1, P2) in the sequence, that is, the natural logarithm of P1 divided by P2. These five calculated values ​​constitute the output of the event pitch response feature.

[0049] Then, step S44 is executed. The event roll response feature is extracted from the local roll velocity time series. The feature types (peak-to-peak value, RMS value, decay time, overshoot count, and logarithmic decay rate) and calculation method are exactly the same as those described for pitch angular velocity in S43, but applied to the roll angular velocity data. These five calculated values ​​constitute the output of the event roll response feature.

[0050] Finally, step S45 is performed. The three sets of numerical features, namely the event vertical response feature, the event pitch response feature, and the event roll response feature, are arranged and connected in a predetermined fixed order to form a single, longer numerical list. For example, the six values ​​of the event vertical response feature can be placed first, followed by the five values ​​of the pitch response feature, and finally the five values ​​of the roll response feature. This final numerical list containing 6 plus 5 plus 5, which equals 16 elements, is the event vehicle posture response feature encoding vector.

[0051] In step S5, a corresponding healthy characteristic reference vector is queried from a database of similar event characteristics under healthy condition based on the average vehicle speed value of the impact event and the event type of the impact event. It should be appreciated that the vehicle's response to road impact in terms of vertical vibration, pitch, roll, etc. is not only dependent on the health of the shock absorber, but also significantly influenced by the vehicle's speed at the time of impact and the type of impact event itself, e.g. is it a circular speed bump, a long strip speed bump, a pothole, etc. If the characteristics of the current event are compared directly to a fixed ideal characteristic without considering these working condition factors, a misjudgment can occur due to the difference in working conditions. Therefore, by using the average vehicle speed value at the time of the current impact event and the event type of the impact as an index, a healthy characteristic reference vector under similar speed and impact type can be found by querying a pre-established database containing a large number of characteristics corresponding to different working conditions under healthy condition. This reference vector represents the response characteristics that the shock absorber should have under the current specific conditions when it is in an ideal healthy state, providing the most relevant comparison object for subsequent accurate assessment of the current shock absorber performance degradation level.

[0052] In detail, step S5 can be implemented by first obtaining the following inputs: the average vehicle speed value of the impact event, which is calculated from the vehicle posture data event segment (which covers a pre-defined time period before and after the impact, e.g. from 0.5 seconds before the impact start point T-start to 2.0 seconds after T-after) clipped in the previous step S3. Specifically, the algorithm extracts all vehicle speed readings recorded within the 2.5 seconds time segment and calculates the arithmetic mean. For example, if the instantaneous speed sequence of the vehicle within the 2.5 seconds event segment is [40, 41, 42, 43, 42, 41, 40] km / h, the calculated average vehicle speed value can be 41.2 km / h. The event type of the impact event, which is used to differentiate road impacts of different nature. Its determination can rely on pre-defined rules, e.g. by analyzing the characteristics of the impact event vertical acceleration waveform such as pulse width, peak symmetry, etc. to automatically classify it into Type A: standard speed bump, Type B: single pothole, Type C: continuous bumping, etc. The rules and categories of this classification need to be pre-set. For example, the current impact event is identified as Type A: standard speed bump. The database of similar event features under healthy condition, which is a core pre-set resource. The construction of this database needs to be completed when the vehicle shock absorber is confirmed to be in brand new or good health condition. By driving a healthy vehicle to perform impact tests on various pre-set event types of road surface, e.g. various specifications of speed bumps, potholes, and in various pre-set vehicle speed intervals, e.g. 20-35 km / h, 35-50 km / h, 50-65 km / h. For each impact under healthy condition, clip the event segment according to step S3, and then extract its event vehicle posture response feature encoding vector (containing 16 feature values) according to step S4. Then, store these feature vectors under healthy condition, together with their corresponding [vehicle speed interval, event type] as index, into the database.

[0053] The query process is as follows: the average vehicle speed value obtained by the current impact event, for example, 41.2 kilometers per hour, and the determined event type, for example, type A: standard deceleration strip, are taken as query inputs. In the first step, the average vehicle speed value is classified into a preset vehicle speed interval according to the average vehicle speed value. For example, if the preset vehicle speed interval includes 35-50 kilometers per hour, 41.2 kilometers per hour falls into this interval. In the second step, a combined key, that is, [vehicle speed interval: 35-50 kilometers per hour, event type: type A: standard deceleration strip], is used to perform an accurate or most similar matching query in the health state of the same event feature database. If there is a record in the database that completely matches the combined key, for example: {vehicle speed interval: 35-50 kilometers per hour, event type: type A: standard deceleration strip, health feature reference vector: Vh=[h1, h2,..., h16]}, then the vector Vh is the output result obtained by the query in this step, that is, the corresponding health feature reference encoding vector.

[0054] In step S6, a deviation value between the event vehicle posture response feature encoding vector and the health feature reference encoding vector is calculated. Accordingly, the feature encoding vector representing the current impact response and the feature reference encoding vector representing the health shock absorber under similar working conditions have been obtained respectively before this step. In order to quantitatively compare the actual performance of the vehicle shock absorber under the current impact event with the expected performance under the ideal health state, so as to evaluate the degree of performance degradation of the shock absorber, the present application can obtain one or a group of specific numerical values, the deviation value, by calculating the difference between the two vectors. The deviation value directly reflects the degree to which the current shock absorber response deviates from the health standard. A smaller deviation value means that the shock absorber state is close to health, while a significant deviation value indicates potential performance decline or failure, which provides a key numerical basis for subsequent shock absorber state diagnosis and early warning decision-making.

[0055] Specifically, in one specific example of the present application, calculating the deviation value between the event vehicle posture response feature encoding vector and the health feature reference encoding vector comprises: calculating the Mahalanobis distance between the event vehicle posture response feature encoding vector and the health feature reference encoding vector as the deviation value.

[0056] In detail, step S6 can be implemented by the following formula:

[0057]

[0058] wherein V is the event vehicle posture response feature encoding vector, V1 is the health feature reference encoding vector, T is the vector transposition operation, S -1 is the inverse covariance matrix between V and V1, and k is the deviation value. That is, the formula is the Mahalanobis distance formula for calculating the deviation value k. Here, S-1 is key, which represents the variance and correlation between the multi-dimensional characteristics of health data. By introducing S -1 , the Mahalanobis distance can eliminate the influence of the dimension and correlation of each characteristic quantity, so as to more accurately measure the statistical distance of V deviating from the center of the health data group represented by V1. Therefore, using the Mahalanobis distance to calculate the deviation value can more accurately and robustly evaluate the real difference between the current shock absorber response and the health state response, avoid misjudgment caused by different scales of features or inherent correlation between features, and improve the accuracy and reliability of the warning.

[0059] In particular, for the response characteristics in the impact event time period, due to the dynamic threshold difference between the high positive peak value and the high negative peak value of the impact event itself, and due to the anisotropic distribution of the response characteristics in this time period, such as the peak value, the root mean square value as a digital statistical characteristic, the number of times, the frequency value as a number of times statistical characteristic, etc., it will cause anisotropic distortion of the feature distribution, thereby causing the overall distribution of the event vehicle posture response feature encoding vector to deviate, resulting in nonlinear disturbance of its covariance structure, which affects the calculation accuracy of the Mahalanobis distance as the covariance correlation distribution distance measure.

[0060] Based on this, in another specific example of the present application, the deviation value between the event vehicle posture response feature encoding vector and the health feature reference encoding vector is calculated, comprising:

[0061] The mean value of each feature value in the event vehicle posture response feature encoding vector is normalized to obtain a normalized event vehicle posture response feature encoding vector, that is:

[0062] v′ i =v i / μ

[0063] Where v i is each feature value in the event vehicle posture response feature encoding vector, μ is the mean value of all feature values v i , v′ i is each feature value in the normalized event vehicle posture response feature encoding vector, that is, for the event vehicle posture response feature encoding vector V composed of the event vertical response feature, the event pitch response feature and the event roll response feature, and each feature value is denoted as v i , first, the anisotropic distribution feature is normalized, for example, considering the covariance structure, mean value normalization is adopted;

[0064] The distribution density function of each feature value in the normalized event vehicle posture response feature encoding vector under the overall distribution is measured to obtain a distribution density event vehicle posture response feature encoding vector, that is:

[0065]

[0066] wherein, Φ i is each feature value in the distribution density event vehicle posture response feature encoding vector, that is, considering the mean normalization, μ is also a vector value v i , so by performing the distribution density function calculation, it is essentially to perform the cumulative calculation of the distribution density of the impact event in its predetermined time period, thereby playing the role of converting the deviation caused by the distribution distortion into the displacement in the density integral space;

[0067] performing structure-preserving embedding of each feature value in the distribution density event vehicle posture response feature encoding vector from the density integral space to the covariance structure to obtain an event vehicle posture response feature encoding correction vector, that is:

[0068]

[0069] wherein, exp is the exponential function value with the natural constant e as the base, v″ i is each feature value in the event vehicle posture response feature encoding correction vector, so as to calculate the curvature representation of the density integral in the covariance structure, so as to improve the dynamic covariance fluctuation stability in the predetermined time period in the structured distribution transfer process, thereby reducing the nonlinear disturbance of the event vehicle posture response feature encoding vector in the covariance distribution structure, and improving the calculation accuracy of the Mahalanobis distance;

[0070] calculate the Mahalanobis distance between the event vehicle posture response feature encoding correction vector and the health feature reference encoding vector as the deviation value, that is:

[0071]

[0072] wherein, V′ is the event vehicle posture response feature encoding correction vector, V1 is the health feature reference encoding vector, T is the vector transposition operation, S -1 is the inverse covariance matrix between V′ and V1, and k is the deviation value.

[0073] In step S7, it is determined whether to generate a new energy automobile shock absorber state abnormality early warning prompt based on the deviation value. It can be understood that the deviation value directly reflects the difference between the current shock absorber performance and the health benchmark. When this difference accumulates or instantaneously reaches a preset dangerous threshold, it indicates that the shock absorber may have a fault or significant performance degradation, and timely intervention is needed. Therefore, in order to convert the shock absorber state from continuous numerical evaluation to discrete early warning decision based on the quantitative deviation calculated in the foregoing, the present application determines whether to generate a new energy automobile shock absorber state abnormality early warning prompt based on the deviation value. That is, the early warning is intended to inform the user or maintenance personnel to pay attention to this abnormality in order to take preventive maintenance or repair measures, thereby ensuring vehicle driving safety, ride comfort, and preventing fault expansion. This decision is a key link to realize intelligent monitoring and active early warning of the shock absorber state, ensuring that the monitoring result can be converted into effective action guidance.

[0074] Specifically, in one specific example of the present application, based on the deviation value, it is determined whether to generate a new energy automobile shock absorber state abnormality early warning prompt, comprising: in response to the deviation value being greater than or equal to a preset threshold, determining to generate a new energy automobile shock absorber state abnormality early warning prompt.

[0075] In detail, step S7 can be implemented in the following way: the input of this step is the deviation value calculated in the previous step. At the same time, a key preset parameter is the preset threshold. The preset threshold is a key parameter, and its setting directly affects the sensitivity and accuracy of the early warning. This threshold is based on statistical analysis of the deviation value distribution under a large number of shock absorber health and different degrees of fault state, for example, by receiver operating characteristic (ROC) curve analysis, a critical point that can better balance the false negative rate and false positive rate is selected; or combined with vehicle dynamics professional knowledge, maintenance experience and expected early warning advance to set. For example, by analyzing a large amount of data, it is found that when the deviation value calculated by Mahalanobis distance is generally less than 2.5, it represents health, and when it continuously exceeds 3.0, it means that the performance of the shock absorber has appeared negligible degradation. Therefore, the preset threshold can be set to 3.0, of course, this is only an example, which can be limited according to the actual situation, and this embodiment does not constitute a specific limitation.

[0076] First, the deviation value calculated in the previous step is obtained. Then, the deviation value is compared with a preset threshold value preset and stored in the program. If the deviation value is greater than or equal to the preset threshold value, the condition is determined to be met, and it is determined that a new energy vehicle shock absorber state abnormality early warning prompt is generated. This early warning prompt can be sending a specific warning signal or text information to the vehicle human-computer interaction interface, such as the instrument panel or the central display screen, or reporting the early warning information to the remote monitoring platform through the vehicle-mounted communication module. For example, if the deviation value calculated in the previous step is 3.8, and the preset threshold value for triggering the early warning is 3.0. Because 3.8 is greater than or equal to 3.0, the condition is met, and an action is output, that is, it is determined that a new energy vehicle shock absorber state abnormality early warning prompt is generated. Conversely, if the deviation value is 2.1, which is less than 3.0, no early warning prompt is generated, and the process ends or waits for the analysis of the next impact event.

[0077] In summary, the new energy vehicle shock absorber state intelligent early warning method based on attitude monitoring based on the embodiments of the present application is illustrated, which can accurately distinguish normal driving vibration from abnormal impact working conditions by real-time collection of multi-dimensional attitude data such as vehicle vertical acceleration, pitch angle velocity and roll angle velocity, and adaptive detection of impact events by combining dynamic threshold algorithm. In the feature extraction stage, the peak-to-peak value, root mean square value, damping ratio and other multi-dimensional physical characteristics are combined to code, which can fully represent the dynamic attenuation characteristics of the shock absorber in the impact event, overcoming the defect of insufficient sensitivity of single feature. By establishing a feature database of the same type of event under healthy state, and using Mahalanobis distance to calculate the deviation of multi-feature space, not only the statistical correlation between features is considered, but also the adaptive matching of reference response mode under different vehicle speeds and impact types is considered, which significantly improves the robustness of abnormal detection. Especially through the optimization of the quadratic feature embedded by the distribution density function and the covariance structure, the influence of noise interference on the feature vector is effectively eliminated, so that the system can still accurately capture the performance degradation trend of the shock absorber under complex working conditions, and realize early fault warning. The overall scheme forms a closed-loop diagnosis logic through multi-dimensional data fusion, dynamic event detection and intelligent feature comparison, which can effectively solve the technical problems of poor environmental adaptability and low diagnosis accuracy of traditional methods.

[0078] The above has described various embodiments of the present disclosure, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes are obvious to those skilled in the art without departing from the scope and spirit of the described embodiments. The selection of the terms used herein is intended to best explain the principles, practical applications, or improvements to the technology in the market of the embodiments, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring, characterized in that: include: Obtaining a time series data stream of vehicle attitude data, where the vehicle attitude data includes the vehicle's vertical acceleration, pitch angular velocity, and roll angular velocity; Extracting a vehicle vertical acceleration time series data stream from the vehicle posture data time series data stream, and performing impact event detection on the vehicle vertical acceleration time series data stream based on a preset rule; In response to detecting an impact event, intercepting a time series data stream of vehicle posture data for a preset time period before and after the impact event to obtain a vehicle posture data event segment; Extracting an event vertical response feature, an event pitch response feature, and an event roll response feature from the vehicle posture data event segment, and combining the event vertical response feature, the event pitch response feature, and the event roll response feature to obtain an event vehicle posture response feature encoding vector; Based on the average vehicle speed value of the impact event and the event type of the impact event, querying the corresponding health feature benchmark coding vector from a similar event feature database in a healthy state; Calculating a deviation value between the event vehicle posture response feature coding vector and the health feature reference coding vector; Based on the deviation value, it is determined whether to generate an abnormal state warning prompt of the shock absorber of the new energy vehicle.

2. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: The vehicle posture data time series data stream is collected by an IMU sensor.

3. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: Performing impact event detection on the vehicle vertical acceleration time series data stream based on preset rules includes: Input the current vehicle speed value into the dynamic threshold function to obtain the current dynamic threshold; determining a data point in the vehicle vertical acceleration time series data stream at which the data point exceeds the current dynamic threshold for the first time, and marking the data point as a starting point of a potential event; Traversing the vehicle vertical acceleration time series data stream backward along the starting point of the potential event and determining whether the absolute value of the positive peak value and the absolute value of the negative peak value are both greater than the current dynamic threshold; In response to the absolute value of the positive peak value and the absolute value of the negative peak value being simultaneously greater than the current dynamic threshold, calculating the pulse width based on the timestamp of the positive peak value and the timestamp of the negative peak value; When the pulse width meets the pulse width condition, the potential event is marked as a shock event.

4. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: Extracting an event vertical response feature, an event pitch response feature, and an event roll response feature from the vehicle posture data event segment, and combining the event vertical response feature, the event pitch response feature, and the event roll response feature to obtain an event vehicle posture response feature encoding vector, including: Extracting a local time series of vehicle vertical acceleration, a local time series of pitch angular velocity, and a local time series of roll angular velocity from the vehicle posture data event segment; Extracting the event vertical response feature from the vehicle vertical acceleration local time series; extracting the event pitch response feature from the pitch angular velocity local time series; extracting the event roll response feature from the roll angular velocity local time series; The event vertical response feature, the event pitch response feature, and the event roll response feature are combined to obtain the event vehicle posture response feature encoding vector.

5. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 4 is characterized in that: The vertical response characteristics of the event include peak-to-peak value, root mean square value, overshoot, number of oscillations, main vibration frequency and damping ratio estimation; the pitch response characteristics of the event include peak-to-peak value, root mean square value, decay time, number of overshoots and logarithmic decay rate; the roll response characteristics of the event include peak-to-peak value, root mean square value, decay time, number of overshoots and logarithmic decay rate.

6. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: Calculating the deviation value between the event vehicle posture response feature coding vector and the health feature benchmark coding vector includes: calculating the Mahalanobis distance between the event vehicle posture response feature coding vector and the health feature benchmark coding vector as the deviation value.

7. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: Calculating a deviation value between the event vehicle posture response feature coding vector and the health feature reference coding vector includes: performing mean normalization processing on each eigenvalue in the event vehicle posture response feature coding vector to obtain a normalized event vehicle posture response feature coding vector; Measuring the distribution density function of each eigenvalue in the normalized event vehicle posture response feature coding vector under the overall distribution to obtain a distribution density event vehicle posture response feature coding vector; Performing structure-preserving embedding from density integral space to covariance structure on each eigenvalue in the distribution density event vehicle posture response feature coding vector to obtain an event vehicle posture response feature coding correction vector; The Mahalanobis distance between the event vehicle posture response feature coding correction vector and the health feature reference coding vector is calculated as the deviation value.

8. The intelligent early warning method for shock absorber status of new energy vehicles based on posture monitoring according to claim 1 is characterized in that: Based on the deviation value, determining whether to generate an abnormal state warning prompt for the shock absorber of the new energy vehicle includes: in response to the deviation value being greater than or equal to a preset threshold, determining to generate an abnormal state warning prompt for the shock absorber of the new energy vehicle.

Images