Vehicle load estimation method, device, apparatus and storage medium

Abstract

The application discloses an estimation method, device and equipment of vehicle load and a storage medium, relates to the technical field of load measurement and deep learning, and comprises the following steps: acquiring strain data and vehicle state data of a vehicle leaf spring; performing data processing on the strain data and the vehicle state data to obtain target strain data, target vehicle state data and a steady state flag, wherein the data processing comprises alignment and resampling, filtering and temperature correction; constructing time sequence features and scalar features for load estimation according to the target strain data, the target vehicle state data and the steady state flag; and inputting the time sequence features and the scalar features into a load prediction model to obtain a vehicle load estimation value. Through the deep learning model fusing physical priori and data driving, high-precision estimation of the vehicle load under different working conditions is realized, and the accuracy, stability and cross-model generalization capability of the load estimation are improved.

Description

Technical Field

[0001] This application relates to the technical field of load measurement and deep learning, and in particular to a method, apparatus, device and storage medium for estimating vehicle load. Background Technology

[0002] Accurate and real-time estimation of the load (or axle load / total weight) of trucks, tractors, and trailers with leaf spring suspensions is of great significance in logistics management, safety monitoring, and road and bridge weighing. Currently, commonly used estimation methods mainly include the following categories: Estimation based on vehicle dynamics models: This method uses the vehicle's motion equations to calculate the load through information such as acceleration and gradient. This method is sensitive to road surface excitation, driving operations, and sensor noise, and the estimation results under static and dynamic conditions are inconsistent, making it difficult to meet the accuracy requirements of measurement. Direct force measurement methods: Such as installing force sensors or weighing bridges on the suspension, saddle, or axle ends. Although this method can directly measure the force, it usually requires structural modifications to the chassis, resulting in high costs and complex installation, making it difficult to promote on a large scale in existing vehicles. Weighbridge weighing: As a static weighing standard, it has high accuracy, but it cannot achieve real-time online monitoring during vehicle operation, limiting its application in dynamic load management.

[0003] Therefore, improving the accuracy, stability, and cross-model generalization ability of load estimation is an urgent problem to be solved. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, device, and storage medium for estimating vehicle load, aiming to solve the technical problem of how to improve the accuracy, stability, and cross-vehicle generalization capability of load estimation.

[0005] To achieve the above objectives, this application proposes a method for estimating vehicle load capacity, the method comprising:

[0006] Acquire strain data of the vehicle leaf springs and vehicle status data; The strain data and vehicle state data are processed to obtain target strain data, target vehicle state data and steady-state flags. The data processing includes alignment and resampling, filtering and temperature correction. Based on the target strain data, the target vehicle state data, and the steady-state indicator, a time-series feature and a scalar feature for load estimation are constructed. The time-series features and the scalar features are input into the load prediction model to obtain the estimated vehicle load.

[0007] In one embodiment, the step of processing the strain data and vehicle state data to obtain target strain data, target vehicle state data, and steady-state indicators includes: The strain data and the vehicle state data are time-aligned and resampled to obtain a signal time series with a uniform sampling rate. The strain data and vehicle state data in the signal time series are subjected to multi-level filtering to obtain filtered strain data and target vehicle state data. The filtered strain data is temperature-compensated according to a preset temperature compensation model to obtain the temperature-compensated target strain data. Based on the target vehicle state data and the target strain data, a steady-state window determination is performed on the current data acquisition time window to obtain a steady-state flag.

[0008] In one embodiment, the step of time-aligning and resampling the strain data and the vehicle state data to obtain a signal time series with a uniform sampling rate includes: The strain data and the vehicle status data are aligned according to a preset target sampling frequency, and the jitter time of the aligned strain data and vehicle status data is obtained. The strain data and the vehicle state data are time-aligned according to the jitter time using a preset strategy to obtain a signal time series with a uniform sampling rate. The preset strategy includes any one of the following: a hold strategy, a linear interpolation strategy, and a missing label strategy.

[0009] In one embodiment, the step of performing temperature compensation on the filtered strain data according to a preset temperature compensation model to obtain temperature-compensated target strain data includes: Obtain the strain gauge temperature of the vehicle leaf spring; The temperature drift is calculated based on the preset temperature compensation model, the preset reference temperature, and the strain gauge temperature. The filtered strain data is corrected based on the temperature drift to obtain temperature-compensated target strain data, which includes left-side temperature-compensated strain and right-side temperature-compensated strain.

[0010] In one embodiment, the step of determining a steady-state flag by analyzing the current data acquisition time window based on the target vehicle state data and the target strain data includes: Get the duration of the data in the current data acquisition time window; When the duration is greater than or equal to a preset stabilization time and the vehicle speed, longitudinal acceleration, lateral acceleration, and steering angular velocity in the target vehicle state data all meet the corresponding preset steady-state conditions, the time window is determined to be a steady-state window, and a steady-state flag is added to the time window.

[0011] In one embodiment, the step of constructing time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state indicator includes: Acquire temperature data and vehicle configuration information; Based on the left-side temperature-compensated strain and right-side temperature-compensated strain in the target strain data, calculate the strain sum, strain difference, and left-right strain ratio. The time-series features are constructed based on the target vehicle state data, the left-side temperature-compensated strain, the right-side temperature-compensated strain, the sum of strains, the strain difference, and the left-right strain ratio. The target vehicle state data includes vehicle speed, longitudinal acceleration, lateral acceleration, pitch angle, and roll angle. Within the time window indicated by the steady-state indicator, window statistical features and a steady-state score are calculated based on the statistical features of one or more data in the time-series features. Based on the window statistical features, the steady-state score, the temperature data, and the vehicle configuration information, scalar features are constructed.

[0012] In one embodiment, after the step of inputting the time-series features and the scalar features into the load prediction model to obtain the vehicle load estimate, the method further includes: When the steady-state flag indicates that the current state is a steady-state window, the vehicle load estimate output by the load prediction model is exponentially smoothed according to a preset smoothing coefficient to obtain a smoothed vehicle load estimate. The vehicle load estimate is updated based on the smoothed vehicle load estimate.

[0013] Furthermore, to achieve the above objectives, this application also proposes a vehicle load estimation device, the device comprising: The data acquisition module is used to acquire strain data of the vehicle leaf springs and vehicle status data; The data processing module is used to process the strain data and vehicle state data to obtain target strain data, target vehicle state data and steady-state flags. The data processing includes alignment and resampling, filtering and temperature correction. The feature construction module is used to construct time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag. The load prediction module is used to input the time-series features and the scalar features into the load prediction model to obtain the estimated vehicle load.

[0014] In addition, to achieve the above objectives, this application also proposes a vehicle load estimation device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the vehicle load estimation method as described above.

[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements the steps of the vehicle load estimation method described above.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the vehicle load estimation method described above.

[0017] This application provides a method for estimating vehicle load. The method includes: acquiring strain data of a vehicle leaf spring and vehicle state data; processing the strain data and vehicle state data to obtain target strain data, target vehicle state data, and a steady-state indicator, wherein the data processing includes alignment and resampling, filtering, and temperature correction; constructing temporal features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state indicator; and inputting the temporal features and the scalar features into a load prediction model to obtain an estimated vehicle load value. In summary, this application, by employing a deep learning model that integrates physical priors and data-driven approaches, achieves high-precision estimation of vehicle load under different operating conditions, improving the accuracy, stability, and cross-vehicle generalization ability of load estimation. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart illustrating the first embodiment of the vehicle load estimation method of this application; Figure 2 This is a diagram illustrating the data acquisition architecture in one embodiment of the vehicle load estimation method of this application. Figure 3This is a schematic diagram of the load prediction model in one embodiment of the vehicle load estimation method of this application; Figure 4 A flowchart illustrating the second embodiment of the vehicle load estimation method of this application; Figure 5 A flowchart illustrating the third embodiment of the vehicle load estimation method of this application; Figure 6 This is a schematic diagram of the module structure of the vehicle load estimation device according to an embodiment of this application; Figure 7 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the vehicle load estimation method in the embodiments of this application.

[0021] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0023] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0024] The main solution of this application embodiment is as follows: acquiring strain data of the vehicle leaf spring and vehicle state data; performing data processing on the strain data and vehicle state data to obtain target strain data, target vehicle state data, and steady-state flag, wherein the data processing includes alignment and resampling, filtering, and temperature correction; constructing time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag; and inputting the time-series features and the scalar features into the load prediction model to obtain the vehicle load estimate.

[0025] Accurate and real-time estimation of the load (or axle load / total weight) of trucks, tractors, and trailers with leaf spring suspensions is of great significance in logistics management, safety monitoring, and road and bridge weighing. Currently, commonly used estimation methods mainly include the following categories: Estimation based on vehicle dynamics models: This method uses the vehicle's motion equations to calculate the load through information such as acceleration and gradient. This method is sensitive to road surface excitation, driving operations, and sensor noise, and the estimation results under static and dynamic conditions are inconsistent, making it difficult to meet the accuracy requirements of measurement. Direct force measurement methods: Such as installing force sensors or weighing bridges on the suspension, saddle, or axle ends. Although this method can directly measure the force, it usually requires structural modifications to the chassis, resulting in high costs and complex installation, making it difficult to promote on a large scale in existing vehicles. Weighbridge weighing: As a static weighing standard, it has high accuracy, but it cannot achieve real-time online monitoring during vehicle operation, limiting its application in dynamic load management. Therefore, how to improve the accuracy, stability, and cross-vehicle generalization ability of load estimation is a problem that urgently needs to be solved.

[0026] It should be noted that the executing entity in this embodiment can be a vehicle load estimation system, a computing service device with data processing, network communication, and program execution functions, or an electronic device capable of realizing the aforementioned vehicle load estimation function, etc. This embodiment does not specifically limit it in this way. The following uses a vehicle load estimation system as an example to describe this embodiment and the following embodiments.

[0027] Based on this, embodiments of this application provide a method for estimating vehicle load capacity, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the vehicle load estimation method of this application.

[0028] In this embodiment, the method for estimating the vehicle load includes steps S10 to S40: Step S10: Obtain strain data of the vehicle leaf spring and vehicle status data.

[0029] It should be noted that, as Figure 2 As shown, in this step, raw data is acquired in real time through a sensing and acquisition system deployed on the vehicle. Specifically, strain gauges are symmetrically mounted on the leaf springs on both sides of each axle to be monitored (e.g., front axle, middle axle, rear axle) to form a Wheatstone bridge (full bridge), and temperature sensors are installed nearby to measure the temperature near the strain gauges. Simultaneously, the vehicle's motion state is acquired through an onboard inertial measurement unit (IMU). A dedicated electronic control unit (ECU) is responsible for providing a stable excitation voltage or current to the strain bridge, amplifying the differential millivolt-level signal output by the bridge, and transmitting it through a high-resolution analog-to-digital converter (ADC, such as a 24-bit ADC). The ECU performs sampling and digitization on the original strain signal using an ADC (Digital Converter for Detection and Digitization). At the same time, the ECU performs preliminary digital filtering (such as amplitude limiting, Hampel filtering, and low-pass filtering) and downsampling on the original strain signal, and makes a preliminary steady-state probability judgment based on signals such as vehicle speed and acceleration, and marks it as a steady-state signal.

[0030] Additionally, it should be noted that the strain data mainly refers to the digital sequence of the original voltage signal obtained by measuring and converting it through a strain gauge bridge, caused by the deformation of the leaf springs, and then sampled by an ADC. It typically includes left-side and right-side strain, corresponding to the real-time deformation of the left and right leaf springs of the same axle, respectively. The vehicle status data includes at least the vehicle speed (v), longitudinal acceleration (ax), lateral acceleration (ay), yaw rate, pitch angle, and roll angle provided by the IMU.

[0031] Step S20: Perform data processing on the strain data and vehicle state data to obtain target strain data, target vehicle state data and steady-state flag. The data processing includes alignment and resampling, filtering and temperature correction.

[0032] It should be noted that in this step, after receiving asynchronous data packets from different ECU channels, the platform layer of the system performs data alignment and resampling to unify multiple signals to the same time baseline and sampling rate, thereby eliminating the impact of jitter and packet loss in network transmission. Next, the aligned strain data undergoes multi-stage filtering to suppress pulse interference, power frequency noise, and high-frequency vibrations, retaining the low-frequency useful signals reflecting load changes. Then, using near-field temperature data acquired from sensors, temperature compensation is applied to the filtered strain data to eliminate the drift of strain gauge and adhesive layer characteristics caused by changes in ambient temperature, obtaining stable and reliable target strain data (i.e., the temperature-compensated strain εL_T on the left and the temperature-compensated strain εR_T on the right). Similarly, the vehicle state data also undergoes corresponding low-pass filtering to obtain the target vehicle state data. Furthermore, based on the processed vehicle speed, acceleration, angular velocity, and the inherent volatility of the strain signal, the system performs precise steady-state window determination and generates a steady-state flag. This flag indicates whether the current moment is in a quasi-static condition where the vehicle is running smoothly and suitable for high-precision measurement.

[0033] Additionally, it should be noted that the target strain data (εL_T, εR_T) are the final results of the original strain data after alignment, resampling, multi-level filtering, zero-point correction and temperature compensation, and their unit is usually microstrain (με).

[0034] Step S30: Construct time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag.

[0035] It should be noted that, in order to fully utilize the spatiotemporal information and statistical characteristics of the data, two types of features are constructed from the processed data and input into the prediction model. The system calculates derived features with clear physical meaning based on the target strain data, such as the sum of strains, strain difference, and left-right strain ratio. These features are then combined with the target vehicle state data (i.e., vehicle speed (v), longitudinal acceleration (ax), lateral acceleration (ay), yaw rate (yaw_rate), pitch angle (pitch), and roll angle (roll)) at time steps to form a multidimensional temporal feature sequence. Secondly, within a set time window (e.g., 2-5 seconds), statistical analysis is performed on the temporal features (especially strain-related features), calculating their mean, variance, peak-to-trough values, root mean square, trend slope, etc., to obtain window statistical features. Simultaneously, temperature data (such as strain gauge temperature, ambient temperature, ECU temperature), a steady-state score calculated based on the target vehicle state data and strain fluctuations, and a vehicle configuration embedding vector representing the vehicle's inherent attributes (mapped from vehicle model, axle type, leaf spring model, etc.) are combined to form scalar features.

[0036] Step S40: Input the time series features and the scalar features into the load prediction model to obtain the vehicle load estimate.

[0037] It should be noted that in this step, the system inputs the constructed temporal features and scalar features into a pre-trained deep learning model—the load prediction model. For example... Figure 3 As shown, the model employs a dual-branch structure: one branch (such as 1D-CNN or TCN, including 3–4 dilated convolutional residual blocks (kernel=3, dilation=1 / 2 / 4( / 8), channel width 32–64, output h_seq)) is used to extract deep patterns of temporal features; the other branch (such as a multilayer perceptron MLP, output h_scalar) is used to process scalar features. The outputs of the two branches are fused and then used by a regression head network to estimate the vehicle load (including axle load estimates and total vehicle weight estimates), ultimately outputting both the axle load estimate (y_axle_i) for each axle and the total vehicle weight estimate (y_total). The total vehicle weight estimate is equal to the sum of the axle loads of all axles.

[0038] Understandably, the model learned a high-precision nonlinear mapping relationship between complex features and actual loads during training, and learned from different vehicle configurations, thus possessing good cross-vehicle generalization ability.

[0039] In one feasible implementation, after step S40, the method further includes: Step S50: When the steady-state flag indicates that the current state is a steady-state window, the vehicle load estimate output by the load prediction model is exponentially smoothed according to a preset smoothing coefficient to obtain a smoothed vehicle load estimate.

[0040] It should be noted that, in order to obtain a more stable and less volatile metrological output under steady-state conditions, when the steady-state indicator confirms that the current state is within the steady-state window (e.g., the steady-state score exceeds the preset steady-state score threshold and lasts for ≥2 seconds), the real-time output vehicle load estimate (axle load or total vehicle weight estimate) is subjected to exponential smoothing filtering. The smoothing formula is shown in Equation 1: (Formula 1) in, This is the estimated vehicle load value output by the current model. This is the smoothed vehicle load estimate. This is the smoothed vehicle load estimate from the previous time step. This is a preset smoothing coefficient (e.g., a value between 0.2 and 0.4). Upon first reaching steady state, It can be initialized to the average value of the model output within the steady-state window. This step effectively suppresses small fluctuations caused by random noise and improves the reliability of steady-state readings.

[0041] Additionally, it should be noted that exponential smoothing is a time series filtering method, and its smoothing effect is determined by the smoothing coefficient. control, The larger the value, the faster the response to the current value but the worse the stability; conversely, the smaller the value, the stronger the smoothing effect but the greater the response delay. This step is activated only when the system is in a steady-state window and is designed to output stable values ​​for measurement purposes.

[0042] Step S60: Update the vehicle load estimate based on the smoothed vehicle load estimate.

[0043] It should be noted that in this step, the system dynamically selects the output value based on the operating conditions. Within the steady-state window, the system outputs a smoothed vehicle load estimate as the final effective load estimate, marked as high confidence. In non-steady-state windows (such as during driving), the system directly outputs the original estimate from the model (or performs only very slight smoothing), marked as low confidence, clearly indicating its purpose as a trend reference rather than precise measurement. In this way, the system achieves a comprehensive consideration of steady-state high-precision measurement and dynamic trend estimation. This step essentially completes the final release of the estimation results and confidence management. The update mechanism ensures that the output results match the actual operating state of the vehicle, satisfying the stringent accuracy requirements when stationary / low-speed, while also guaranteeing continuous monitoring of load change trends during driving.

[0044] Additionally, it should be noted that after the load prediction model outputs the estimated axle load (y_axle_i) and the estimated total vehicle weight (y_total) for each axle, the system calculates an index called the consistency residual, r_cons. The specific calculation method is shown in Formula 2: (Formula 2) This involves subtracting the sum of all axle load estimates from the total vehicle weight estimate directly output by the model. This step is performed automatically after each model inference, and the result serves as an internal health monitoring indicator. It is understood that the consistency residual (r_cons) should ideally be close to zero. Its absolute value reflects the degree of internal consistency of the model's predictions. A significantly increased |r_cons| may indicate abnormal data input for a specific axle (such as sensor failure, strong signal interference), prediction bias in that operating condition, or misalignment of feature data from different axles. Therefore, this residual is used as an effective auxiliary indicator for model self-checking and fault diagnosis.

[0045] Additionally, it should be noted that to ensure long-term measurement accuracy, the system supports and executes periodic calibration procedures. Calibration consists of three stages: Initial Installation Calibration: After the system is first installed on the vehicle, a zeroing operation is performed under no-load conditions. Subsequently, the vehicle is driven onto the weighbridge or a mobile axle load scale is used to weigh the vehicle under various load conditions, including no-load (0%), partial load (e.g., 30%, 60%), and full load (100%), while simultaneously recording data such as strain output by the system. Through multi-point data regression, the zero-point offset, load-strain proportionality coefficient, and coefficients (k1, k2) from the aforementioned temperature compensation model for each channel are calculated and stored. This process may also include left and right off-center load condition tests to verify the system's off-center load identification capability. Periodic Recalibration: During vehicle use (e.g., monthly or quarterly), when the vehicle passes the weighbridge, the system can automatically or manually trigger a recalibration process. The system uses the weighbridge weighing result as the true value and compares it with its current estimated value. If the deviation exceeds a preset threshold, a calibration algorithm is activated to adjust relevant parameters (mainly zero point and proportional coefficient) in small steps with limited amplitude to avoid parameter abrupt changes due to a single abnormal data point. Online fine-tuning: During daily vehicle operation, the system continuously monitors operating conditions. When it detects that the vehicle is in a long-term idling and steady-state condition (such as parking at night), it can automatically perform small-amplitude zero-point drift compensation to cope with the slow aging of sensors and circuits.

[0046] It is understood that calibration is the process of establishing an accurate mapping relationship between the output value of the measurement system and the physical true value. Recalibration is a parameter correction process to maintain this accurate relationship and offset the drift of the system over time. This method, through a hierarchical calibration strategy (initial multi-point calibration, periodic triggered recalibration, and online no-load fine-tuning), greatly reduces the maintenance cost of frequent manual calibration while ensuring high accuracy, thus achieving a balance between accuracy and availability.

[0047] This embodiment provides a method for estimating vehicle load. The method includes: acquiring strain data of the vehicle leaf spring and vehicle state data; processing the strain data and vehicle state data to obtain target strain data, target vehicle state data, and a steady-state flag; the data processing includes alignment and resampling, filtering, and temperature correction; constructing temporal and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag; and inputting the temporal and scalar features into a load prediction model to obtain an estimated vehicle load value. In summary, this embodiment, by employing a deep learning model that integrates physical priors and data-driven approaches, achieves high-precision estimation of vehicle load under different operating conditions, improving the accuracy, stability, and cross-vehicle generalization ability of load estimation.

[0048] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 , Figure 4 This is a flowchart illustrating the second embodiment of the vehicle load estimation method of this application. Step S20 specifically includes: Step S201: Time-align and resample the strain data and the vehicle state data to obtain a signal time series with a uniform sampling rate.

[0049] It should be noted that the uniform sampling rate signal time series refers to a data matrix formed by precisely aligning all sensor signals in the time dimension, arranged at fixed time intervals (e.g., 40 milliseconds corresponding to 25 Hz). This process eliminates the phase difference caused by asynchronous acquisition and transmission, ensuring that subsequent filtering, feature calculation, and model inference are based on the physical state at the same moment.

[0050] In one feasible implementation, step S201 specifically includes: Step A10: Align the strain data and the vehicle state data according to the preset target sampling frequency, and obtain the jitter time of the aligned strain data and vehicle state data.

[0051] It should be noted that in this step, the system presets a target sampling frequency (e.g., Fs = 25 Hz). During data alignment, the system calculates the difference between the timestamp of each actually arriving data point and the ideal equally spaced time grid point; this difference is the jitter time (Δt). For example, if a signal point that should appear at t=1.000s actually arrives at t=1.015s, its jitter time is +15ms. The system calculates or records this jitter time for each data point of each data channel to guide subsequent correction strategies. Furthermore, it should be noted that the jitter time reflects the deviation of the actual data arrival time from the theoretical sampling time. It is mainly caused by factors such as the uncertainty of network transmission delay and the slight randomness of ECU scheduling.

[0052] Step A20: Based on the jitter time, the strain data and the vehicle state data are time-aligned using a preset strategy to obtain a signal time series with a uniform sampling rate. The preset strategy includes any one of the following: a hold strategy, a linear interpolation strategy, and a missing label strategy.

[0053] It should be noted that in this step, the system employs a tiered processing strategy based on the magnitude of the jitter time Δt: **Maintenance Strategy:** When |Δt| < 10ms (small jitter), the jitter is considered acceptable, and the timestamp of the data point is not corrected; it is directly assigned to the nearest theoretical time grid point, and its value remains unchanged. **Linear Interpolation Strategy:** When 10ms ≤ |Δt| < 200ms (medium jitter), to more accurately reconstruct the signal value at the theoretical sampling time, the system uses adjacent data points before and after this data point to calculate an estimate of the theoretical sampling time through linear interpolation. **Missing Data Strategy:** When |Δt| ≥ 200ms or the packet loss time is too long, the data in that time period is considered unreliable, and it is marked as missing data (NaN) at the corresponding time grid point.

[0054] Understandably, the preset strategies are layered solutions designed for timing problems of varying severity. The preservation strategy avoids unnecessary computation and improves efficiency; the linear interpolation strategy achieves a good balance between accuracy and complexity and can effectively correct moderate timing deviations; the missing data labeling strategy ensures the robustness of the system in the event of severe data anomalies and prevents erroneous data from contaminating subsequent processes.

[0055] Step S202: Perform multi-level filtering on the strain data and vehicle state data in the signal time series to obtain filtered strain data and target vehicle state data.

[0056] It should be noted that the aligned data still contains various types of noise. The system employs a cascaded multi-stage filtering chain for processing. For strain data, the system uses a Hampel filter (or a similar robust filter) to identify and remove pulse-like outliers caused by electromagnetic interference, etc. Then, a low-pass filter (such as a second-order Butterworth low-pass filter) is used to filter out high-frequency noise (such as high-frequency road surface excitation and power frequency interference) above the load change frequency. Finally, a Savitzky-Golay filter is applied for local polynomial smoothing, further reducing random noise while preserving signal trends and peak characteristics. For vehicle state data (such as acceleration and angular velocity), low-pass filtering is mainly performed to smooth high-frequency fluctuations, facilitating subsequent steady-state assessment.

[0057] Step S203: Perform temperature compensation on the filtered strain data according to the preset temperature compensation model to obtain the temperature-compensated target strain data.

[0058] It should be noted that the resistance of the strain gauge, the properties of the adhesive layer, and the elastic modulus of the leaf spring all change with temperature, causing a drift in the strain measurement values ​​that is independent of the load (temperature drift). To eliminate this effect, the system employs a preset temperature compensation model. This model takes the strain gauge temperature (measured by a temperature sensor mounted close to the strain gauge) as input, calculates a temperature drift, and then subtracts this drift from the filtered strain data to obtain the target strain data that is only related to the mechanical load and whose temperature influence has been corrected.

[0059] In one feasible implementation, step S203 specifically includes: Step B10: Obtain the strain gauge temperature of the vehicle leaf spring.

[0060] It should be noted that the strain gauge temperature refers to a direct physical quantity measuring the temperature of the strain gauge itself and the adjacent bonding area (including the temperature of the left and right strain gauges). Using a proximity temperature probe instead of ambient temperature allows for more accurate capture of heat source changes that cause strain gauge output drift.

[0061] Step B20: Calculate the temperature drift based on the preset temperature compensation model, the preset reference temperature, and the strain gauge temperature.

[0062] It should be noted that the system will call the temperature compensation coefficients (k1, k2) and reference temperature (T0) pre-stored in the model parameter library. The real-time strain gauge temperature (T) will be substituted into the compensation model (e.g., Δε_temp = k1). (T-T0) +k2 (T-T0)^2) can be used to calculate in real time the spurious change in strain measurement caused by the current temperature relative to the reference temperature, i.e., the temperature drift.

[0063] Step B30: Correct the filtered strain data according to the temperature drift to obtain the temperature-compensated target strain data, which includes the left-side temperature-compensated strain and the right-side temperature-compensated strain.

[0064] It should be noted that in this step, the system subtracts the corresponding temperature drift from each filtered strain signal (e.g., the strain on the left side of the left axis εL_filt, and the strain on the right side of the left axis εR_filt). Each strain has its own independent temperature probe and compensation coefficient. Specifically, the calculations are: εL_T = εL_filt - Δε_temp_L, εR_T = εR_filt - Δε_temp_R. The calculated εL_T and εR_T are the left-side temperature-compensated strain and the right-side temperature-compensated strain, which together constitute the target strain data. The purpose of this step is to eliminate the interference of temperature on the strain data, thus maintaining a stable load-strain relationship.

[0065] Step S204: Based on the target vehicle state data and the target strain data, determine the steady-state window of the current data acquisition time window to obtain a steady-state flag.

[0066] It should be noted that the system analyzes data within a sliding time window (e.g., 2-5 seconds) to determine whether the vehicle is in a preset steady-state condition. The judgment is based on two main factors: first, whether the vehicle's kinematic state is stable; and second, whether the strain signal itself exhibits minimal fluctuations. The system checks whether the data at all times within the window continuously meets a series of preset steady-state conditions. If these conditions are met, the time window is determined to be a steady-state window, and a steady-state flag is generated (e.g., set to 1 or TRUE); otherwise, the flag is set to non-steady-state (0 or FALSE).

[0067] Understandably, the purpose of this step is to identify periods when vehicles are nearly stationary or traveling at a constant, low speed in a straight line without significant acceleration, deceleration, or turning. During these periods, road disturbances and vehicle dynamics are minimal, allowing strain-based load estimation to achieve accuracy close to that of static weighbridge measurements.

[0068] In one feasible implementation, step S204 specifically includes: Step C10: Obtain the duration of the data in the current data acquisition time window.

[0069] It should be noted that the duration refers to the time during which the collected data remains unchanged within a preset time window. A brief moment of stability may be accidental, and stability for a certain period of time (e.g., ≥2 seconds) is required to effectively filter out transient interference and improve the reliability of steady-state judgment.

[0070] Step C20: When the duration is greater than or equal to the preset steady-state time and the vehicle speed, longitudinal acceleration, lateral acceleration and steering angular velocity in the target vehicle state data all meet the corresponding preset steady-state conditions, the time window is determined to be a steady-state window, and a steady-state flag is added to the time window.

[0071] It should be noted that, once sufficient data is available, the system will check whether the following preset steady-state conditions are continuously met throughout the entire duration: vehicle speed (v) is below a threshold, for example, v < 5 km / h; the absolute value of longitudinal acceleration (|ax|) is below a threshold, for example, |ax| < 0.03 g; the absolute value of lateral acceleration (|ay|) is below a threshold, for example, |ay| < 0.03 g; and the absolute value of yaw rate (|yaw_rate|) is below a threshold, for example, |yaw_rate| < 2 deg / s. If all checked items are met simultaneously, the system will determine the current time window as a steady-state window and mark the center or end time of that window with a steady-state flag.

[0072] Additionally, it should be noted that the preset steady-state conditions are a set of quantified physical thresholds. The vehicle speed condition ensures that the vehicle is moving at low speed or stationary; the acceleration and angular velocity conditions ensure that the vehicle does not exhibit significant linear acceleration or rotational motion.

[0073] In this embodiment, by comprehensively utilizing time alignment, multi-level filtering, temperature compensation, and steady-state window determination techniques, the problem of inaccurate estimation by traditional strain methods caused by asynchronous data, noise interference, temperature drift, and vehicle dynamics is effectively solved. This achieves high-quality purification and intelligent identification of operating conditions for the original measurement signal, providing a reliable data foundation for subsequent high-precision and robust load estimation.

[0074] Based on the first and second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to that in embodiments one and two above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 5 , Figure 5 This is a flowchart illustrating the third embodiment of the vehicle load estimation method of this application. Step S30 specifically includes: Step S301: Obtain temperature data and vehicle configuration information.

[0075] It should be noted that the temperature data includes strain gauge temperature T, ambient temperature T_amb, and ECU temperature T_ecu, used to capture any residual temperature drift or thermal hysteresis. The vehicle configuration information typically includes vehicle model / platform, axle type (e.g., front axle, drive rear axle, steering axle), and leaf spring model (including part number, number of leaves, rated load, etc.). This information is converted into fixed-dimensional learnable embedding vectors before being input into the model, enabling the same model to adapt to the different characteristics of various vehicles.

[0076] Step S302: Calculate the sum of strains, strain difference, and left-right strain ratio based on the left-side temperature-compensated strain and right-side temperature-compensated strain in the target strain data.

[0077] It should be noted that in this step, the system calculates three physically meaningful features in real time based on the target strain data (εL_T and εR_T) on both sides of each axle. The calculation formulas are as follows: strain sum Σ = εL_T + εR_T; strain difference Δ = εL_T - εR_T; left / right strain ratio R / L = εR_T / (εL_T + eps), where eps is a constant to prevent division by zero (e.g., 10 με). After the calculation is completed, the left / right strain ratio R / L is usually truncated (e.g., limited to the range of [-10, 10]).

[0078] It is understandable that the strain (Σ) has a strong monotonic relationship with the total load on the shaft and is the core feature characterizing the total load; the strain difference (Δ) directly reflects the off-center load of the left and right wheels and the possible installation asymmetry; the left and right strain ratio (R / L) is more sensitive to off-center load and has a weak correlation with the total load, providing independent off-center load information and enhancing the robustness of the model under asymmetric load.

[0079] Step S303: Construct time-series features based on the target vehicle state data, the left-side temperature-compensated strain, the right-side temperature-compensated strain, the sum of strains, the strain difference, and the left-right strain ratio. The target vehicle state data includes vehicle speed, longitudinal acceleration, lateral acceleration, pitch angle, and roll angle.

[0080] It should be noted that in this step, the system stacks a series of time-correlated signals at the same time interval to form a multi-dimensional time-series array. Specifically, for each sampling time t, the following set of signal values ​​is selected to form a feature vector: [left temperature-compensated strain εL_T(t), right temperature-compensated strain εR_T(t), strain sum Σ(t), strain difference Δ(t), left / right strain ratio R / L(t), vehicle speed v(t), longitudinal acceleration ax(t), lateral acceleration ay(t), pitch angle pitch(t), roll angle roll(t)]. The first five are strain-related features, and the last five are vehicle dynamic features. Arranging the feature vectors from multiple consecutive time intervals (e.g., within a 2- to 5-second window) in chronological order constitutes the time-series feature.

[0081] Understandably, the temporal features capture the key dynamic processes and instantaneous states in load estimation. They preserve the complete form of the signal's changes over time, enabling subsequent temporal models (such as 1D-CNN and TCN) to learn the complex spatiotemporal relationships between load, vehicle motion, and strain response.

[0082] Step S304: Within the time window indicated by the steady-state flag, calculate the window statistical characteristics and steady-state score based on the statistical characteristics of one or more data in the time series characteristics.

[0083] It should be noted that, for the specific time window identified by the steady-state indicator, the system performs statistical analysis on the time-series characteristic sequences within that window. First, a series of statistical characteristics are calculated for the key physical quantity sequences within the window (especially εL_T, εR_T, Σ, Δ), such as mean, variance, range (difference between maximum and minimum values), root mean square value, and slope of the linear trend. Second, by combining vehicle dynamics (v, |ax|, |ay|, |yaw_rate|), strain sum (Σ), and variance or range, a quantified steady-state score is calculated using threshold comparison and weighted summation to characterize the degree of static state within that window.

[0084] Step S305: Construct scalar features based on the window statistical features, the steady-state score, the temperature data, and the vehicle configuration information.

[0085] It should be noted that in this step, the system concatenates all non-time-series information or information representing the overall window attributes (mean, variance, range, root mean square value, and slope of the linear trend) into a one-dimensional feature vector. Specifically, the window statistical features (flattened statistical quantities such as the mean and variance of multiple physical quantities), steady-state score, temperature data (T, T_amb, T_ecu), and configuration embedding vectors mapped from vehicle configuration information are connected to form the final scalar feature.

[0086] In this embodiment, by constructing a time-series and scalar dual-branch feature that integrates physical characteristics (strain sum / difference / ratio) and statistical laws (window statistics, steady-state scoring) from preprocessed data, a unified representation of load, off-center load, and vehicle configuration differences is achieved. This solves the problems of poor model generalization ability and inaccurate estimation caused by neglecting dynamic details, off-center load effects, and vehicle model differences in traditional methods, and significantly improves the accuracy, robustness, and cross-vehicle adaptability of load estimation.

[0087] This application also provides a vehicle load estimation device, please refer to... Figure 6 The vehicle load estimation device includes: The data acquisition module 10 is used to acquire strain data of the vehicle leaf spring and vehicle status data; Data processing module 20 is used to process the strain data and vehicle state data to obtain target strain data, target vehicle state data and steady-state flag. The data processing includes alignment and resampling, filtering and temperature correction. The feature construction module 30 is used to construct time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data and the steady-state flag; The load prediction module 40 is used to input the time series features and the scalar features into the load prediction model to obtain the estimated vehicle load.

[0088] The vehicle load estimation device provided in this application, employing the vehicle load estimation method in the above embodiments, can solve the technical problem of how to improve the accuracy, stability, and cross-vehicle generalization ability of load estimation. Compared with the prior art, the beneficial effects of the vehicle load estimation device provided in this application are the same as those of the vehicle load estimation method provided in the above embodiments, and other technical features in the vehicle load estimation device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0089] In one embodiment, the data processing module 20 is further configured to perform time alignment and resampling on the strain data and the vehicle state data to obtain a signal time series with a uniform sampling rate; perform multi-level filtering on the strain data and vehicle state data in the signal time series to obtain filtered strain data and target vehicle state data; perform temperature compensation on the filtered strain data according to a preset temperature compensation model to obtain temperature-compensated target strain data; and perform steady-state window judgment on the current data acquisition time window based on the target vehicle state data and the target strain data to obtain a steady-state flag.

[0090] In one embodiment, the data processing module 20 is further configured to align the strain data and the vehicle state data according to a preset target sampling frequency, and obtain the jitter time of the aligned strain data and the vehicle state data; and to perform time alignment of the strain data and the vehicle state data according to the jitter time using a preset strategy to obtain a signal time series with a uniform sampling rate, wherein the preset strategy includes any one of a preservative strategy, a linear interpolation strategy, and a missing label strategy.

[0091] In one embodiment, the data processing module 20 is further configured to acquire the strain gauge temperature of the vehicle leaf spring; calculate the temperature drift based on a preset temperature compensation model, a preset reference temperature, and the strain gauge temperature; and correct the filtered strain data based on the temperature drift to obtain temperature-compensated target strain data, wherein the target strain data includes left-side temperature-compensated strain and right-side temperature-compensated strain.

[0092] In one embodiment, the data processing module 20 is further configured to obtain the duration of data in the current data acquisition time window; when the duration is greater than or equal to a preset stable time and the vehicle speed, longitudinal acceleration, lateral acceleration and steering angular velocity in the target vehicle state data all meet the corresponding preset steady-state conditions, the time window is determined to be a steady-state window, and a steady-state flag is added to the time window.

[0093] In one embodiment, the feature construction module 30 is further configured to acquire temperature data and vehicle configuration information; calculate the sum of strains, strain difference, and left-right strain ratio based on the left-side temperature-compensated strain and right-side temperature-compensated strain in the target strain data; construct a time-series feature based on the target vehicle state data, the left-side temperature-compensated strain, the right-side temperature-compensated strain, the sum of strains, the strain difference, and the left-right strain ratio, wherein the target vehicle state data includes vehicle speed, longitudinal acceleration, lateral acceleration, pitch angle, and roll angle; within the time window indicated by the steady-state indicator, calculate window statistical features and a steady-state score based on the statistical features of one or more data in the time-series feature; and construct a scalar feature based on the window statistical features, the steady-state score, the temperature data, and the vehicle configuration information.

[0094] In one embodiment, the load prediction module 40 is further configured to, when the steady-state flag indicates that the current state is a steady-state window, perform exponential smoothing on the vehicle load estimate output by the load prediction model according to a preset smoothing coefficient to obtain a smoothed vehicle load estimate; and update the vehicle load estimate based on the smoothed vehicle load estimate.

[0095] This application provides a vehicle load estimation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the vehicle load estimation method in Embodiment 1 above.

[0096] The following is for reference. Figure 7 The diagram illustrates a structural schematic of a vehicle load estimation device suitable for implementing embodiments of this application. The vehicle load estimation device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The vehicle load estimation device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0097] like Figure 7As shown, the vehicle load estimation device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM (Read Only Memory) 1002 or a program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the vehicle load estimation device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the vehicle load estimation device to communicate wirelessly or wiredly with other devices to exchange data. Although a vehicle load estimation device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0098] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0099] The vehicle load estimation device provided in this application, employing the vehicle load estimation method described in the above embodiments, can solve the technical problem of how to improve the accuracy, stability, and cross-vehicle generalization ability of load estimation. Compared with the prior art, the beneficial effects of the vehicle load estimation device provided in this application are the same as those of the vehicle load estimation method provided in the above embodiments, and other technical features in this vehicle load estimation device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0100] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0101] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0102] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the vehicle load estimation method in the above embodiments.

[0103] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0104] The aforementioned computer-readable storage medium may be included in the vehicle load estimation device; or it may exist independently and not be installed in the vehicle load estimation device.

[0105] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a vehicle load estimation device, cause the vehicle load estimation device to: acquire strain data of the vehicle leaf spring and vehicle state data; perform data processing on the strain data and vehicle state data to obtain target strain data, target vehicle state data, and a steady-state flag, wherein the data processing includes alignment and resampling, filtering, and temperature correction; construct temporal features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag; and input the temporal features and the scalar features into a load prediction model to obtain a vehicle load estimate.

[0106] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0107] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0108] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0109] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described vehicle load estimation method. This addresses the technical problem of improving the accuracy, stability, and cross-vehicle generalization ability of load estimation. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the vehicle load estimation method provided in the above embodiments, and will not be elaborated upon here.

[0110] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the vehicle load estimation method described above.

[0111] The computer program product provided in this application can solve the technical problem of how to improve the accuracy, stability, and cross-vehicle generalization ability of load estimation. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the vehicle load estimation method provided in the above embodiments, and will not be repeated here.

[0112] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for estimating the load capacity of a vehicle, characterized in that, The method includes: Acquire strain data of the vehicle leaf springs and vehicle status data; The strain data and vehicle state data are processed to obtain target strain data, target vehicle state data and steady-state flags. The data processing includes alignment and resampling, filtering and temperature correction. Based on the target strain data, the target vehicle state data, and the steady-state indicator, a time-series feature and a scalar feature for load estimation are constructed. The time-series features and the scalar features are input into the load prediction model to obtain the estimated vehicle load.

2. The method as described in claim 1, characterized in that, The step of processing the strain data and vehicle state data to obtain target strain data, target vehicle state data, and steady-state indicators includes: The strain data and the vehicle state data are time-aligned and resampled to obtain a signal time series with a uniform sampling rate. The strain data and vehicle state data in the signal time series are subjected to multi-level filtering to obtain filtered strain data and target vehicle state data. The filtered strain data is temperature-compensated according to a preset temperature compensation model to obtain the temperature-compensated target strain data. Based on the target vehicle state data and the target strain data, a steady-state window determination is performed on the current data acquisition time window to obtain a steady-state flag.

3. The method as described in claim 2, characterized in that, The step of performing time alignment and resampling on the strain data and the vehicle state data to obtain a signal time series with a uniform sampling rate includes: The strain data and the vehicle status data are aligned according to a preset target sampling frequency, and the jitter time of the aligned strain data and vehicle status data is obtained. The strain data and the vehicle state data are time-aligned according to the jitter time using a preset strategy to obtain a signal time series with a uniform sampling rate. The preset strategy includes any one of the following: a hold strategy, a linear interpolation strategy, and a missing label strategy.

4. The method as described in claim 2, characterized in that, The step of performing temperature compensation on the filtered strain data according to a preset temperature compensation model to obtain temperature-compensated target strain data includes: Obtain the strain gauge temperature of the vehicle leaf spring; The temperature drift is calculated based on the preset temperature compensation model, the preset reference temperature, and the strain gauge temperature. The filtered strain data is corrected based on the temperature drift to obtain temperature-compensated target strain data, which includes left-side temperature-compensated strain and right-side temperature-compensated strain.

5. The method as described in claim 2, characterized in that, The step of determining the steady-state window based on the target vehicle state data and the target strain data to obtain the steady-state indicator includes: Get the duration of the data in the current data acquisition time window; When the duration is greater than or equal to a preset stabilization time and the vehicle speed, longitudinal acceleration, lateral acceleration, and steering angular velocity in the target vehicle state data all meet the corresponding preset steady-state conditions, the time window is determined to be a steady-state window, and a steady-state flag is added to the time window.

6. The method as described in claim 1, characterized in that, The step of constructing time-series and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state indicator includes: Acquire temperature data and vehicle configuration information; Based on the left-side temperature-compensated strain and right-side temperature-compensated strain in the target strain data, calculate the strain sum, strain difference, and left-right strain ratio. The time-series features are constructed based on the target vehicle state data, the left-side temperature-compensated strain, the right-side temperature-compensated strain, the sum of strains, the strain difference, and the left-right strain ratio. The target vehicle state data includes vehicle speed, longitudinal acceleration, lateral acceleration, pitch angle, and roll angle. Within the time window indicated by the steady-state indicator, window statistical features and a steady-state score are calculated based on the statistical features of one or more data in the time-series features. Based on the window statistical features, the steady-state score, the temperature data, and the vehicle configuration information, scalar features are constructed.

7. The method as described in claim 1, characterized in that, After the step of inputting the time-series features and the scalar features into the load prediction model to obtain the vehicle load estimate, the method further includes: When the steady-state flag indicates that the current state is a steady-state window, the vehicle load estimate output by the load prediction model is exponentially smoothed according to a preset smoothing coefficient to obtain a smoothed vehicle load estimate. The vehicle load estimate is updated based on the smoothed vehicle load estimate.

8. A device for estimating the load of a vehicle, characterized in that, The device includes: The data acquisition module is used to acquire strain data of the vehicle leaf springs and vehicle status data; The data processing module is used to process the strain data and vehicle state data to obtain target strain data, target vehicle state data and steady-state flags. The data processing includes alignment and resampling, filtering and temperature correction. The feature construction module is used to construct time-series features and scalar features for load estimation based on the target strain data, the target vehicle state data, and the steady-state flag. The load prediction module is used to input the time-series features and the scalar features into the load prediction model to obtain the estimated vehicle load.

9. A vehicle load estimation device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method for estimating the vehicle load as claimed in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the vehicle load estimation method as described in any one of claims 1 to 7.

Images