Weight measuring system in vehicle motion state and control method thereof

Abstract

The invention discloses a weight measuring system in a vehicle motion state and a control method thereof, and relates to the field of vehicle-mounted weighing, weight data are collected from a strain gauge sensor and a tilt angle sensor, and an algorithm processor is used for processing the weight data and compensating a dynamic effect; and fitting a neural network model by using a multi-section function, predicting a vehicle load measurement value by training the neural network model, and outputting the vehicle load measurement value in real time. The high precision of the neural network is reserved, the efficiency of the system is remarkably improved, delay is reduced, and the problem that a traditional system is inaccurate in measurement in a motion state is solved. According to the method, the neural network model is fitted through the multi-segment function, through approximation fitting, the calculation efficiency and the real-time performance are improved, meanwhile, the measurement accuracy is kept, powerful technical support is provided for application of the vehicle-mounted weighing system, delay is reduced, and the method is more suitable for the vehicle-mounted weighing system with the high real-time performance requirement.

Description

Technical Field

[0001] The present invention relates to the field of vehicle weighing, and particularly to a weight measurement system and its control method in the vehicle's moving state. Background Art

[0002] In existing vehicle weighing control algorithms, PID control is used. However, in a highly dynamic environment, the performance of the PID controller may be limited because its parameters are usually fixed and cannot adapt to rapidly changing conditions. Neural network control using machine learning methods is used, but it requires a large amount of data for training and may be computationally intensive in a real-time system, resulting in delays. Sliding mode control is used, but there may be high-frequency oscillations, affecting the smoothness of the control signal.

[0003] Due to factors such as uneven loading when the vehicle is loaded and bumps during driving, traditional weighing systems are difficult to provide accurate weight data. The present invention integrates motion sensors and weight sensors and uses an algorithm based on a BP neural network for data processing to achieve accurate measurement of the load when the vehicle is in a moving state.

[0004] In view of this, the present invention provides a weight measurement system and its control method in the vehicle's moving state. Summary of the Invention

[0005] The object of the present invention is to provide a weight measurement system and its control method in the vehicle's moving state to solve the deficiencies in the background art.

[0006] According to one aspect of the present invention, a weight measurement control method in the vehicle's moving state is provided, including the following steps:

[0007] Step S1: Collect the original input data of the vehicle, where the original input data includes weight data, motion state data, vehicle speed information, and driving environment data;

[0008] Step S2: Use an algorithm processor to process the original input data for signal processing and extract and compensate for dynamic effects;

[0009] Step S3: Use a multi-segment function fitting neural network model to predict the vehicle load measurement value by training the neural network model and output the vehicle load measurement value in real time.

[0010] As a preferred technical solution of the present invention, the system parameters are automatically adjusted according to the vehicle type and driving environment.

[0011] As a preferred technical solution of the present invention, it also has a self-calibration function and can adapt to sensor drift.

[0012] As a preferred technical solution of the present invention, according to the feedback data in actual applications, continuously optimize the parameters of the neural network model and the piecewise function through continuous learning and adjustment.

[0013] According to another aspect of the present invention, there is provided a weight measurement system under vehicle motion state, based on the implementation of a weight measurement control method under vehicle motion state, including a data acquisition module, a signal processor, an algorithm processing module, and a signal output end. Each module is connected through a communication module.

[0014] The data acquisition module collects the original input data of the vehicle, and the original input data includes weight data, motion state data, vehicle speed information, and driving environment data.

[0015] The signal processor performs signal processing on the original input data to obtain the compensation dynamic effect of the vehicle motion state on the weight data.

[0016] The algorithm processing module uses a piecewise function to fit the neural network model, and uses the trained neural network model to fit and predict the vehicle load measurement value for the original input data.

[0017] The parameter resource library uses a piecewise function to fit the neural network model, and determines the optimal parameters of each piecewise function by minimizing the fitting error.

[0018] The signal output end outputs the calculated vehicle load measurement value result to the in-vehicle display screen or the remote monitoring platform.

[0019] In the above technical solution, the technical effects and advantages provided by the present invention are as follows:

[0020] The present invention not only retains the high precision of the neural network, but also significantly improves the efficiency of the system, reduces the delay, and solves the problem of inaccurate measurement of the traditional system under the motion state. By using a piecewise function to fit the neural network model, through this approximation fitting, the calculation efficiency and real-time performance are improved, while maintaining the measurement accuracy, which provides strong technical support for the application of the in-vehicle weighing system, reduces the delay, and makes it more suitable for the in-vehicle weighing system with high real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings.

[0022] Figure 1 It is a schematic diagram of the in-vehicle weighing process in the method of the present invention;

[0023] Figure 2 is the specific training flowchart of the model in the method of the present invention;

[0024] Figure 3 is the schematic flowchart of the process in the method of the present invention;

[0025] Figure 4 is the framework diagram of the weight measurement control system in the method of the present invention. Specific embodiments

[0026] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Embodiment 1, please refer to Figure 4 As shown, a weight measurement system in a vehicle motion state in this embodiment includes a data acquisition module, a signal processor, an algorithm processing module, and a signal output end, and each module is connected through a communication module;

[0028] The data acquisition module acquires the original input data of the vehicle, and the original input data includes weight data, motion state data, vehicle speed information, and driving environment data.

[0029] Specifically, in the motion state, the driving environment data (such as road conditions, weather, etc.) when the vehicle is driving, and when the current vehicle is driving, the weight data of the vehicle is acquired based on a strain gauge sensor, and the strain signal of the vehicle under different load states is used to directly reflect the weight change of the vehicle.

[0030] Based on an inclination sensor, the motion state data of the vehicle (such as dynamic effects such as bumps and tilts) is acquired to compensate for the interference of the weight measurement caused by dynamic effects (such as bumps and partial loads) during the driving of the vehicle;

[0031] The vehicle speed information includes vehicle speed and vehicle acceleration; the vehicle speed and vehicle acceleration are respectively acquired based on a speed sensor and an acceleration sensor; and the driving environment data is acquired based on an image acquisition device; in practical applications, the vehicle speed information and the driving environment data can be obtained in real time by a driving recorder.

[0032] The signal processor performs signal processing on the original input data to obtain the compensation dynamic effect of the vehicle motion state on the weight data;

[0033] It should be noted that during vehicle driving, the influence of dynamic effects (such as jolts, tilts, etc.) on weight measurement is compensated to improve the accuracy of load measurement.

[0034] Specifically, the acquisition logic for compensating dynamic effects is as follows:

[0035] Based on the original input data, strain signals of the vehicle in different load states and the vehicle's motion state are captured, and information required to compensate for dynamic effects is extracted.

[0036] Analyze the vehicle motion state data to identify and quantify the analysis results of the dynamic effects on the vehicle during driving.

[0037] Establish a compensation model. According to the analysis results, one or more compensation models are established to simulate and predict the influence of dynamic effects on weight measurement. Based on physical principles, empirical formulas, or machine learning algorithms.

[0038] Model training and optimization. Use the collected data to train and optimize the compensation model, predict the influence of dynamic effects, and adjust the model parameters.

[0039] During vehicle driving, the compensation model is applied in real-time to compensate the weight measurement data.

[0040] The acquisition logic for compensating dynamic effects can ensure accurate and reliable weight measurement under the vehicle's motion state. This logic is an indispensable part of the vehicle load measurement system, and it helps to solve the problem of inaccurate measurement in traditional systems under motion states.

[0041] Algorithm processing module. Use a multi-segment function to fit a neural network model, and use the trained neural network model to fit the original input data to predict the vehicle load measurement value.

[0042] It should be noted that multi-segment function fitting can better adapt to data changes in different load ranges and working conditions, and can output the vehicle load measurement value in real-time and accurately. This method can effectively compensate for the errors caused by dynamic effects, improve the accuracy and reliability of load measurement, and is applicable to application scenarios such as vehicle load monitoring.

[0043] Specifically, by collecting data of the vehicle under different loads, speeds, and road conditions, a neural network model is trained to accurately predict the load. Then, the trained neural network model is fitted with multi-segment linear or polynomial functions, and the optimal parameters of each segment function are determined by minimizing the fitting error.

[0044] In practical applications, the data of the weight sensor and the motion sensor are input into the data processing unit in real time. The system selects the corresponding piecewise function according to the original input data for calculation, quickly obtains the load result, and transmits it to the in-vehicle display screen or the remote monitoring platform through the communication module. Since the computational complexity of the piecewise function is much lower than that of the neural network, the system can complete the load calculation within milliseconds, meeting the real-time requirement. At the same time, by optimizing the design of the piecewise function, it maintains a measurement accuracy similar to that of the neural network model. In addition, the system also has adaptability and can automatically adjust the parameters of the piecewise function according to the vehicle type and driving environment to ensure applicability under different conditions. Through this embodiment, the present invention demonstrates how to fit the neural network model with a piecewise function to achieve efficient and accurate in-vehicle weighing control, solve the problem of low efficiency of the traditional neural network, and significantly improve the real-time performance and adaptability of the system.

[0045] Predict the vehicle load measurement value of the original input data through the trained neural network model. The vehicle load measurement value represents the real-time load of the vehicle. The influence of dynamic effects (such as bumps, tilts, etc.) on the weight measurement is eliminated through the neural network model, ensuring the accuracy of the measurement.

[0046] Specifically, for the neural network model, a BP neural network model is trained using the collected data. During the training process, the neural network learns how to predict the load based on the sensor data and adapt to different dynamic environments (such as bumps, partial loads, etc.).

[0047] The parameter resource library uses a piecewise function to fit the neural network model. By minimizing the fitting error, the optimal parameters of each piece of the function are determined to simplify the computational complexity and maintain high accuracy.

[0048] It should be noted that: to determine the optimal parameters of each piece of the function by minimizing the fitting error and implement the piecewise function fitting the neural network model in the in-vehicle weighing system, the following steps are included:

[0049] Use the preprocessed data to train a neural network model (such as a BP neural network) so that it can learn the non-linear relationship between the vehicle load and the sensor output.

[0050] Fit the output of the trained neural network model with a piecewise linear or polynomial function. The piecewise function should be able to cover the entire range of the output of the neural network model, and each piece of the function can well approximate the output of the neural network within its respective interval.

[0051] For each piece of the function, determine its optimal parameters by minimizing the fitting error, where:

[0052] Define the error function: Usually, the mean square error (MSE) or other similar error metrics are used as the error function.

[0053] Parameter optimization: Use optimization algorithms (such as gradient descent, genetic algorithm, particle swarm optimization, etc.) to adjust the parameters of each segment of the function to minimize the error function.

[0054] Verification and adjustment: Evaluate the fitting effect on the validation set and adjust the parameters of the model structure or optimization algorithm if necessary.

[0055] Each segment of the function can well approximate the output of the neural network within its respective interval. The selection of the segmentation points is reasonable to avoid large fitting errors near the segmentation points, and the generalization ability of the model is ensured through methods such as cross-validation. The calculation of multiple segments of functions (especially linear functions) is usually much simpler than the forward propagation of the neural network. By using multiple segments of functions instead of the complete neural network model, the computational complexity can be significantly reduced.

[0056] In practical applications, according to the real-time sensor data of the vehicle, select the corresponding segmented function for calculation to quickly obtain the load result. This method can complete the load calculation within milliseconds, meeting the real-time requirements.

[0057] Specifically, in practical applications, the data of the weight sensor and the motion sensor are input into the data processing unit in real time; select the corresponding segmented function for calculation according to the original input data to quickly obtain the load result; use techniques such as mean filtering to separate the true weight from the fluctuations caused by motion and compensate for the dynamic effects.

[0058] It should be noted that: The system has self-adaptability and can automatically adjust the parameters of the segmented function according to the vehicle type and driving environment to ensure applicability under different conditions. The calculated load result is transmitted to the in-vehicle display screen or the remote monitoring platform through the communication module. It can complete the load calculation within milliseconds, meeting the real-time requirements; the system has a self-calibration function and can adapt to the drift of the sensor to ensure accuracy during long-term use. It realizes the accurate measurement of the load under the vehicle's moving state and solves the problems of inaccurate measurement and excessive delay of traditional systems in dynamic environments.

[0059] Automatically adjust the system parameters according to the vehicle type and driving environment; that is to say, the weight distribution and driving characteristics of different vehicles (such as cars, trucks, buses, etc.) are different, and the performance of the same vehicle will also vary under different environments (such as urban roads, highways, mountainous areas, etc.). Therefore, it is necessary to be able to identify these differences and adjust the parameters accordingly. That is, identify the vehicle type according to information such as the vehicle model, size, load capacity, etc., and collect driving environment data through sensors (such as GPS, meteorological sensors, road condition sensors, etc.). According to the identified vehicle type and the perceived environmental conditions, the system automatically adjusts the parameters of the neural network model and the segmented function to adapt to the specific vehicle and environment.

[0060] It also has a self - calibration function and can adapt to the drift of the sensor; the drift of the sensor refers to the offset of the sensor output over time. Therefore, regularly performing a calibration procedure to achieve the self - calibration function can ensure the long - term stable operation of the system. By comparing the sensor output with a known reference value, the drift of the sensor is monitored in real - time. Once drift is detected, the system automatically adjusts the sensor output or model parameters to compensate for the drift.

[0061] According to the feedback data in actual applications, continuously optimize the parameters of the neural network model and the piece - wise function. Through continuous learning and adjustment, the system can adapt to more vehicle types and complex environmental conditions, further improving the measurement accuracy and real - time performance. By continuously learning and adjusting, the system can better adapt to different vehicle types and complex environmental conditions, improve the measurement accuracy and real - time performance, compare the actual measurement results with the expected results, and adjust the model parameters according to the differences, automatically adjusting the model parameters to minimize the prediction error.

[0062] The adaptive adjustment, self - calibration function, and continuous optimization mechanism in the vehicle - mounted weighing system can adapt to different vehicle types and driving environments, improve the measurement accuracy and real - time performance, and more accurate and reliable load measurement results are of great significance for improving transportation efficiency, ensuring driving safety, etc.

[0063] The signal output terminal outputs the calculated vehicle load measurement result to the vehicle - mounted display screen or the remote monitoring platform, completes the load calculation within milliseconds, continues to monitor the vehicle state, prepares for the next round of data collection and processing, and optimizes the system performance according to the feedback in actual applications.

[0064] This embodiment forms a closed - loop system, covering the whole process from data collection, processing, calculation to result output. Through technologies such as multi - sensor fusion, neural network models, piece - wise function fitting, and dynamic compensation, the system can efficiently and accurately measure the load under the vehicle's moving state, and has adaptive and self - calibration functions to ensure applicability and accuracy under different conditions.

[0065] Embodiment 2

[0066] This embodiment provides a training method for a neural network model, including the following steps:

[0067] 1. Data pre - processing includes:

[0068] Filtering and mean processing: The original 60,000 groups of sensor signals (in this embodiment, 4 - way pressure signals collected by strain - gauge sensors corresponding to 4 mechanical transmission structures and 1 - way inclination XY - axis collected by an inclination sensor) are filtered to remove noise, and the moving mean is calculated to smooth the data.

[0069] Standardization: Select 200 groups of typical data and perform Z-score standardization (mean is 0, standard deviation is 1) to eliminate the dimensional difference.

[0070] Dataset division: After standardization, the data is divided into 8:2. 160 groups are used as the training set for parameter learning, and 40 groups are used as the validation set to evaluate the generalization performance.

[0071] 2. The network topology design corresponding to the neural network model includes:

[0072] Input layer: 5 neurons, corresponding to the 5-dimensional input of the sensor signal.

[0073] Hidden layer: 4 neurons, and the activation function is Sigmoid (formula: f(x) = 1 + e - x1) for non-linear feature mapping.

[0074] Output layer: 1 neuron (weight prediction value), and a linear activation function (f(x) = x) is used to directly output continuous values.

[0075] 3. The parameter initialization settings are as follows:

[0076] Weights and biases: The weight matrix from the input layer to the hidden layer (5×4) and the weight matrix from the hidden layer to the output layer (4×1) are randomly initialized (such as a normal distribution), and the bias terms are initialized to 0 or small random numbers.

[0077] Hyperparameter settings: Learning rate is 0.001, the maximum number of iterations is 1000 times, the training error threshold is 0.0001, and the number of samples for a single batch training is all the training data.

[0078] 4. The training process (forward propagation and backward propagation) includes:

[0079] Forward propagation: The input layer receives 5-dimensional data, and after weighted summation, it is input into the hidden layer: After being activated by Sigmoid, the output of the hidden layer is a h = Sigmoid(z h ).

[0080] The output layer receives the output of the hidden layer and obtains the prediction value through linear weighting:

[0081] Error calculation: Calculate the mean square error (MSE) between the prediction value and the true weight: where N is the number of training samples (160).

[0082] Backward propagation: Output layer gradient: The gradient of the error with respect to the output layer weights The bias gradient is

[0083] Hidden layer gradient: Calculate the gradient from the input layer to the hidden layer through the chain rule, considering the Sigmoid derivative Sigmoid -1 = a h ·(1 - a h ), Update formula:

[0084] Parameter update: Adopt the gradient descent method, and adjust the weights and biases according to the learning rate: where η = 0.001.

[0085] 5. Termination condition

[0086] Error convergence: If the training error E ≤ 0.0001, terminate the training in advance when the condition is reached at 600 times. (Maximum iteration: Force termination after reaching 1000 iterations to prevent overfitting.)

[0087] 6. Verification and model evaluation are as follows:

[0088] Verification set test: Input 40 groups of verification data into the trained network, calculate the MSE of the predicted value and the true value, and evaluate the generalization ability of the model. The following parameters are obtained

[0089]

[0090] Performance analysis: Compare the errors of the training set and the verification set, indicating that the model fits well. The system has high precision, fast response speed and wide fault tolerance mechanism.

[0091] The following conclusions can be drawn:

[0092] (1) After training, obtain the function mapping relationship between the sensor output voltage and the weight, and approximate and fit this function relationship using multiple linear functions. The approximation condition is based on the angular relationship between the angles of adjacent linear functions and the tangents of the linear functions. The specific method is to establish an optimization objective function: where θ j is the turning angle between adjacent segments, and the regularization coefficient λ = 0.01 controls the balance between smoothness and fitting accuracy.

[0093] (2) Piecewise slope constraint iteration: Apply |m j - m j+1 | / |1 + m j+1 - m j | ≤ tanθ j 2 to the slopes m max of adjacent segments, and solve the constrained least squares problem through the Lagrange multiplier method to ensure that the turning angle θ < 5° to obtain the fitting function.

[0094] The selection of the piecewise function needs to be achieved according to the following method. Dominant variable selection: Analyze the sensitivity of each sensor to the output, and select the left rear sensor as the basis for segmentation.

[0095] Segmentation point optimization: Based on the voltage range of the left rear sensor, use dynamic programming or greedy algorithm to divide the interval to ensure that the slope change between adjacent segments satisfies the included angle constraint condition:

[0096]

[0097] where m1 and m2 are the slopes of adjacent segments, and Δf' = m2 - m1.

[0098] Piecewise fitting and constraint verification Re - perform multiple linear regression on the data of each sub - interval to update the intercept and coefficients.

[0099] By iteratively adjusting the segmentation point, ensure that the derivative change between adjacent segments satisfies the included angle condition to avoid mutations.

[0100] Obtain a functional relationship similar to the following:

[0101] y = 71.827+0.514×left front + 1.781×left rear + 0.452×right rear - 0.025×right front.

[0102] Then write the multi - segment function into the single - chip microcomputer. When different voltages are collected, the function selects the mapping function of different weights. The single - chip microcomputer uses the binary search method for searching: achieving interval positioning with a time complexity of O(logN) for the ordered segmented interval. Hardware acceleration: Use the DMA or FPU unit of the single - chip microcomputer to accelerate floating - point operations, and finally obtain the actual weight. Through dominant variable segmentation, included angle constraint control, and optimization during the operation of the single - chip microcomputer, high - precision and low - latency calculation of the sensor voltage - to - weight mapping is achieved. This method significantly reduces the computational complexity while ensuring the fitting accuracy.

[0103] Embodiment 3

[0104] The parts not detailed in this embodiment are as shown in Embodiment 1, and it can achieve high - precision load measurement under the vehicle's motion state, and is applicable to various vehicle types and complex environmental conditions. As Figures 1-3 shown, this embodiment provides a weight measurement control method under the vehicle's motion state, including the following steps:

[0105] Step S1: Collect the original input data of the vehicle under different loads, speeds, and road conditions. The original input data includes weight data, motion state data, vehicle speed information, and driving environment data, and use an algorithm processor to process the original input data to compensate for dynamic effects;

[0106] Combine the data from the inclination sensor and the strain gauge sensor to compensate for the dynamic effects (such as bumps, uneven loads, etc.) during vehicle driving;

[0107] Specifically, as Figure 1 shown, the entire process from obtaining experimental data to determining the optimal non-linear model:

[0108] Obtain multiple sets of experimental data, and collect experimental data of different vehicles under different conditions.

[0109] Based on multiple sets of experimental data, use a neural network to train and obtain a non-linear model: Use the collected data to train the neural network and establish a model describing the non-linear relationship between vehicle load and sensor output.

[0110] Approximate and fit the non-linear model with a multi-terminal linear function, and evaluate and determine the optimal non-linear model: Approximate and fit the trained non-linear model with a multi-terminal linear function, and determine the optimal model by comparing the evaluation results with the actual counterweight.

[0111] Step S2: Use a multi-segment function to fit the neural network model, and predict the vehicle load measurement value by training the neural network model, and output the vehicle load measurement value in real time.

[0112] Specifically, the fitting logic of the piecewise function:

[0113] Fit the trained neural network model with a multi-segment linear or polynomial function, and determine the optimal parameters of each segment function by minimizing the fitting error, so as to reduce the computational complexity while maintaining high accuracy.

[0114] The data of the weight sensor and the motion sensor are input into the data processing unit in real time, and the system selects the corresponding piecewise function for calculation according to the original input data, and quickly obtains the load result.

[0115] Automatically adjust the parameters of the piecewise function according to the vehicle type and driving environment to ensure applicability under different conditions.

[0116] As Figure 2 shown, it shows the process of developing and testing a vehicle load measurement system based on a BP neural network. Each step from data acquisition to final system installation and testing:

[0117] Acquisition signal output: Collect the original signal data from the vehicle's sensors for training the BP neural network model.

[0118] Preprocess the data set: Perform operations such as cleaning, conversion, and normalization on the collected original data to improve the data quality and the efficiency of subsequent processing.

[0119] Debugging the BP neural network algorithm: Divide the preprocessed dataset into a training set and a test set. Use the training set data to train the neural network model and the test set data to evaluate the prediction ability of the model. Adjust and optimize the parameters and structure of the BP (backpropagation) neural network to effectively learn from the data, and divide, train, and predict the dataset to obtain a weighing system model:

[0120] Optimizing the model: Based on the performance of the model on the test set, further adjust and optimize the model to improve its accuracy and generalization ability.

[0121] Adding a compensation function model: On the basis of the neural network model, add a compensation function model to compensate for possible measurement errors or environmental impacts, thereby improving the measurement accuracy.

[0122] Installing the system on a vehicle for actual testing: Install the developed system on a vehicle for testing in an actual environment. For the vehicle load measurement system based on the BP neural network, ensure that the system can provide accurate and reliable measurement results in actual applications.

[0123] This embodiment solves the problems of low accuracy and excessive delay of traditional neural network control in vehicle weighing systems. By fusing the data of inclination sensors and strain gauges, it effectively compensates for the dynamic effects during vehicle driving, realizes high-precision load measurement, is applicable to various vehicle types and complex environmental conditions, and through multi-sensor fusion and BP neural network algorithm processing, realizes accurate measurement of vehicle load under dynamic conditions and solves the problem of inaccurate measurement in the moving state. As Figure 3 shown, the data acquisition and processing process and how to use the BP neural network for vehicle weighing:

[0124] Data acquisition: Based on the collection of "multiple groups of experimental data" in step S1.

[0125] Vehicle load measurement system algorithm based on BP neural network: Use the BP neural network algorithm to process the collected data to predict the vehicle load.

[0126] Generating a function model: Generate a function model through neural network processing to predict the vehicle load according to the input sensor data.

[0127] Sensor data input and output of actual weight: Demonstrates how sensor data is input into the system and how the system outputs the predicted load value.

[0128] Embodiment 3

[0129] For the parts not described in detail in this embodiment, refer to the description in Embodiment 1. This embodiment provides an electronic device, including: a processor and a memory, where a computer program that can be called by the processor is stored in the memory;

[0130] By calling the computer program stored in the memory, the processor executes the above-mentioned weight measurement control method under the vehicle motion state.

[0131] This electronic device may vary greatly due to different configurations or performances, and can include one or more processors (Central Processing Units, CPU) and one or more memories. Among them, at least one computer program is stored in the memory, and the at least one computer program is loaded and executed by the processor to implement the weight measurement control method provided by each of the above method embodiments. This electronic device can also include other components for implementing the functions of the device. For example, this electronic device can also have components such as wired or wireless network interfaces and input / output interfaces for input / output. Details are not described in this application embodiment.

[0132] This embodiment provides a computer-readable storage medium, on which a rewritable computer program is stored;

[0133] When the computer program runs on a computer device, the computer device is enabled to execute the above-mentioned weight measurement control method under the vehicle motion state.

[0134] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including at least one computer program, and the at least one computer program can be executed by a processor to complete the weight measurement control method in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (Read-Only Memory, abbreviated as: ROM), a random access memory (Random Access Memory, abbreviated as: RAM), a compact disc read-only memory (Compact Disc Read-Only Memory, abbreviated as: CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0135] In an exemplary embodiment, a computer program product or a computer program is further provided. The computer program product or the computer program includes one or more program codes, and the one or more program codes are stored in a computer-readable storage medium. One or more processors of an electronic device can read the one or more program codes from the computer-readable storage medium, and the one or more processors execute the one or more program codes, so that the electronic device can execute the above-mentioned method for weight measurement control in a vehicle motion state.

[0136] It should be understood that in various embodiments of the present application, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0137] It should be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0138] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A weight measurement control method in a vehicle's motion state, characterized in that, It includes the following steps: Step S1: Collect the original input data of the vehicle, where the original input data includes weight data, motion state data, vehicle speed information, and driving environment data; Step S2: Use an algorithm processor to process the original input data for signal processing and extract the compensation for dynamic effects; Step S3: Use a multi-segment function to fit a neural network model, and predict the vehicle load measurement value by training the neural network model, and output the vehicle load measurement value in real time.

2. The weight measurement control method in a vehicle's moving state according to claim 1, wherein: The acquisition logic of the compensation for dynamic effects: Capture the original input data of the vehicle in different load states based on the original input data, analyze the vehicle motion state data, and identify and quantify the analysis results of the dynamic effects suffered by the vehicle during driving; Establish a compensation model. According to the analysis results, establish one or more compensation models to simulate and predict the impact of dynamic effects on weight measurement; Model training and optimization. Use the collected data to train and optimize the compensation model, predict the impact of dynamic effects, adjust the model parameters, and during the vehicle driving process, apply the compensation model to compensate the weight measurement data in real time.

3. A weight measurement control method in a vehicle moving state according to claim 2, characterized in that: Automatically adjust the system parameters according to the vehicle type and driving environment. Determine the vehicle's driving environment data according to the vehicle's weight distribution and driving characteristics; Compare the actual measurement results with the expected results, adjust the model parameters according to the differences, and automatically adjust the model parameters to minimize the prediction error; Automatically adjust the parameters of the neural network model and the piecewise function to adapt to specific vehicles and environments.

4. A weight measurement control method in a vehicle moving state according to claim 3, characterized in that: It also has a self-calibration function. Set to execute a calibration program regularly to achieve the self-calibration function, and it can adapt to the drift of the sensor.

5. A weight measurement control method under vehicle motion state according to claim 4, characterized in that: According to the feedback data in actual applications, continuously optimize the parameters of the neural network model and the piecewise function. The training method of the neural network model is as follows: Data preprocessing; Filter, average, and standardize the original input data to eliminate the dimension difference. After standardization, the data is divided into a training set and a test set at a ratio of 8:2; Determine the network topology design corresponding to the neural network model. The original input data is the input layer, the non-linear function is the hidden layer, and the output layer corresponds to the weight prediction value; Initialize the parameters of the neural network model, and use forward propagation and backward propagation for training, so that the training terminates in advance when the conditions are met, or is forced to terminate after reaching the preset number of iterations to prevent overfitting.

6. A weight measurement system in a vehicle's motion state, which is realized based on the weight measurement control method in any one of claims 1-4, and is characterized in that: It includes a data acquisition module, a signal processor, an algorithm processing module, and a signal output terminal. Each module is connected through a communication module. Data acquisition module, which collects the original input data of the vehicle. The original input data includes weight data, motion state data, vehicle speed information, and driving environment data; Signal processor, which performs signal processing on the original input data to obtain the compensation for dynamic effects of the vehicle motion state on the weight data; Algorithm processing module, which uses a multi-segment function to fit a neural network model, and fits and predicts the vehicle load measurement value with the trained neural network model for the original input data.

7. A weight measurement system in a vehicle's motion state according to claim 6, characterized in that: It also includes a parameter resource library. Use a multi-segment function to fit a neural network model, and determine the optimal parameters of each segment function by minimizing the fitting error.

8. A weight measurement system in a vehicle's motion state according to claim 7, characterized in that: It also includes a signal output terminal, which outputs the calculated vehicle load measurement result to the in-vehicle display screen or the remote monitoring platform.

Images