Abnormality self-diagnosis method for highway traffic overload control detection equipment

Abstract

The invention provides an abnormity self-diagnosis method for highway traffic overload control detection equipment, which comprises the following steps of: acquiring time domain and frequency domain characteristics of a sensor data sample in a laboratory as a source domain; collecting related data of the dynamic truck scale of the overload control station, and calculating corresponding time domain features and frequency domain features as target domains; fusing the time domain feature and the frequency domain feature to obtain a corresponding source domain sample and a label-free target domain sample; mapping to a feature space through kernel function processing and a migration component analysis method TCA to obtain a source domain sample feature and a target domain sample feature; training a preset WKNN classifier by using the marked features of the source domain and the target domain after TCA feature migration, and outputting a target domain pseudo tag; and the target domain pseudo-label sample and the source domain are combined to re-train the WKNN classifier, the trained WKNN classifier is obtained, the trained WKNN classifier is used to carry out abnormity diagnosis on the dynamic truck scale, and the method can realize real-time abnormity self-diagnosis of overload treatment equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of abnormality detection of overweight control site equipment, and in particular to an abnormality self-diagnosis method for overweight control detection equipment for highway traffic. Background Art

[0002] Overweight vehicle control devices are crucial for ensuring road safety, maintaining traffic order, and reducing traffic accidents. They can promptly detect overloaded vehicles, preventing them from damaging infrastructure such as bridges and roads, thereby reducing repair and restoration costs. As overweight vehicle control efforts continue to advance, an increasing number of off-site inspection sites are being established. These sites are equipped with a variety of detection equipment, such as lidar for profile detection, and weighing instruments, load cells, and coil vehicle inspection devices for overload detection. However, these devices are prone to malfunctioning when exposed to harsh environments or improperly installed or used, resulting in the inability to properly weigh passing vehicles. Therefore, regular calibration of measuring instruments such as dynamic truck scales and profile measurement devices used in road transportation is crucial for monitoring and enforcing traffic regulations, preventing illegal overloading, improving the effectiveness of traffic equipment, and ensuring the safety and reliability of the road network. However, since dynamic truck scales, contour measuring devices and other measuring instruments in road transportation are widely distributed and numerous, traditional detection methods require manual inspection of each one, which has low detection efficiency and increases the difficulty and complexity of the detection work. At present, some fuzzy algorithms are used to calculate abnormal situations, but these algorithms are not accurate and have low practicality during actual use.

[0003] Therefore, a more efficient and accurate automatic detection method for equipment failures at overweight control stations is needed to better ensure road safety, maintain traffic order, and reduce traffic accidents. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an abnormality self-diagnosis method for highway traffic overload control detection equipment to solve the problems of low efficiency and low detection accuracy of existing equipment abnormalities.

[0005] The present invention provides a method for self-diagnosis of abnormalities in highway traffic overload control detection equipment, the method comprising the following steps: Step S1: collecting the time domain and frequency domain features of the weighing sensor data samples under various fault types in the laboratory as the source domain; Step S2: Collect relevant data of dynamic truck scales at overweight control sites, calculate the output value of each weighing sensor through the association model, and calculate the time domain characteristics and frequency domain characteristics based on the output values of each weighing sensor as the target domain; Step S3: Fusing the time domain features and frequency domain features of the source domain and the target domain respectively to obtain processed source domain samples and unlabeled target domain samples; Step S4: Process the source domain samples and target domain samples respectively through the kernel function, and then map them to the feature space through the transfer component analysis method TCA to obtain the source domain sample features and the target domain sample features; Step S5: Use the labeled features of the source domain and target domain after TCA feature migration to train the preset WKNN classifier and output the target domain pseudo label; Step S6: Merge the pseudo-labeled samples of the target domain with the source domain to form an extended annotated dataset, and retrain the WKNN classifier using the extended dataset to obtain a trained WKNN classifier. Step S7: collect relevant data from any dynamic truck scale at the overweight vehicle control site in real time, pre-process it, and input it into the trained WKNN classifier to output the predicted fault type.

[0006] Furthermore, the fault types include substantial faults and performance degradation faults, and the substantial faults include weighing sensor faults, foundation and scale platform faults, and junction box and line faults.

[0007] Furthermore, the relevant data of the dynamic truck scale at the overweight vehicle control site includes environmental detection data, weighing instrument data, multiple sensor data used for weighing each dynamic truck scale, lidar data, camera and coil vehicle inspection device.

[0008] Furthermore, the step S2 is specifically as follows: Step S21: Collect the values of multiple weighing sensors corresponding to the dynamic truck scale at the overweight control station, and use a Butterworth filter to filter out dynamic signal interference; Step S22: Calculate the output value of each weighing sensor using a correlation model; specifically, the correlation model is: ; in, is the output value of the jth weighing sensor of the dynamic truck scale, is the value of the nth weighing sensor of the dynamic truck scale collected, M is the number of weighing sensors corresponding to the current dynamic truck scale, For the corresponding The coefficient between the load cells and the remaining M-1 load cells, , 、 and They are the vehicle speed factor, load type factor and load factor in dynamic conditions, is the correlation weight of these three influencing factors; Step S23: Calculate the time domain and frequency domain features based on the output values of each weighing sensor as the target domain. The time domain features include the root mean square value, pulse index, kurtosis index, margin index, peak index, average amplitude, and waveform index. The frequency domain features include standard deviation frequency, center of gravity frequency, root mean square frequency, and mean frequency. The calculation expressions of each feature are as follows: RMS value: ; Pulse indicator: ; Kurtosis index: ; Margin index: ; Peak index: ; Average amplitude: ; Waveform indicators: ; Standard Deviation Frequency: ; Center of gravity frequency: ; RMS frequency: ; Mean frequency: ; in, Indicates the amplitude of the i-th point of a sensor output signal, N indicates the number of signal points, represents the frequency value of the i-th line of the power spectrum, Indicates the reference output value.

[0009] Furthermore, before step S3, the method further includes: performing PCA or KPCA feature dimensionality reduction analysis on the source domain and the target domain respectively to obtain the variance explanation rate of each principal component, and then fusing the time domain features with the frequency domain features to obtain labeled weighing sensor samples and unlabeled truck scale samples respectively, wherein a reference model of the characteristic signals under different fault types is established to determine the threshold range under different faults.

[0010] Furthermore, in step S4, “mapping to the feature space through TCA feature migration to obtain source domain sample features and target domain sample features” is specifically as follows: TCA feature migration is used to map source and target domain samples to the reproducing kernel Hilbert space. By minimizing the maximum mean difference between source and target domain samples in the feature subspace, a dimensionality-reduced feature subspace is obtained. The distribution difference between the source and target domains is reduced in this subspace, and source domain sample features and target domain sample features with a certain degree of similarity are obtained, thereby realizing cross-domain migration of fault knowledge.

[0011] Furthermore, the preprocessing in step S7 includes using a Butterworth filter to filter out dynamic signal interference from the collected weighing sensors, obtaining the output values of each weighing sensor of the dynamic vehicle scale through a correlation model, and then calculating the time domain features and frequency domain features through the output values of each sensor, and then fusing the time domain features and frequency domain features through dimensionality reduction processing.

[0012] Furthermore, the method also includes using the collected relevant data of the dynamic vehicle scale to synchronously judge abnormal behavior, and further screening the output of step S7. If there is no abnormal behavior, it is judged as the corresponding fault type. Otherwise, a vehicle abnormal behavior warning is directly issued; wherein, the abnormal vehicle behavior includes: crossing lane behavior, wrong-way behavior, edge pressing behavior, S-shaped behavior, parking behavior, scale jumping behavior, scale dragging behavior, scale grinding behavior, queuing behavior and jack behavior. The judgment of these abnormal behaviors is achieved by deeply integrating the perception data of the roadside equipment, and combining the multi-channel data of the weighing system with optical measurement methods to judge the vehicle weight and behavior.

[0013] Furthermore, the specific process of judging abnormal behavior and the way to handle it include: The lane-crossing behavior is determined based on its characteristics, which include: the vehicle straddling the centerline of a lane, with part of its weight being taken from the adjacent lane, and the weight measured by the current lane's weighing equipment being inaccurate. The lane-crossing behavior is determined by estimating the vehicle's position and status based on multi-channel sensor signals from all weighing equipment within the station, determining its driving position and direction through coil signals, and comparing the vehicle's size and position measured by lidar with the simultaneous weighing signals to determine whether the vehicle is in a lane-crossing state. The lane-crossing behavior is processed as follows: the system indicates this abnormal state, and sensor signals matching the vehicle's gross weight based on position are accumulated and uploaded. The wrong-way behavior is judged based on the characteristics of the wrong-way behavior, which are: the vehicle crosses the center line of the lane, the entire weight is obtained by the weighing equipment of the adjacent opposite lane, there is no weighing data in the current lane, and an error occurs in the weighing process of the opposite lane; the wrong-way behavior is judged in the following way: the vehicle position and status are estimated based on the multi-channel sensor signals of all weighing equipment in the station, the driving position and direction are judged by the coil signal, and the vehicle size, position and driving direction obtained by the laser radar are compared with the weighing signal at the same time to determine whether the vehicle is in a wrong-way state; the wrong-way behavior is processed as follows: the abnormal state is prompted in the system, and the sensor signal of the opposite lane belonging to the total weight of the vehicle is matched according to the position and accumulated and uploaded, and the rear camera of the vehicle is called as the front camera to identify the vehicle license plate; The edge pressing behavior is judged based on its characteristics, which are: the weighing equipment does not cover the entire width of the lane, and the vehicle is partially pressed on the road surface without a weighing platform; the edge pressing behavior is judged by judging whether the balance of the left and right weights exceeds a threshold based on the multi-channel sensor signal of the weighing equipment, and comparing the vehicle size, position and driving direction measured by the lidar with the weighing signal at the same time to judge whether the vehicle is in an edge pressing state; the edge pressing behavior is processed as follows: the abnormal state is prompted in the system, and the left half of the vehicle weight belonging to the vehicle is matched according to the position as the reference half weight, and twice the weight is uploaded as temporary data for subsequent review; The S-shaped behavior is judged based on its characteristics, which include: the vehicle passing through the weighing area in a swaying route, so that the two wheels on the same axle of the vehicle do not pass through the weighing area in the lane at the same time; the S-shaped behavior is judged by determining the characteristics of the axle weight and wheel weight peaks based on the multi-channel sensor signals of all weighing equipment at the site, combining the vehicle's driving trajectory measured by lidar with the weighing signals of the same period to determine whether the vehicle is in an S-shaped driving state; the S-shaped behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when the vehicle passed through the weighing area are uploaded as temporary data for subsequent review; The parking behavior is judged based on the characteristics of the parking behavior, which are: parking in the weighing area for a long time, resulting in erroneous signals from the ground sensor coil vehicle detector, errors in the dynamic vehicle scale program for axle weight measurement, and inability to correctly calculate the matching total weight; the parking behavior is judged by estimating the vehicle state based on the multi-channel sensor signal of the weighing equipment, judging whether the vehicle is in a driving state through the coil signal, confirming the vehicle trajectory obtained by laser radar measurement, and comparing it with the weighing signal at the same time to judge whether the vehicle is in a parked state; the parking behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when all axles of the vehicle pass through the weighing area are uploaded as data; The weight-jump behavior is determined based on its characteristics, which include: the vehicle's acceleration causes it to lift its head, shifting its center of gravity rearward to reduce the measured front axle weight; the weight-jump behavior is determined by determining its acceleration behavior based on a vehicle speed-time curve measured by a laser radar; and the weight-jump behavior is processed as follows: the abnormal state is prompted in the system; The weighing behavior is determined based on its characteristics, which include: the vehicle is towed on the weighing device by locking the trailer axle, increasing the lateral force on the weighing sensor and causing the load carrier to become stuck, thereby reducing the measured trailer weight. The weighing behavior is determined by determining the acceleration and deceleration behavior based on the vehicle speed-time curve measured by the laser radar. The weighing behavior is processed as follows: the abnormal state is prompted in the system; The scale-wasting behavior is judged based on its characteristics, which include: the vehicle repeatedly moving forward and backward in the weighing area, resulting in errors in the accumulated axle count and total weight. The scale-wasting behavior is judged by estimating the vehicle's forward and backward state and the weight of the axles that are repeatedly weighed based on the weight information of all weighing devices in the station. The speed-time curve measured by the lidar is then compared with the weighing signal at the same time to determine the weight of the vehicle that has been repeatedly weighed. The scale-wasting behavior is processed as follows: the abnormal state is prompted in the system, and the repeatedly weighed weight and the repeatedly calculated axle count are eliminated. The queuing behavior is determined based on its characteristics: multiple vehicles queuing closely together, reducing the distance between vehicles, and intentionally interfering with the vehicle separator, causing weight errors and preventing the camera from recognizing license plates. The queuing behavior is determined by estimating the vehicle separation status based on the weight information of all weighing devices within the station, combined with the vehicle separation information obtained by lidar recognition. The queuing behavior is then processed by notifying the system of the abnormal status and matching the weights of the vehicles in the queue based on spatiotemporal information. The jack behavior is judged based on the characteristics of the jack behavior, which are: by installing a jack device at the bottom of the vehicle, the weight is transferred to the non-weighing area, thereby reducing the measured axle weight; the jack behavior is judged in the following way: based on the weight of each axle obtained by the weighing equipment, it is estimated whether the axle weight imbalance is caused by the jack, and at the same time, combined with the speed-time curve measured by the laser radar, it is judged whether the jack is operated to avoid inspection by stopping the vehicle; the jack behavior is processed as follows: the abnormal state is prompted in the system.

[0014] Furthermore, the method further includes step 8, after a fault is detected, using an intelligent fault-tolerant method for a dynamic weighing system based on an improved particle swarm optimization generalized regression neural network under a multi-correlation model to measure and estimate passing vehicles, as follows: Eliminate deviation data caused by failure; The Lévy flight-modified particle swarm algorithm is used to search for the optimal solution to optimize the generalized regression neural network, and the LPSO-GRNN network model matching the truck scale data samples is obtained, and the output estimation network is established based on this model. A multi-correlation model is used to determine the expected characteristic values under different faults, and then the output prediction network is used to obtain the expected output estimate of the faulty sensor, thereby realizing the measurement estimate of the passing vehicles.

[0015] The advantages of the present invention are that it can realize real-time abnormal self-diagnosis of overweight control equipment, improve the efficiency and accuracy of diagnosing equipment abnormalities, so as to timely maintain faulty equipment and ensure stable operation of the equipment, which is of great significance to ensuring road safety, maintaining traffic order and reducing traffic accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0017] Figure 1 This is a flowchart of an execution method of an abnormality self-diagnosis method of a highway traffic overload control detection device according to the present invention; Figure 2 This is a schematic diagram of the principle of an abnormality self-diagnosis method of highway traffic overload control detection equipment according to the present invention; Figure 3 This is a principle block diagram of the model building corresponding to the abnormal self-diagnosis method of highway traffic overload control detection equipment of the present invention. DETAILED DESCRIPTION

[0018] The embodiment of the present application provides an abnormality self-diagnosis method for highway traffic overload control detection equipment, which can realize automatic diagnosis and analysis of abnormal data and quickly locate the cause of the abnormality.

[0019] The technical solution in the embodiments of this application has the following general idea: a correlation model is used to calculate the sensor output values of a dynamic truck scale to calculate time and frequency domain features as the target domain. Simultaneously, sensor data under various types of faults simulated in the laboratory is used to obtain the source domain. To address the difficulty of online detection of dynamic faults in truck scales while in service, a dynamic weighing system fault detection method based on kernel principal component analysis (KPCA) under a multi-correlation model is proposed. A Butterworth filter is used to filter out dynamic signal interference. KPCA and PCA are used to achieve dimensionality reduction and screening of features. Feature indicators and corresponding control limits are determined to construct a comprehensive fault evaluation method. To address the problem of insufficient on-site truck scale fault data collection, which leads to reduced diagnostic performance, a cross-sensor fault diagnosis method based on transfer component analysis (TCA) combined with a weighted K-nearest neighbor algorithm (WKNN) under multi-feature fusion is adopted. TCA is used to map sensor samples and truck scale samples into a feature space, thereby reducing distribution differences. WKNN is used as a classifier to achieve data classification and matching. This method achieves a diagnostic accuracy of 93.33% when the support set sample size for each fault condition is 3, effectively avoiding the problem of unbalanced sample distribution. The present invention also diagnoses abnormal behavior in data anomalies, eliminating false fault outputs caused by abnormal behavior, thereby improving the accuracy of fault analysis. Furthermore, by estimating the expected output of the faulty sensor, it improves the accuracy of vehicle prediction under fault conditions.

[0020] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0021] like Figures 1 to 3 As shown, this embodiment provides a method for self-diagnosis of abnormalities in highway traffic overload control detection equipment, the method comprising the following steps: Step S1: collecting the time domain and frequency domain features of the weighing sensor data samples under various fault types in the laboratory as the source domain; Step S2: Collect relevant data of dynamic truck scales at overweight control sites, calculate the output value of each weighing sensor through the association model, and calculate the time domain characteristics and frequency domain characteristics based on the output values of each weighing sensor as the target domain; Step S3: Fusing the time domain features and frequency domain features of the source domain and the target domain respectively to obtain processed source domain samples and unlabeled target domain samples; Step S4: Process the source and target domain samples separately using kernel functions, then map them to the feature space using TCA feature transfer to obtain source and target domain sample features. Preferably, TCA feature transfer is used to map the source and target domain samples to a reproducing kernel Hilbert space. By minimizing the maximum mean difference between the source and target domain samples in the feature subspace, a reduced-dimensional feature subspace is obtained. The distribution difference between the source and target domains is reduced in this subspace, thereby enabling cross-domain transfer of fault knowledge.

[0022] Specifically, for vibrating weighing sensors, the vibration signals of the weighing sensors are first acquired. The vibration signals of the weighing sensors with known fault states are used as source domain samples, while the vibration signals of the dynamic weighing system with unknown fault states are used as target domain samples. The TCA algorithm is used to process the differences between the source and target domain samples, thereby increasing the similarity between the source and target domain data samples. The migrated source domain data is then identified using the maximum mean difference (MMD) method, addressing the issue of insufficient labeled data in the target domain, i.e., a limited number of actual fault samples. This results in a cross-device fault diagnosis model for dynamic weighing systems based on TCA and WKNN.

[0023] Step S5: Use the labeled features of the source and target domains after TCA feature migration to train the preset WKNN classifier and output the pseudo-label of the target domain. Through WKNN, the cross-domain transfer of fault knowledge is realized. After the original labeled source domain has gone through the previous series of operations, the WKNN algorithm can effectively capture the potential feature differences between the source and target domains, allowing part of the target domain to also be labeled, and semi-supervised learning is performed again. Finally, the unlabeled class with the largest prediction probability in the target domain is labeled with a pseudo-label, so that all target domains are labeled.

[0024] Step S6: Filter the target domain pseudo labels to obtain filtered target domain pseudo label samples, merge the filtered target domain pseudo label samples with the source domain to form an extended annotated dataset, and retrain the WKNN classifier using the extended dataset to obtain a trained WKNN classifier. Step S7: collect relevant data from any dynamic truck scale at the overweight vehicle control site in real time, pre-process it, and input it into the trained WKNN classifier to output the predicted fault type.

[0025] The above steps S1 and S2 are executed in no particular order.

[0026] Preferably, the fault types include substantial faults and performance degradation faults, and the substantial faults include weighing sensor faults, foundation and weighing platform faults, and junction box and line faults.

[0027] Preferably, the relevant data of the dynamic truck scale at the overweight vehicle control site includes environmental detection data, weighing instrument data, multiple sensor data used for weighing each dynamic truck scale, lidar data, camera and coil vehicle detector.

[0028] Preferably, the step S2 is specifically as follows: Step S21: Collect the values of multiple weighing sensors corresponding to the dynamic truck scale at the overweight control station, and use a Butterworth filter to filter out dynamic signal interference; Step S22: Calculate the output value of each weighing sensor using a correlation model; specifically, the correlation model is: ; The above formula is expanded to: ; in, is the output value of the jth weighing sensor of the dynamic truck scale, is the value of the nth weighing sensor of the dynamic truck scale collected, M is the number of weighing sensors corresponding to the current dynamic truck scale, For the corresponding The coefficient between the load cells and the remaining M-1 load cells, , 、 and They are the vehicle speed factor, load type factor and load factor in dynamic conditions, is the correlation weight of these three influencing factors; Step S23: Calculate the time domain and frequency domain features based on the output values of each weighing sensor as the target domain. The time domain features include the root mean square value, pulse index, kurtosis index, margin index, peak index, average amplitude, and waveform index. The frequency domain features include standard deviation frequency, center of gravity frequency, root mean square frequency, and mean frequency. The calculation expressions of each feature are as follows: RMS value: ; Pulse indicator: ; Kurtosis index: ; Margin index: ; Peak index: ; Average amplitude: ; Waveform indicators: ; Standard Deviation Frequency: ; Center of gravity frequency: ; RMS frequency: ; Mean frequency: ; in, Indicates the amplitude of the i-th point of a sensor output signal, N indicates the number of signal points, represents the frequency value of the i-th line of the power spectrum, Indicates the reference output value.

[0029] Preferably, before step S3, the method further includes: performing PCA or KPCA feature dimensionality reduction analysis on the source domain and the target domain respectively to obtain the variance explanation rate of each principal component, and then fusing the time domain features with the frequency domain features to obtain labeled weighing sensor samples and unlabeled truck scale samples respectively, wherein a reference model of the characteristic signals under different fault types is established to determine the threshold range under different faults.

[0030] Through the above steps S1-S3, the time domain and frequency domain features of the source domain and target domain samples are extracted, and they are fused through the PCA feature fusion technology to obtain the labeled source domain sample feature set and the unlabeled target domain sample feature set, thus realizing feature extraction and fusion.

[0031] Preferably, the preprocessing in step S7 includes using a Butterworth filter to filter out dynamic signal interference from the collected weighing sensors, obtaining the output values of each weighing sensor of the dynamic vehicle scale through a correlation model, and then calculating the time domain characteristics and frequency domain characteristics through the output values of each sensor, and then fusing the time domain characteristics and frequency domain characteristics through dimensionality reduction processing.

[0032] Preferably, the method further includes using the collected relevant data of the dynamic vehicle scale to synchronously judge abnormal behavior, and further screening the output of step S7. If there is no abnormal behavior, it is judged as the corresponding fault type. Otherwise, a vehicle abnormal behavior warning is directly issued; wherein, the abnormal vehicle behavior includes: crossing lane behavior, wrong-way behavior, edge pressing behavior, S-shaped behavior, parking behavior, scale jumping behavior, scale dragging behavior, scale grinding behavior, queuing behavior and jack behavior. The judgment of these abnormal behaviors is achieved by deeply integrating the perception data of the roadside equipment, and combining the multi-channel data of the weighing system with optical measurement methods to judge the vehicle weight and behavior.

[0033] Preferably, the specific process of judging abnormal behavior and the method of handling it include: The lane-crossing behavior is determined based on its characteristics, which include: the vehicle straddling the centerline of a lane, with part of its weight being taken from the adjacent lane, and the weight measured by the current lane's weighing equipment being inaccurate. The lane-crossing behavior is determined by estimating the vehicle's position and status based on multi-channel sensor signals from all weighing equipment within the station, determining its driving position and direction through coil signals, and comparing the vehicle's size and position measured by lidar with the simultaneous weighing signals to determine whether the vehicle is in a lane-crossing state. The lane-crossing behavior is processed as follows: the system indicates this abnormal state, and sensor signals matching the vehicle's gross weight based on position are accumulated and uploaded. The wrong-way behavior is judged based on the characteristics of the wrong-way behavior, which are: the vehicle crosses the center line of the lane, the entire weight is obtained by the weighing equipment of the adjacent opposite lane, there is no weighing data in the current lane, and an error occurs in the weighing process of the opposite lane; the wrong-way behavior is judged in the following way: the vehicle position and status are estimated based on the multi-channel sensor signals of all weighing equipment in the station, the driving position and direction are judged by the coil signal, and the vehicle size, position and driving direction obtained by the laser radar are compared with the weighing signal at the same time to determine whether the vehicle is in a wrong-way state; the wrong-way behavior is processed as follows: the abnormal state is prompted in the system, and the sensor signal of the opposite lane belonging to the total weight of the vehicle is matched according to the position and accumulated and uploaded, and the rear camera of the vehicle is called as the front camera to identify the vehicle license plate; The edge pressing behavior is judged based on its characteristics, which are: the weighing equipment does not cover the entire width of the lane, and the vehicle is partially pressed on the road surface without a weighing platform; the edge pressing behavior is judged by judging whether the balance of the left and right weights exceeds a threshold based on the multi-channel sensor signal of the weighing equipment, and comparing the vehicle size, position and driving direction measured by the lidar with the weighing signal at the same time to judge whether the vehicle is in an edge pressing state; the edge pressing behavior is processed as follows: the abnormal state is prompted in the system, and the left half of the vehicle weight belonging to the vehicle is matched according to the position as the reference half weight, and twice the weight is uploaded as temporary data for subsequent review; The S-shaped behavior is judged based on its characteristics, which include: the vehicle passing through the weighing area in a swaying route, so that the two wheels on the same axle of the vehicle do not pass through the weighing area in the lane at the same time; the S-shaped behavior is judged by determining the characteristics of the axle weight and wheel weight peaks based on the multi-channel sensor signals of all weighing equipment at the site, combining the vehicle's driving trajectory measured by lidar with the weighing signals of the same period to determine whether the vehicle is in an S-shaped driving state; the S-shaped behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when the vehicle passed through the weighing area are uploaded as temporary data for subsequent review; The parking behavior is judged based on the characteristics of the parking behavior, which are: parking in the weighing area for a long time, resulting in erroneous signals from the ground sensor coil vehicle detector, errors in the dynamic vehicle scale program for axle weight measurement, and inability to correctly calculate the matching total weight; the parking behavior is judged by estimating the vehicle state based on the multi-channel sensor signal of the weighing equipment, judging whether the vehicle is in a driving state through the coil signal, confirming the vehicle trajectory obtained by laser radar measurement, and comparing it with the weighing signal at the same time to judge whether the vehicle is in a parked state; the parking behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when all axles of the vehicle pass through the weighing area are uploaded as data; The weight-jump behavior is determined based on its characteristics, which include: the vehicle's acceleration causes it to lift its head, shifting its center of gravity rearward to reduce the measured front axle weight; the weight-jump behavior is determined by determining its acceleration behavior based on a vehicle speed-time curve measured by a laser radar; and the weight-jump behavior is processed as follows: the abnormal state is prompted in the system; The weighing behavior is determined based on its characteristics, which include: the vehicle is towed on the weighing device by locking the trailer axle, increasing the lateral force on the weighing sensor and causing the load carrier to become stuck, thereby reducing the measured trailer weight. The weighing behavior is determined by determining the acceleration and deceleration behavior based on the vehicle speed-time curve measured by the laser radar. The weighing behavior is processed as follows: the abnormal state is prompted in the system; The scale-wasting behavior is judged based on its characteristics, which include: the vehicle repeatedly moving forward and backward in the weighing area, resulting in errors in the accumulated axle count and total weight. The scale-wasting behavior is judged by estimating the vehicle's forward and backward state and the weight of the axles that are repeatedly weighed based on the weight information of all weighing devices in the station. The speed-time curve measured by the lidar is then compared with the weighing signal at the same time to determine the weight of the vehicle that has been repeatedly weighed. The scale-wasting behavior is processed as follows: the abnormal state is prompted in the system, and the repeatedly weighed weight and the repeatedly calculated axle count are eliminated. The queuing behavior is determined based on its characteristics: multiple vehicles queuing closely together, reducing the distance between vehicles, and intentionally interfering with the vehicle separator, causing weight errors and preventing the camera from recognizing license plates. The queuing behavior is determined by estimating the vehicle separation status based on the weight information of all weighing devices within the station, combined with the vehicle separation information obtained by lidar recognition. The queuing behavior is then processed by notifying the system of the abnormal status and matching the weights of the vehicles in the queue based on spatiotemporal information. The jack behavior is judged based on the characteristics of the jack behavior, which are: by installing a jack device at the bottom of the vehicle, the weight is transferred to the non-weighing area, thereby reducing the measured axle weight; the jack behavior is judged in the following way: based on the weight of each axle obtained by the weighing equipment, it is estimated whether the axle weight imbalance is caused by the jack, and at the same time, combined with the speed-time curve measured by the laser radar, it is judged whether the jack is operated to avoid inspection by stopping the vehicle; the jack behavior is processed as follows: the abnormal state is prompted in the system.

[0034] Preferably, the method further includes step 8, after detecting a fault, using an intelligent fault-tolerant method for a dynamic weighing system based on an improved particle swarm optimization generalized regression neural network under a multi-correlation model to measure and estimate passing vehicles, specifically as follows: Eliminate deviation data caused by failure; The Levy flight improved particle swarm algorithm (LPSO) was used to search for the optimal solution to optimize the generalized regression neural network (GRNN). The LPSO-GRNN network model that matched the truck scale data samples was obtained, and the output prediction network was established based on this model. A multi-correlation model is used to determine the expected characteristic values under different fault conditions. The output prediction network is then used to obtain the expected output value of the faulty sensor, thereby achieving an estimated measurement value for the vehicle. The expected characteristic values refer to the signal's mean frequency, standard deviation frequency, pulse index, and root mean square frequency, which are characteristic values expected due to a sensor fault. The expected output value is the measurement value that the sensor would output if it were fault-free under the current load.

[0035] The technical solution provided in the embodiments of the present application has at least the following technical effects or advantages: the present invention uses the TCA algorithm to process the differences between source domain samples and target domain samples, thereby increasing the similarity between source domain data samples and target domain samples, and discriminates the migrated source domain data through the maximum mean difference, and also marks the target domain features, thereby solving the problem of insufficient supply of labeled data in the target domain, that is, the small number of actual fault samples. The distribution difference between the source domain and the target domain is reduced in this subspace, thereby realizing the cross-domain migration of fault knowledge, and forming a dynamic weighing system cross-device fault diagnosis model based on TCA-WKNN. At the same time, combined with the abnormal behavior analysis algorithm, it can realize automatic diagnosis and analysis of abnormal conditions of equipment at overweight control sites, while improving diagnostic accuracy, improving the efficiency of equipment fault self-diagnosis, quickly locating fault points and fault causes, realizing efficient maintenance, improving equipment operation stability, and improving the work efficiency of overweight control site equipment.

[0036] Although the specific embodiments of the present invention are described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and are not intended to limit the scope of the present invention. Equivalent modifications and changes made by those skilled in the art in accordance with the spirit of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for self-diagnosis of abnormalities in highway traffic overload control detection equipment, characterized by: The method comprises the following steps: Step S1: collecting the time domain and frequency domain features of the weighing sensor data samples under various fault types in the laboratory as the source domain; Step S2: Collect relevant data of dynamic truck scales at overweight control sites, calculate the output value of each weighing sensor through the association model, and calculate the time domain characteristics and frequency domain characteristics based on the output values of each weighing sensor as the target domain; Step S3: Fusing the time domain features and frequency domain features of the source domain and the target domain respectively to obtain processed source domain samples and unlabeled target domain samples; Step S4: Process the source domain samples and target domain samples respectively through the kernel function, and then map them to the feature space through the transfer component analysis method TCA to obtain the source domain sample features and the target domain sample features; Step S5: Use the labeled features of the source domain and target domain after TCA feature migration to train the preset WKNN classifier and output the target domain pseudo label; Step S6: Merge the pseudo-labeled samples of the target domain with the source domain to form an extended annotated dataset, and retrain the WKNN classifier using the extended dataset to obtain a trained WKNN classifier. Step S7: collect relevant data from any dynamic truck scale at the overweight vehicle control site in real time, pre-process it, and input it into the trained WKNN classifier to output the predicted fault type.

2. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: The fault types include substantial faults and performance degradation faults. The substantial faults include weighing sensor faults, foundation and scale platform faults, and junction box and line faults.

3. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: The relevant data of the dynamic vehicle scale at the overweight vehicle control site includes environmental detection data, weighing instrument data, multiple sensor data used for weighing each dynamic vehicle scale, lidar data, camera and coil vehicle inspection device.

4. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: The step S2 is specifically as follows: Step S21: Collect the values of multiple weighing sensors corresponding to the dynamic truck scale at the overweight control station, and use a Butterworth filter to filter out dynamic signal interference; Step S22: Calculate the output value of each weighing sensor using a correlation model; specifically, the correlation model is: ; in, is the output value of the jth weighing sensor of the dynamic truck scale, is the value of the nth weighing sensor of the dynamic truck scale collected, M is the number of weighing sensors corresponding to the current dynamic truck scale, For the corresponding The coefficient between the load cells and the remaining M-1 load cells, , 、 and They are the vehicle speed factor, load type factor and load factor in dynamic conditions, is the correlation weight of these three influencing factors; Step S23: Calculate the time domain and frequency domain features based on the output values of each weighing sensor as the target domain. The time domain features include the root mean square value, pulse index, kurtosis index, margin index, peak index, average amplitude, and waveform index. The frequency domain features include standard deviation frequency, center of gravity frequency, root mean square frequency, and mean frequency. The calculation expressions of each feature are as follows: RMS value: ; Pulse indicator: ; Kurtosis index: ; Margin index: ; Peak index: ; Average amplitude: ; Waveform indicators: ; Standard Deviation Frequency: ; Center of gravity frequency: ; RMS frequency: ; Mean frequency: ; in, Indicates the amplitude of the i-th point of a sensor output signal, N indicates the number of signal points, represents the frequency value of the i-th line of the power spectrum, Indicates the reference output value.

5. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: Before step S3, the method also includes: performing PCA or KPCA feature dimensionality reduction analysis on the source domain and the target domain respectively to obtain the variance explanation rate of each principal component, and then fusing the time domain features with the frequency domain features to obtain labeled weighing sensor samples and unlabeled vehicle scale samples respectively, wherein a reference model of the characteristic signals under different fault types is established to determine the threshold range under different faults.

6. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: In step S4, "mapping to the feature space through TCA feature migration to obtain source domain sample features and target domain sample features" is specifically as follows: TCA feature migration is used to map source and target domain samples to the reproducing kernel Hilbert space. By minimizing the maximum mean difference between source and target domain samples in the feature subspace, a dimensionality-reduced feature subspace is obtained. The distribution difference between the source and target domains is reduced in this subspace, and source domain sample features and target domain sample features with a certain degree of similarity are obtained, thereby realizing cross-domain migration of fault knowledge.

7. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: The preprocessing in step S7 includes using a Butterworth filter to filter out dynamic signal interference from the collected weighing sensors, obtaining the output values of each weighing sensor of the dynamic vehicle scale through a correlation model, and then calculating the time domain features and frequency domain features through the output values of each sensor, and then fusing the time domain features and frequency domain features through dimensionality reduction processing.

8. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1 is characterized by: The method also includes using the collected relevant data of the dynamic vehicle scale to synchronously judge abnormal behavior, and further screening the output of step S7. If there is no abnormal behavior, it is judged as the corresponding fault type. Otherwise, a vehicle abnormal behavior warning is directly issued; wherein, the abnormal vehicle behavior includes: crossing lane behavior, wrong-way behavior, edge pressing behavior, S-shaped behavior, parking behavior, scale jumping behavior, scale dragging behavior, scale grinding behavior, queuing behavior and jacking behavior. The judgment of these abnormal behaviors is achieved by deeply integrating the perception data of the roadside terminal equipment, and combining the multi-channel data of the weighing system with optical measurement methods to judge the vehicle weight and behavior.

9. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 8, characterized in that: The specific process of judging abnormal behavior and its handling methods include: The lane-crossing behavior is determined based on its characteristics, which include: the vehicle straddling the centerline of a lane, with part of its weight being taken from the adjacent lane, and the weight measured by the current lane's weighing equipment being inaccurate. The lane-crossing behavior is determined by estimating the vehicle's position and status based on multi-channel sensor signals from all weighing equipment within the station, determining its driving position and direction through coil signals, and comparing the vehicle's size and position measured by lidar with the simultaneous weighing signals to determine whether the vehicle is in a lane-crossing state. The lane-crossing behavior is processed as follows: the system indicates this abnormal state, and sensor signals matching the vehicle's gross weight based on position are accumulated and uploaded. The wrong-way behavior is judged based on the characteristics of the wrong-way behavior, which are: the vehicle crosses the center line of the lane, the entire weight is obtained by the weighing equipment of the adjacent opposite lane, there is no weighing data in the current lane, and an error occurs in the weighing process of the opposite lane; the wrong-way behavior is judged in the following way: the vehicle position and status are estimated based on the multi-channel sensor signals of all weighing equipment in the station, the driving position and direction are judged by the coil signal, and the vehicle size, position and driving direction obtained by the laser radar are compared with the weighing signal at the same time to determine whether the vehicle is in a wrong-way state; the wrong-way behavior is processed as follows: the abnormal state is prompted in the system, and the sensor signal of the opposite lane belonging to the total weight of the vehicle is matched according to the position and accumulated and uploaded, and the rear camera of the vehicle is called as the front camera to identify the vehicle license plate; The edge pressing behavior is judged based on its characteristics, which are: the weighing equipment does not cover the entire width of the lane, and the vehicle is partially pressed on the road surface without a weighing platform; the edge pressing behavior is judged by judging whether the balance of the left and right weights exceeds a threshold based on the multi-channel sensor signal of the weighing equipment, and comparing the vehicle size, position and driving direction measured by the lidar with the weighing signal at the same time to judge whether the vehicle is in an edge pressing state; the edge pressing behavior is processed as follows: the abnormal state is prompted in the system, and the left half of the vehicle weight belonging to the vehicle is matched according to the position as the reference half weight, and twice the weight is uploaded as temporary data for subsequent review; The S-shaped behavior is judged based on its characteristics, which include: the vehicle passing through the weighing area in a swaying route, so that the two wheels on the same axle of the vehicle do not pass through the weighing area in the lane at the same time; the S-shaped behavior is judged by determining the characteristics of the axle weight and wheel weight peaks based on the multi-channel sensor signals of all weighing equipment at the site, combining the vehicle's driving trajectory measured by lidar with the weighing signals of the same period to determine whether the vehicle is in an S-shaped driving state; the S-shaped behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when the vehicle passed through the weighing area are uploaded as temporary data for subsequent review; The parking behavior is judged based on the characteristics of the parking behavior, which are: parking in the weighing area for a long time, resulting in erroneous signals from the ground sensor coil vehicle detector, errors in the dynamic vehicle scale program for axle weight measurement, and inability to correctly calculate the matching total weight; the parking behavior is judged by estimating the vehicle state based on the multi-channel sensor signal of the weighing equipment, judging whether the vehicle is in a driving state through the coil signal, confirming the vehicle trajectory obtained by laser radar measurement, and comparing it with the weighing signal at the same time to judge whether the vehicle is in a parked state; the parking behavior is processed as follows: the abnormal state is prompted in the system, and all weights measured by the weighing equipment during the period when all axles of the vehicle pass through the weighing area are uploaded as data; The weight-jump behavior is determined based on its characteristics, which include: the vehicle's acceleration causes it to lift its head, shifting its center of gravity rearward to reduce the measured front axle weight; the weight-jump behavior is determined by determining its acceleration behavior based on a vehicle speed-time curve measured by a laser radar; and the weight-jump behavior is processed as follows: the abnormal state is prompted in the system; The weighing behavior is determined based on its characteristics, which include: the vehicle is towed on the weighing device by locking the trailer axle, increasing the lateral force on the weighing sensor and causing the load carrier to become stuck, thereby reducing the measured trailer weight. The weighing behavior is determined by determining the acceleration and deceleration behavior based on the vehicle speed-time curve measured by the laser radar. The weighing behavior is processed as follows: the abnormal state is prompted in the system; The scale-wasting behavior is judged based on its characteristics, which include: the vehicle repeatedly moving forward and backward in the weighing area, resulting in errors in the accumulated axle count and total weight. The scale-wasting behavior is judged by estimating the vehicle's forward and backward state and the weight of the axles that are repeatedly weighed based on the weight information of all weighing devices in the station. The speed-time curve measured by the lidar is then compared with the weighing signal at the same time to determine the weight of the vehicle that has been repeatedly weighed. The scale-wasting behavior is processed as follows: the abnormal state is prompted in the system, and the repeatedly weighed weight and the repeatedly calculated axle count are eliminated. The queuing behavior is determined based on its characteristics: multiple vehicles queuing closely together, reducing the distance between vehicles, and intentionally interfering with the vehicle separator, causing weight errors and preventing the camera from recognizing license plates. The queuing behavior is determined by estimating the vehicle separation status based on the weight information of all weighing devices within the station, combined with the vehicle separation information obtained by lidar recognition. The queuing behavior is then processed by notifying the system of the abnormal status and matching the weights of the vehicles in the queue based on spatiotemporal information. The jack behavior is judged based on the characteristics of the jack behavior, which are: by installing a jack device at the bottom of the vehicle, the weight is transferred to the non-weighing area, thereby reducing the measured axle weight; the jack behavior is judged in the following way: based on the weight of each axle obtained by the weighing equipment, it is estimated whether the axle weight imbalance is caused by the jack, and at the same time, combined with the speed-time curve measured by the laser radar, it is judged whether the jack is operated to avoid inspection by stopping the vehicle; the jack behavior is processed as follows: the abnormal state is prompted in the system.

10. The abnormality self-diagnosis method of highway traffic overload control detection equipment according to claim 1, characterized in that: The method further includes step 8, after a fault is detected, using an intelligent fault-tolerant method for a dynamic weighing system based on an improved particle swarm optimization generalized regression neural network under a multi-correlation model to measure and estimate passing vehicles, specifically as follows: Eliminate deviation data caused by failure; The Lévy flight-modified particle swarm algorithm is used to search for the optimal solution to optimize the generalized regression neural network, and the LPSO-GRNN network model matching the truck scale data samples is obtained, and the output estimation network is established based on this model. A multi-correlation model is used to determine the expected characteristic values under different faults, and then the output prediction network is used to obtain the expected output estimate of the faulty sensor, thereby realizing the measurement estimate of the passing vehicles.

Images