Method for calibrating dynamic weighing precision of heavy truck

Abstract

The invention discloses a method for calibrating the dynamic weighing precision of a heavy truck, relates to the technical field of dynamic weighing, and solves the technical problems that an existing dynamic weighing algorithm is poor in calculation precision and reliability, and a large error is generated due to the fact that a measurement result is influenced by a driving state when the truck is under different driving conditions. Comprising the steps of periodically obtaining dynamic weighing data of a heavy truck as a sample set, and performing model training on a GRU neural network to obtain a dynamic weighing calibration model M; performing weighted fusion processing on the data of the plurality of load sensors to obtain measurement weighing information of the heavy truck; then various driving parameter values during measurement are substituted into the dynamic weighing calibration model M to calibrate measured weighing information so as to improve the dynamic weighing precision; the communication state between each load sensor and the controller is verified in real time through the vehicle-mounted terminal, and an administrator is prompted to overhaul or replace the load sensors in time, so that the dynamic weighing precision is improved.

Description

Technical Field

[0001] The invention relates to the technical field of dynamic weighing, in particular to a method for calibrating the dynamic weighing accuracy of a heavy truck. Background Art

[0002] Vehicle dynamic weighing is to weigh the vehicle without stopping or slowing down. Compared with the traditional static weighing measurement method, vehicle dynamic weighing is not only fast, efficient, and has little impact on highway throughput, but also can obtain relevant information such as the number of axles, axle weight, and vehicle speed while obtaining vehicle weight;

[0003] The traditional dynamic weighing algorithm only relies on the signal generated by a single sensor to denoise and process to obtain weighing data, and usually obtains the weighing data after correction based on a correction coefficient given to the entire lane; although this method can calculate the weighing data, the calculation accuracy and reliability are poor due to the inconsistency of the sensor itself, road conditions, polishing conditions, etc.; at the same time, under different driving conditions of the vehicle, the measurement results will be affected by the driving state and produce large errors; based on the above shortcomings, the present invention proposes a method for calibrating the dynamic weighing accuracy of heavy trucks. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a method for calibrating the dynamic weighing accuracy of a heavy truck.

[0005] To achieve the above object, a first aspect of the present invention provides a method for calibrating the dynamic weighing accuracy of a heavy truck, comprising the following steps:

[0006] Step 1: Periodically obtain the dynamic weighing data of heavy trucks as a sample set, perform model training on the GRU neural network, and obtain the dynamic weighing calibration model M; the dynamic weighing data includes various driving parameter values ​​of the heavy truck during measurement, measured weighing information, and target weighing information;

[0007] Step 2: Using distributed on-board sensors deployed at various locations on the heavy-duty truck, the driving parameter values ​​of the heavy-duty truck are collected in real time; the driving parameter values ​​include the rotation speed of the driving wheel hub, the driving speed, and the acceleration change rate in each direction;

[0008] Step 3: Use multiple load sensors installed on the axles to measure the load on each axle of the heavy truck in real time, that is, weight data; perform weighted fusion processing on the weight data to obtain the measured weighing information of the heavy truck;

[0009] Step 4: Input the various driving parameter values ​​during measurement into the dynamic weighing calibration model M to obtain the weighing calibration coefficient; calibrate the measured weighing information according to the weighing calibration coefficient to obtain the calibrated weighing information; then transmit the calibrated weighing information to the display module for real-time display;

[0010] Step 5: During the driving of the heavy truck, the communication status between each load sensor and the controller is verified in real time through the vehicle terminal; if the communication problem of the load sensor is serious, a communication abnormality signal is generated;

[0011] The vehicle-mounted terminal is used to send a communication abnormality signal to the controller. After receiving the communication abnormality signal, the controller controls the alarm module to sound an alarm and displays the location information of the corresponding load sensor on the display module to prompt the administrator to inspect or replace the load sensor.

[0012] Furthermore, the specific training steps of the dynamic weighing calibration model M are as follows:

[0013] Dividing the sample set into a training set, a test set, and a verification set according to a set ratio; the set ratio includes 2:1:1, 3:1:1, and 4:3:1;

[0014] Modify the state units in the GRU network structure so that each output state unit first enters the activation function before performing the next step of calculation, thereby constructing an improved GRU neural network;

[0015] The training set, validation set, and test set are input into the GRU neural network as historical feature values ​​for model training to obtain the dynamic weighing calibration model M.

[0016] Furthermore, the improved GRU neural network makes the state first enter an activation function during the calculation gate process, and also needs to pass through an activation function at the output end to complete the final output; the activation function selects the SELU function so that the excitation value of the neuron automatically converges to zero mean and unit variance.

[0017] Furthermore, the specific verification steps of the vehicle terminal are:

[0018] The vehicle-mounted terminal sends a verification configuration message to each load sensor according to a preset verification cycle; wherein the verification configuration message carries a first signal quality threshold;

[0019] After receiving the verification configuration message, the load sensor immediately sends a second synchronization signal to the controller; after monitoring the second synchronization signal, the controller determines the signal quality of the second synchronization signal and compares it with the first signal quality threshold to obtain a corresponding quality difference value Z1;

[0020] Compare the quality difference Z1 with the set value; if the quality difference Z1 is greater than the set value, it means that the signal is severely attenuated and a low-quality signal is generated;

[0021] Within a preset time period, if the number of occurrences of low-quality signals reaches a preset first number or a predetermined proportion or the number of occurrences of continuous low-quality signals reaches a preset second number, it is determined that the load sensor communication problem is serious and a communication abnormality signal is generated; wherein the preset second number is less than the preset second number.

[0022] Furthermore, the specific steps of weighted fusion processing include:

[0023] The weighted fusion algorithm is used to fuse the data of multiple load sensors to obtain the weight coefficient;

[0024] By determining the position of each axle of the vehicle, the corresponding sensor value is multiplied by the weight coefficient and added together to obtain the vehicle's measurement and weighing information.

[0025] Furthermore, when acquiring a sample set, a measurement error threshold is set. When the measurement error between the measured weighing information and the target weighing information is greater than or equal to the measurement error threshold, the corresponding dynamic weighing data is not used as a reference; the target weighing information is the weighing result obtained when the heavy truck is in a stationary state.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] The present invention periodically obtains the dynamic weighing data of heavy trucks as a sample set, performs model training on the GRU neural network, and obtains a dynamic weighing calibration model M; obtains the measured weighing information of the heavy truck by weighted fusion processing on the data of multiple load sensors; then substitutes the various driving parameter values ​​during measurement into the dynamic weighing calibration model M to obtain the weighing calibration coefficient; calibrates the measured weighing information according to the weighing calibration coefficient to improve the dynamic weighing accuracy; at the same time, during the driving of the heavy truck, the communication status between each load sensor and the controller is verified in real time through the on-board terminal. If the communication problem of the load sensor is serious, the administrator is promptly prompted to repair or replace the load sensor, thereby improving the dynamic weighing accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1The present invention is a principle block diagram of a method for calibrating dynamic weighing accuracy of a heavy truck. DETAILED DESCRIPTION

[0030] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] See also Figure 1 The first aspect of the present invention provides a method for calibrating the dynamic weighing accuracy of a heavy truck, comprising the following steps:

[0032] Step 1: Periodically obtain the dynamic weighing data of heavy trucks as a sample set, train the GRU neural network model, and obtain the dynamic weighing calibration model M;

[0033] Dynamic weighing data includes various driving parameter values ​​of heavy trucks during measurement, measured weighing information and target weighing information; the target weighing information is a relatively accurate weighing result obtained when the heavy truck is in a stationary state;

[0034] In this embodiment, not all dynamic weighing data acquired within the cycle time are qualified. For this reason, a measurement error threshold is set in this application. When the measurement error between the measured weighing information and the target weighing information is greater than or equal to the measurement error threshold, the corresponding dynamic weighing data is not used as a reference;

[0035] It should be noted that the specific training steps of the dynamic weighing calibration model M are as follows:

[0036] The sample set is divided into a training set, a test set, and a validation set according to a set ratio; the set ratios include 2:1:1, 3:1:1, and 4:3:1;

[0037] Modify the state units in the GRU network structure so that each output state unit first enters the activation function before performing the next step of calculation, thereby constructing an improved GRU neural network;

[0038] Among them, the improved GRU neural network makes the state enter an activation function first in the process of calculating the gate, and also needs to pass an activation function at the output end to complete the final output; the activation function selects the SELU function so that the excitation value of the neuron automatically converges to zero mean and unit variance;

[0039] The training set, validation set and test set are input into the GRU neural network as historical feature values ​​for model training to obtain a dynamic weighing calibration model M;

[0040] Step 2: The vehicle-mounted sensors distributed in various locations of the heavy-duty truck are used to collect the driving parameter values ​​of the heavy-duty truck in real time; the driving parameter values ​​include the rotation speed value of the driving wheel hub, the driving speed, and the acceleration change rate in all directions; it can be understood that the vehicle-mounted sensors include rotation speed sensors, speed sensors, and acceleration sensors;

[0041] Step 3: Use multiple load sensors installed on the axles to measure the load on each axle of the heavy truck in real time, that is, weight data; perform weighted fusion processing on the weight data to obtain the measured weighing information of the heavy truck; wherein the weighted fusion processing specifically includes:

[0042] The weighted fusion algorithm is used to fuse the data of multiple load sensors to obtain the weight coefficient;

[0043] By determining the position of each axle of the vehicle, the corresponding sensor value is multiplied by the weight coefficient and added together to obtain the vehicle's measurement and weighing information;

[0044] Step 4: Calibrate the measured weighing information according to the dynamic weighing calibration model M to obtain the calibrated weighing information, and then transmit the calibrated weighing information to the display module for real-time display; specifically:

[0045] Input the various driving parameter values ​​during measurement into the dynamic weighing calibration model M to obtain the weighing calibration coefficient;

[0046] The measured weighing information is calibrated according to the weighing calibration coefficient to obtain calibrated weighing information;

[0047] In this embodiment, since the controller processes the measured weight information of the heavy truck according to the measured data of the load sensor, it is very important to ensure high reliability and low latency communication between the load sensor and the controller so that the measured data can be quickly fed back to the controller;

[0048] Step 5: During the driving of the heavy truck, the communication status between each load sensor and the controller is verified in real time through the vehicle terminal. The specific verification steps are as follows:

[0049] The vehicle-mounted terminal sends a verification configuration message to each load sensor according to a preset verification cycle; wherein the verification configuration message carries a first signal quality threshold;

[0050] After receiving the verification configuration message, the load sensor immediately sends a second synchronization signal to the controller; after monitoring the second synchronization signal, the controller determines the signal quality of the second synchronization signal and compares it with the first signal quality threshold to obtain a corresponding quality difference value Z1;

[0051] The quality difference here can reflect the attenuation of the signal during the transmission process; compare the quality difference Z1 with the set value; if the quality difference Z1 is greater than the set value, it means that the signal is severely attenuated and a low-quality signal is generated;

[0052] Within a preset time period, if the number of occurrences of low-quality signals reaches a preset first number or a predetermined proportion or the number of occurrences of consecutive low-quality signals reaches a preset second number, it is determined that the communication problem of the load sensor is serious and a communication abnormality signal is generated; wherein the preset second number is less than the preset second number;

[0053] The vehicle-mounted terminal is used to send abnormal communication signals to the controller. After receiving the abnormal communication signals, the controller controls the alarm module to sound an alarm and displays the location information of the corresponding load sensor on the display module to prompt the administrator to inspect or replace the load sensor to improve the dynamic weighing accuracy.

[0054] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0055] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for calibrating the dynamic weighing accuracy of heavy trucks, It is characterized in that include: Step 1: Periodically obtain the dynamic weighing data of heavy trucks as a sample set, perform model training on the GRU neural network, and obtain the dynamic weighing calibration model M; the dynamic weighing data includes various driving parameter values ​​of the heavy truck during measurement, measured weighing information, and target weighing information; Step 2: Using distributed on-board sensors deployed at various locations on the heavy-duty truck, the driving parameter values ​​of the heavy-duty truck are collected in real time; the driving parameter values ​​include the rotation speed of the driving wheel hub, the driving speed, and the acceleration change rate in each direction; Step 3: Use multiple load sensors installed on the axles to measure the load on each axle of the heavy truck in real time, that is, weight data; perform weighted fusion processing on the weight data to obtain the measured weighing information of the heavy truck; Step 4: Input the various driving parameter values ​​during measurement into the dynamic weighing calibration model M to obtain the weighing calibration coefficient; calibrate the measured weighing information according to the weighing calibration coefficient to obtain the calibrated weighing information; then transmit the calibrated weighing information to the display module for real-time display; Step 5: During the driving of the heavy truck, the communication status between each load sensor and the controller is verified in real time through the vehicle terminal; if the communication problem of the load sensor is serious, a communication abnormality signal is generated; The vehicle-mounted terminal is used to send a communication abnormality signal to the controller. After receiving the communication abnormality signal, the controller controls the alarm module to sound an alarm and displays the location information of the corresponding load sensor on the display module to prompt the administrator to inspect or replace the load sensor.

2. A method for calibrating dynamic weighing accuracy of a heavy truck according to claim 1, It is characterized in that The specific training steps of the dynamic weighing calibration model M are as follows: Dividing the sample set into a training set, a test set, and a verification set according to a set ratio; the set ratio includes 2:1:1, 3:1:1, and 4:3:1; Modify the state units in the GRU network structure so that each output state unit first enters the activation function before performing the next step of calculation, thereby constructing an improved GRU neural network; The training set, validation set, and test set are input into the GRU neural network as historical feature values ​​for model training to obtain the dynamic weighing calibration model M.

3. A method for calibrating dynamic weighing accuracy of a heavy truck according to claim 2, It is characterized in that in, The improved GRU neural network makes the state enter an activation function first in the process of calculating the gate, and also needs to pass through an activation function at the output end to complete the final output; the activation function selects the SELU function so that the excitation value of the neuron automatically converges to zero mean and unit variance.

4. A method for calibrating dynamic weighing accuracy of a heavy truck according to claim 1, It is characterized in that The specific verification steps of the vehicle terminal are as follows: The vehicle-mounted terminal sends a verification configuration message to each load sensor according to a preset verification cycle; wherein the verification configuration message carries a first signal quality threshold; After receiving the verification configuration message, the load sensor immediately sends a second synchronization signal to the controller; after monitoring the second synchronization signal, the controller determines the signal quality of the second synchronization signal and compares it with the first signal quality threshold to obtain a corresponding quality difference value Z1; Compare the quality difference Z1 with the set value; if the quality difference Z1 is greater than the set value, it means that the signal is severely attenuated and a low-quality signal is generated; Within a preset time period, if the number of occurrences of low-quality signals reaches a preset first number or a predetermined proportion or the number of occurrences of continuous low-quality signals reaches a preset second number, it is determined that the load sensor communication problem is serious and a communication abnormality signal is generated; wherein the preset second number is less than the preset second number.

5. A method for calibrating dynamic weighing accuracy of a heavy truck according to claim 1, It is characterized in that The specific steps of weighted fusion processing include: The data of multiple load sensors are fused using a weighted fusion algorithm to obtain a weight coefficient; By determining the position of each axle of the vehicle, the corresponding sensor value is multiplied by the weight coefficient and added together to obtain the vehicle's measurement and weighing information.

6. A method for calibrating the dynamic weighing accuracy of a heavy truck according to claim 1, It is characterized in that When acquiring the sample set, a measurement error threshold is set. When the measurement error between the measured weighing information and the target weighing information is greater than or equal to the measurement error threshold, the corresponding dynamic weighing data is not used as a reference; the target weighing information is the weighing result obtained when the heavy truck is in a stationary state.

Images