Heavy-duty truck vehicle-mounted safety monitoring method and device

Abstract

The invention provides a heavy-duty truck vehicle-mounted safety monitoring method and device, and the method comprises the steps: obtaining the original monitoring data of a wheel-rail force through a wheel-rail force monitoring module which is installed on an axle and has a low power consumption characteristic, and carrying out the edge calculation of the original monitoring data, and obtaining a wheel load shedding rate, a derailment coefficient and a wheel axle transverse force; the framework acceleration and the bearing temperature are obtained through the low-power-consumption vibration acceleration sensing module and the bearing temperature sensing module; performing multi-modal fusion on each index to obtain a multi-modal operation state representation vector; obtaining a dynamic threshold value of each index based on an enhanced feature vector and a multi-modal operation state representation vector obtained by an offline self-learning model; and giving current situation analysis and trend analysis of safety monitoring based on a comparison result of each index and a dynamic threshold value. According to the invention, the limitation of independent analysis of single sensor data in the prior art is overcome, the dynamic adjustment of the safety threshold is realized, and the accuracy and reliability of safety monitoring are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of heavy haul railway safety monitoring, and particularly relates to a heavy haul train-mounted safety monitoring method and device. BACKGROUND

[0002] Heavy haul railway transportation plays an important role in the economy, and its safe operation is crucial. In order to ensure the safe operation of heavy haul trains, it is necessary to monitor their running state in real time. Existing heavy haul train safety monitoring technologies mainly focus on ground equipment monitoring, such as detection through TPDS, axle temperature detection stations, etc. However, these ground monitoring methods have certain limitations.

[0003] Firstly, ground monitoring equipment can only monitor the state of the vehicle when it passes through a specific location, and cannot achieve continuous monitoring of the vehicle during operation. Secondly, ground monitoring equipment is easily affected by environmental factors, such as adverse weather which can cause monitoring data distortion. Thirdly, ground monitoring systems are difficult to cover all lines comprehensively, and there are monitoring blind spots. In order to make up for the shortcomings of ground monitoring, train-mounted safety monitoring systems have emerged. However, existing train-mounted safety monitoring systems also have some problems.

[0004] Firstly, the existing train-mounted safety monitoring system often analyzes the monitoring data of multiple safety indicators such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration and bearing temperature independently, lacks effective fusion of multi-source heterogeneous data, and is difficult to accurately reflect the overall running state of heavy haul trains. Due to the complex correlation between various indicators, the abnormality of a single indicator may be affected by other indicators, and isolated analysis of each indicator can easily lead to misjudgment or missed judgment. How to effectively fuse multi-modal information and extract more comprehensive running state features is a difficulty in existing technology.

[0005] Secondly, existing technologies usually use fixed safety thresholds to evaluate monitoring data, lacking dynamic adjustment of the threshold. However, the running state of heavy haul trains is affected by various factors such as line conditions, vehicle load, running speed, environmental factors, etc. Fixed safety thresholds are difficult to adapt to complex and variable operating environments, and are prone to false alarms or missed alarms, reducing the accuracy and reliability of safety monitoring. How to dynamically adjust the safety threshold according to the actual running situation is another challenge faced by existing technology.

[0006] Furthermore, traditional safety monitoring systems have shortcomings in terms of current status analysis and trend analysis, making it difficult to fully utilize monitoring data. Existing systems typically only provide the current safety status assessment results, lacking in-depth mining and analysis of historical data, and thus unable to predict future safety risks. How to effectively utilize historical data to predict safety risk trends and provide decision support for safety management is an area where existing technologies need improvement. Summary of the Invention

[0007] In view of this, this application provides a method and device for on-board safety monitoring of heavy-duty trucks to solve at least one of the aforementioned problems.

[0008] To achieve the above objectives, this application adopts the following approach:

[0009] According to a first aspect of this application, a method for on-board safety monitoring of heavy-duty trucks is provided, the method comprising:

[0010] By using a wheel-rail force monitoring module installed on the axle with low power consumption, the raw monitoring data of wheel-rail force is obtained, and edge computing is performed on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel-axle lateral force.

[0011] The frame acceleration and bearing temperature are obtained through a low-power vibration acceleration sensing module and a bearing temperature sensing module.

[0012] The wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are fused in a multimodal manner to obtain a multimodal operating state characterization vector;

[0013] Based on the enhanced feature vector obtained from the offline self-learning model and the multimodal operating state characterization vector, the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are obtained.

[0014] Based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature, and the dynamic threshold, a current status analysis and trend analysis of safety monitoring are given.

[0015] As an embodiment of this application, the training process of the above-mentioned offline self-learning model is as follows:

[0016] Time series analysis was performed on historical multimodal operating state vectors and operating condition data to obtain trend labels and trend directions for safety monitoring;

[0017] An autoencoder feature extraction model is trained using historical multimodal running state vectors, and a low-dimensional enhanced feature vector is output.

[0018] An enhanced feature vector is generated by combining the low-dimensional enhanced feature vector, the trend label, and the trend direction.

[0019] As an embodiment of this application, the above method performs multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature to obtain a multimodal operating state characterization vector, including:

[0020] The wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are dimensionally aligned and then feature-stitched to obtain a preliminary multimodal input vector.

[0021] A fusion network with two layers of feedforward network is used to fuse the initial multimodal input vector and output a multimodal operating state representation vector.

[0022] As an embodiment of this application, the dynamic thresholds for the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature obtained by the above method based on the enhanced feature vector obtained from the offline self-learning model and the multimodal operating state characterization vector include:

[0023] The enhanced feature vector and the multimodal operating state characterization vector are input into the constructed dynamic threshold prediction model, and the dynamic thresholds corresponding to the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature are output respectively.

[0024] The dynamic threshold prediction model is a neural network. During the training process, the dynamic threshold prediction model uses the multimodal operating state representation vector and the enhanced feature vector output by the offline self-learning model as the training set, and optimizes it with the goal of minimizing the error between the predicted threshold and the actual safety state.

[0025] As an embodiment of this application, the above method provides a current status analysis and trend analysis of safety monitoring based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature, and the dynamic threshold, including:

[0026] The real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are compared with their corresponding dynamic thresholds to determine the current safety status of each indicator.

[0027] Based on the current safety status of each indicator, a comprehensive assessment of the overall safety status of the heavy-duty truck is conducted, and a safety monitoring report is generated. The report includes the current values, dynamic thresholds, safety status, and overall safety assessment results of the wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0028] The real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are combined with historical index data to form a predictive data sequence. Time series analysis is performed on the predictive data sequence to predict the future trend of each index.

[0029] Based on the changing trends of various indicators, the safety status of heavy-duty trucks in the future is predicted for a certain period of time, and a trend analysis report is generated. The report includes historical data, changing trends, prediction results, and safety risk warnings for each indicator.

[0030] As an embodiment of this application, the dynamic threshold in the above method includes a warning value and an alarm value. The method further includes: when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the warning value, sending a warning message to the locomotive driver's cab display terminal and / or the ground data analysis server; and when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the alarm value, sending an alarm message to the locomotive driver's cab display terminal and / or the ground data analysis server.

[0031] According to a second aspect of this application, a vehicle-mounted safety monitoring device for heavy-duty trucks is provided, the device comprising:

[0032] The wheel-rail force monitoring module is used to acquire the raw monitoring data of wheel-rail force and perform edge calculations on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel axle lateral force;

[0033] Vibration acceleration sensing module, used to acquire the acceleration of the structure;

[0034] Bearing temperature sensing module, used to acquire bearing temperature;

[0035] The parameter fusion module is used to perform multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature to obtain a multimodal operating state characterization vector;

[0036] The dynamic threshold acquisition module is used to obtain the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature based on the enhanced feature vector obtained by the offline self-learning model and the multimodal operating state characterization vector.

[0037] The safety monitoring and analysis module is used to provide a current status analysis and trend analysis of safety monitoring based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature and the dynamic threshold.

[0038] As an embodiment of this application, the above-mentioned apparatus further includes: an offline model training unit for training an offline self-learning model, the process of which is as follows:

[0039] Time series analysis was performed on historical multimodal operating state vectors and operating condition data to obtain trend labels and trend directions for safety monitoring;

[0040] An autoencoder feature extraction model is trained using historical multimodal running state vectors, and a low-dimensional enhanced feature vector is output.

[0041] An enhanced feature vector is generated by combining the low-dimensional enhanced feature vector, the trend label, and the trend direction.

[0042] As one embodiment of this application, the above parameter fusion module includes:

[0043] The preliminary fusion unit is used to dimensionally align the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature, and then perform feature stitching to obtain a preliminary multimodal input vector.

[0044] The secondary fusion unit is used to fuse the initial multimodal input vector using a fusion network with two layers of feedforward network, and output a multimodal operating state representation vector.

[0045] As an embodiment of this application, the above-mentioned dynamic threshold acquisition module is specifically used for:

[0046] The enhanced feature vector and the multimodal operating state characterization vector are input into the constructed dynamic threshold prediction model, and the dynamic thresholds corresponding to the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature are output respectively.

[0047] The dynamic threshold prediction model is a neural network. During the training process, the dynamic threshold prediction model uses the multimodal operating state representation vector and the enhanced feature vector output by the offline self-learning model as the training set, and optimizes it with the goal of minimizing the error between the predicted threshold and the actual safety state.

[0048] As one embodiment of this application, the above-mentioned security monitoring and analysis module includes:

[0049] The comparison unit is used to compare the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with the corresponding dynamic thresholds to determine the current safety status of each indicator.

[0050] The current status analysis unit is used to comprehensively evaluate the overall safety status of heavy-duty trucks based on the current safety status of each indicator, and generate a safety monitoring report. The report includes the current values, dynamic thresholds, safety status, and overall safety assessment results of the wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0051] The trend prediction unit is used to combine the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with historical index data to form a prediction data sequence, perform time series analysis on the prediction data sequence, and predict the future trend of each index.

[0052] The trend analysis unit is used to predict the safety status of heavy-duty trucks over a future period of time based on the changing trends of various indicators, and generate a trend analysis report. The report includes historical data, changing trends, prediction results, and safety risk warnings for each indicator.

[0053] As an embodiment of this application, the aforementioned dynamic threshold includes a warning value and an alarm value, and the device further includes:

[0054] The early warning unit is used to send early warning information to the locomotive driver's cab display terminal and / or ground data analysis server when the real-time monitoring value of any one of the following indicators exceeds the early warning value: wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0055] The alarm unit is used to send alarm information to the locomotive driver's cab display terminal and / or ground data analysis server when the real-time monitoring value of any one of the following indicators exceeds the alarm value: wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0056] According to a third aspect of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0057] According to a fourth aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described method.

[0058] According to a fifth aspect of the present invention, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the above-described method.

[0059] As can be seen from the above technical solutions, the heavy-duty truck on-board safety monitoring method and device provided in this application overcomes the limitations of independent analysis of single sensor data in existing technologies. By integrating multi-source heterogeneous data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature, it can more accurately reflect the overall operating status of heavy-duty trucks and reduce false alarms or missed alarms. It achieves dynamic adjustment of safety thresholds, improving the accuracy and reliability of safety monitoring. It overcomes the shortcomings of using fixed safety thresholds in existing technologies, enabling dynamic adjustment of safety thresholds based on actual operating conditions, adapting to complex and changing operating environments, and reducing false alarms or missed alarms. Furthermore, this application enhances the capabilities of status quo and trend analysis, providing decision support for safety management. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0061] Figure 1 This is a flowchart illustrating a method for onboard safety monitoring of heavy-duty trucks provided in an embodiment of this application;

[0062] Figure 2 This is a schematic diagram of the offline self-learning model training process provided in the embodiments of this application;

[0063] Figure 3 This is a schematic diagram of the process for obtaining the multimodal operating state representation vector provided in an embodiment of this application;

[0064] Figure 4 This is a flowchart illustrating the current status and trend analysis of security monitoring provided in the embodiments of this application;

[0065] Figure 5 This is a schematic diagram of the structure of an on-board safety monitoring device for heavy-duty trucks provided in an embodiment of this application;

[0066] Figure 6This is a schematic block diagram of the system configuration of the electronic device provided in the embodiments of the invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the embodiments of this application will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments and descriptions of this application are used to explain this application, but are not intended to limit this application.

[0068] like Figure 1 The diagram shown is a flowchart illustrating a method for on-board safety monitoring of heavy-duty trucks according to an embodiment of this application. The method includes the following steps:

[0069] Step S101: Obtain the raw monitoring data of wheel-rail force by using a wheel-rail force monitoring module installed on the axle with low power consumption, and perform edge calculation on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel-axle lateral force.

[0070] In this embodiment, on heavy-duty trucks, non-destructive continuous force measurement wheelset technology can be used to replace the intermittent force measurement wheelset with drilling. Non-destructive continuous force measurement wheelsets do not require modification of the wheelset, and do not require machining of wire holes or threading holes. Therefore, they do not affect the overall structure of the wheelset, nor its structural strength or fatigue life. Furthermore, compared to intermittent measurement, continuous measurement can test the wheel-rail force at any angle of the wheel, providing richer and more comprehensive data.

[0071] In this embodiment, a low-power wheel-rail force monitoring module installed on the axle of a heavy-duty truck collects raw wheel-rail force monitoring data in real time. This low-power wheel-rail force monitoring module replaces the traditional current collector ring + acquisition device model, reducing power consumption and enabling long-term fixed installation. The force sensors are installed on the inside of the wheel and on the axle, without affecting the normal maintenance of the wheelset.

[0072] The wheel-rail force monitoring module in this embodiment integrates a high-density ring-shaped battery and a low-power wheel-rail force acquisition, analysis, and wireless data transmission module. The collected raw wheel-rail force data can be processed using edge computing on the low-power wheel-rail force acquisition and analysis module to calculate operational safety indicators such as wheel load reduction rate, derailment coefficient, and wheel axle lateral force. Edge computing reduces data transmission volume, lowers communication pressure, and improves response speed. Monitoring data is wirelessly transmitted to a wheel-rail force data receiving module installed on the vehicle body, which is then connected to the monitoring host via a wired connection.

[0073] Step S102: Obtain the frame acceleration and bearing temperature through the low-power vibration acceleration sensing module and the bearing temperature sensing module.

[0074] In this embodiment, a low-power vibration acceleration sensing module and a bearing temperature sensing module are used to obtain the acceleration information of the heavy-duty truck frame and the temperature information of the bearing, respectively.

[0075] The low-power vibration acceleration testing module includes a low-power vibration acceleration sensor and a low-power signal acquisition circuit board. The low-power vibration acceleration sensor utilizes silicon MEMS technology, featuring small size and low drive energy requirements. The vibration acceleration sensor and signal acquisition circuit board together form the vibration acceleration testing module, which is connected to the monitoring host via a wired connection. The bearing temperature sensing module consists of an infrared temperature sensor and a signal acquisition and analysis unit, and it is also connected to the monitoring host via a wired connection.

[0076] Step S103: Perform multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature to obtain a multimodal operating state characterization vector.

[0077] The multimodal data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration and bearing temperature obtained in steps S101 and S102 are fused to obtain a multimodal operating state representation vector that can comprehensively characterize the operating state of heavy-duty trucks. The fused multimodal operating state representation vector can more accurately reflect the overall operating state of heavy-duty trucks and avoid the one-sidedness of a single indicator.

[0078] Step S104: Based on the enhanced feature vector obtained from the offline self-learning model and the multimodal operating state representation vector, obtain the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature.

[0079] In this embodiment, the offline self-learning model in this step can be trained using historical operating data to learn the reasonable threshold ranges for various safety indicators under different operating conditions. The enhanced feature vector can include historical operating data, line information, environmental factors, etc., which are used to improve the accuracy of dynamic threshold prediction. Therefore, the dynamic threshold of this application can be adjusted according to actual operating conditions, adapting to complex and ever-changing operating environments and reducing false alarms or missed alarms.

[0080] Step S105: Based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature and the dynamic threshold, a current status analysis and trend analysis of safety monitoring are given.

[0081] This step compares the real-time monitored wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature with the corresponding dynamic thresholds to determine the current safety status of each indicator and comprehensively assess the overall safety status of the heavy-duty truck. Simultaneously, by combining historical monitoring data, it predicts the future trends of each indicator and assesses the safety status of the heavy-duty truck over a future period.

[0082] In this embodiment, a safety monitoring report can be generated based on the safety status and changing trends of various indicators, and the report can be sent to the locomotive driver's cab display terminal and the ground data analysis server.

[0083] As can be seen from the above technical solution, the on-board safety monitoring method for heavy-duty trucks provided in this application overcomes the limitations of independent analysis of single sensor data in existing technologies. By integrating multi-source heterogeneous data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature, it can more accurately reflect the overall operating status of heavy-duty trucks and reduce false alarms or missed alarms. It achieves dynamic adjustment of safety thresholds, improving the accuracy and reliability of safety monitoring. It overcomes the shortcomings of using fixed safety thresholds in existing technologies, enabling dynamic adjustment of safety thresholds based on actual operating conditions, adapting to complex and changing operating environments, and reducing false alarms or missed alarms. Furthermore, this application enhances the capabilities of status quo and trend analysis, providing decision support for safety management.

[0084] In one embodiment of this application, such as Figure 2 As shown, the offline self-learning model training process in step S104 above is as follows:

[0085] Step S201: Perform time series analysis on historical multimodal operating state vectors and operating condition data to obtain trend labels and trend directions for safety monitoring.

[0086] This step involves performing time series analysis on historical multimodal operating state vectors (including data such as wheel load reduction rate, derailment coefficient, wheel and axle lateral force, frame acceleration, and bearing temperature) and operating condition data. The aim is to uncover the trend information of safety monitoring indicators over time, thereby providing a basis for subsequent dynamic threshold prediction.

[0087] First, collect operational data of heavy-duty trucks over a period of time, including multimodal operational state vectors and operating condition data. Operating condition data may include information such as route type, vehicle load, operating speed, ambient temperature, and weather conditions.

[0088] Then, time series analysis methods are used for analysis. In this embodiment, various time series analysis methods can be used, such as moving average, exponential smoothing, or ARIMA model.

[0089] Next, based on the results of the time series analysis, a trend label is defined for each safety monitoring indicator. For example, the trend label can be divided into three types: "risk rising," "risk falling," and "stable," which respectively indicate that the indicator will show an upward, downward, or stable trend in the future.

[0090] Finally, the direction of the trend can be determined based on the trend label, such as whether it is rising rapidly or slowly, or falling rapidly or slowly. Here, the trend direction can be quantified as a numerical value or as a conclusive textual description.

[0091] Step S202: Train the autoencoder feature extraction model using historical multimodal running state vectors and output a low-dimensional enhanced feature vector.

[0092] In this implementation, the autoencoder feature extraction model is trained using historical multimodal operating state vectors. The purpose is to extract key features from the multimodal data and reduce their dimensionality to obtain low-dimensional enhanced feature vectors.

[0093] An autoencoder is a neural network model consisting of an encoder and a decoder. The encoder compresses the input data into a low-dimensional feature vector, and the decoder restores the original data from the low-dimensional feature vector. The autoencoder can be trained by minimizing the reconstruction error (the difference between the original data and the decoder output). After training, the encoder can extract key features from the input data. The encoder's output is the low-dimensional enhanced feature vector, which contains the main information from the multimodal data and has a dimensionality reduction effect, reducing the complexity of subsequent calculations.

[0094] Step S203: Combine the low-dimensional enhanced feature vector, the trend label, and the trend direction to generate an enhanced feature vector.

[0095] This step combines the trend label and trend direction obtained in step S201 with the low-dimensional enhanced feature vector obtained in step S202 to generate the final enhanced feature vector.

[0096] Trend labels, trend directions, and low-dimensional enhanced feature vectors are fused. The fusion method can be simple concatenation or a more complex neural network model; this embodiment is not limited to either. The fused vector is the enhanced feature vector. It contains both key features of multimodal data and trend information of safety monitoring indicators, enabling a more comprehensive characterization of the operating status of heavy-duty trucks and providing richer information for subsequent dynamic threshold prediction.

[0097] As can be seen, the offline self-learning model training process provided in this embodiment utilizes time series analysis to extract trend information, utilizes an autoencoder to extract key features, and fuses the two to finally obtain an enhanced feature vector, laying the foundation for dynamic threshold prediction.

[0098] In another embodiment of this application, such as Figure 3 As shown, in step S103 above, the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are fused in a multimodal manner to obtain a multimodal operating state characterization vector, including:

[0099] Step S1031: Align the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature dimensionally and then perform feature stitching to obtain a preliminary multimodal input vector.

[0100] Since data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration and bearing temperature may have different dimensions, it is first necessary to align these data dimensions and then stitch them together to form a preliminary multimodal input vector.

[0101] When performing maintenance, firstly, standardize the sampling frequency to ensure that all modal data have the same sampling frequency. If different modal data have different sampling frequencies, interpolation or downsampling methods can be used to unify them. Next, determine a suitable time window and capture data from all modalities within that window. Then, standardize or normalize the data from different modalities to eliminate the influence of units and numerical ranges, making them comparable.

[0102] The data from each modality, after being aligned in dimensions, are concatenated in a specific order to form a long vector. For example, data on wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature can be concatenated sequentially. The concatenated vector is the initial multimodal input vector, containing data information from all modalities.

[0103] Step S1032: Use a fusion network with two layers of feedforward network to fuse the preliminary multimodal input vector and output a multimodal operating state representation vector.

[0104] A fusion network with two layers of feedforward network is used to fuse the initial multimodal input vectors. The purpose is to extract the correlation between multimodal data and generate a vector that can comprehensively represent the operating status of heavy-duty trucks.

[0105] The fusion network structure in this embodiment may include an input layer, a first feedforward network, a second feedforward network, and an output layer. The input layer receives the initial multimodal input vector. The first feedforward network contains multiple neurons that perform linear transformations and nonlinear activations on the input vector. The second feedforward network further performs linear transformations and nonlinear activations on the output of the first network. The output layer outputs a multimodal operating state representation vector. Activation functions such as ReLU, Sigmoid, or Tanh can be used to introduce nonlinear characteristics and enhance the model's expressive power.

[0106] During training, the weights and biases of the fusion network are adjusted using the backpropagation algorithm so that it can learn the correlations between multimodal data.

[0107] The multimodal fusion process provided in this step first integrates data from different modalities through dimension alignment and feature concatenation, then uses a fusion network to extract the correlation between the multimodal data, and finally obtains a vector that can comprehensively represent the operating status of heavy-duty trucks.

[0108] In another embodiment of this application, the step S104 above, which uses the enhanced feature vector obtained from the offline self-learning model and the multimodal operating state characterization vector to obtain the dynamic thresholds for the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature, includes: inputting the enhanced feature vector and the multimodal operating state characterization vector into the constructed dynamic threshold prediction model, and outputting the dynamic thresholds corresponding to the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature, respectively.

[0109] The dynamic threshold prediction model is a neural network. During the training process, the dynamic threshold prediction model uses the multimodal operating state representation vector and the enhanced feature vector output by the offline self-learning model as the training set, and optimizes it with the goal of minimizing the error between the predicted threshold and the actual safety state.

[0110] As described above, the input to the dynamic threshold prediction model in this embodiment includes two parts: an enhanced feature vector and a multimodal operating state representation vector. The enhanced feature vector integrates historical operating data, route information, environmental factors, and trend information learned from historical data. It is the output of the offline self-learning model, providing global and trend-based information. The multimodal operating state representation vector contains real-time monitoring data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature. It reflects the current operating state of the heavy-duty freight truck.

[0111] The dynamic threshold prediction model is a neural network with powerful nonlinear mapping capabilities, enabling it to learn the complex relationship between input data and dynamic thresholds. It uses a multimodal operating state representation vector and enhanced feature vectors output by an offline self-learning model as the training set. Optimization aims to minimize the error between the predicted threshold and the actual safety state, which can be obtained through expert experience, historical accident data, or simulation results. Through backpropagation, the weights and biases of the neural network are continuously adjusted, allowing the model to more accurately predict the dynamic thresholds of various safety indicators. The trained dynamic threshold prediction model outputs dynamic thresholds corresponding to wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature. These dynamic thresholds can be adjusted according to actual operating conditions to adapt to complex and changing operating environments, improving the accuracy and reliability of safety monitoring.

[0112] In another embodiment of this application, such as Figure 4 As shown, in step S105 above, based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature, and the dynamic threshold, the current status analysis and trend analysis of safety monitoring are given, including:

[0113] Step S1051: Compare the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with the corresponding dynamic thresholds to determine the current safety status of each indicator.

[0114] In this step, the real-time monitoring values ​​of wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature are compared with their corresponding dynamic thresholds to determine the current safety status of each indicator. This comparison can be a direct comparison of the real-time monitoring value and the dynamic threshold, or it can involve calculating the difference or ratio between the real-time monitoring value and the dynamic threshold, setting a tolerance range, and determining whether the indicator exceeds the safety range.

[0115] The security status can be divided into multiple levels, such as "normal", "warning", and "alarm", which correspond to different security risk levels. Of course, more detailed security status levels can also be set according to actual needs.

[0116] If the real-time monitoring value is less than the dynamic threshold, the indicator is considered to be in a "normal" state; if the real-time monitoring value is close to the dynamic threshold but has not yet exceeded it, the indicator is considered to be in a "warning" state; if the real-time monitoring value exceeds the dynamic threshold, the indicator is considered to be in an "alarm" state.

[0117] Step S1052: Based on the current safety status of each indicator, comprehensively evaluate the current overall safety status of the heavy-duty truck and generate a safety monitoring report. The report includes the current values ​​of the wheel load reduction rate, the derailment coefficient, the lateral force of the wheel axle, the frame acceleration, and the bearing temperature, as well as the dynamic threshold, safety status, and overall safety assessment results.

[0118] In this step, the overall safety status of the heavy-duty truck is comprehensively assessed based on the current safety status of each indicator, and a safety monitoring report is generated. For example, if all indicators are in a "normal" state, the overall safety status of the heavy-duty truck is considered good. If some indicators are in a "warning" state, it is necessary to closely monitor the changing trends of these indicators and take corresponding preventive measures. If some indicators are in an "alarm" state, it is necessary to take immediate emergency measures to prevent accidents from occurring.

[0119] The safety monitoring report can include the current values ​​of wheel load reduction rate, derailment coefficient, wheel and axle lateral force, frame acceleration, and bearing temperature; the dynamic thresholds corresponding to each indicator; the safety status of each indicator; and the overall safety status assessment results for the heavy-duty truck. Other relevant information, such as route information, vehicle information, and environmental information, can be added to the report as needed.

[0120] Step S1053: Combine the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with historical index data to form a predictive data sequence. Perform time series analysis on the predictive data sequence to predict the future trend of each index.

[0121] In this step, real-time monitoring values ​​of wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature are combined with historical index data to form a predictive data sequence. Time series analysis is then performed on the predictive data sequence to predict the future trends of each index.

[0122] Real-time monitoring values ​​and historical indicator data are arranged chronologically to form a time series. The appropriate length of historical data can be selected as needed. In this embodiment, various time series analysis methods can be used, such as moving average, exponential smoothing, and ARIMA models. The prediction results can be represented by graphs or tables showing the future trends of each indicator, or by providing predicted values ​​and confidence intervals for a future period.

[0123] Step S1054: Based on the changing trends of each indicator, predict the safety status of heavy-duty trucks in the future and generate a trend analysis report. The report includes historical data, changing trends, prediction results, and safety risk warnings for each indicator.

[0124] In this step, based on the changing trends of various indicators, the safety status of heavy-duty trucks is predicted for a future period, and a trend analysis report is generated. If the changing trends of all indicators are relatively stable and will not exceed the dynamic threshold in the future, the future safety status of heavy-duty trucks is considered good. If the changing trends of some indicators are more obvious and may exceed the dynamic threshold in the future, preventive measures need to be taken in advance to reduce safety risks.

[0125] The trend analysis report may include: historical data for each indicator, the changing trends of each indicator, the predicted results of each indicator, and safety risk warnings, such as "In the near future, the derailment coefficient may exceed the dynamic threshold; please check the wheel-rail status." Of course, this embodiment can also add other relevant information to the report as needed, such as recommended operating speeds and maintenance measures.

[0126] By following the above four steps, the safety status of heavy-duty trucks can be comprehensively and dynamically monitored and analyzed, potential safety hazards can be identified in a timely manner, and the safe operation of heavy-duty trucks can be ensured.

[0127] In another embodiment of this application, the dynamic threshold in the above method includes a warning value and an alarm value. The method further includes: when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the warning value, sending a warning message to the locomotive driver's cab display terminal and / or the ground data analysis server; and when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the alarm value, sending an alarm message to the locomotive driver's cab display terminal and / or the ground data analysis server.

[0128] In this embodiment, the warning value is a lower threshold for abnormal indicators, indicating that the operating state of the system or component may deviate from the normal range, but has not yet reached the point where immediate action is required. When the monitored value exceeds the warning value, it indicates a potential risk that requires close monitoring. The alarm value is a higher threshold for abnormal indicators, indicating that the operating state of the system or component has exceeded the safe range and may lead to serious consequences. When the monitored value exceeds the alarm value, immediate measures must be taken to prevent an accident from occurring.

[0129] In this embodiment, the warning value and alarm value are not fixed and can be dynamically adjusted according to the actual operating conditions. Specifically, they are dynamically determined according to step S104.

[0130] The system continuously monitors the real-time values ​​of key indicators such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature. When the real-time monitoring value of any indicator exceeds its corresponding warning value, the system immediately generates a warning message and sends it to the locomotive driver's cab display terminal to remind the driver to pay attention and take appropriate measures. The message is also sent to the ground data analysis server to facilitate analysis and decision-making by ground personnel.

[0131] When the real-time monitoring value of any indicator exceeds its corresponding alarm value, the system immediately generates an alarm message and sends it to the locomotive driver's cab display terminal to remind the driver to take immediate emergency measures. The message is also sent to the ground data analysis server to facilitate remote intervention or other emergency measures by ground personnel.

[0132] Drivers can adjust train operation in a timely manner based on received warnings and alarms, such as reducing speed and checking equipment, to ensure driving safety. Ground personnel can comprehensively analyze the received data to assess the overall safety status of heavy-duty freight cars and take corresponding maintenance or management measures.

[0133] like Figure 5 The diagram shown is a structural schematic of a heavy-duty truck on-board safety monitoring device according to an embodiment of this application. The device includes: a wheel-rail force monitoring module 510, a vibration acceleration sensing module 520, a bearing temperature sensing module 530, a parameter fusion module 540, a dynamic threshold acquisition module 550, and a safety monitoring and analysis module 560. The parameter fusion module 540 is connected to the wheel-rail force monitoring module 510, the vibration acceleration sensing module 520, and the bearing temperature sensing module 530, respectively. The dynamic threshold acquisition module 550 is connected to the parameter fusion module 540 and the safety monitoring and analysis module 560, respectively.

[0134] The wheel-rail force monitoring module 510 is used to acquire the raw monitoring data of wheel-rail force and perform edge calculations on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel axle lateral force.

[0135] Vibration acceleration sensing module 520 is used to acquire the frame acceleration;

[0136] Bearing temperature sensing module 530 is used to acquire bearing temperature;

[0137] The parameter fusion module 540 is used to perform multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature to obtain a multimodal operating state characterization vector;

[0138] The dynamic threshold acquisition module 550 is used to obtain the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature based on the enhanced feature vector obtained by the offline self-learning model and the multimodal operating state characterization vector.

[0139] The safety monitoring and analysis module 560 is used to provide a current status analysis and trend analysis of safety monitoring based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature and the dynamic threshold.

[0140] In one embodiment of this application, the above-mentioned apparatus further includes: an offline model training unit for training an offline self-learning model, the process of which is as follows:

[0141] Time series analysis was performed on historical multimodal operating state vectors and operating condition data to obtain trend labels and trend directions for safety monitoring;

[0142] An autoencoder feature extraction model is trained using historical multimodal running state vectors, and a low-dimensional enhanced feature vector is output.

[0143] An enhanced feature vector is generated by combining the low-dimensional enhanced feature vector, the trend label, and the trend direction.

[0144] In one embodiment of this application, the parameter fusion module includes:

[0145] The preliminary fusion unit is used to dimensionally align the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature, and then perform feature stitching to obtain a preliminary multimodal input vector.

[0146] The secondary fusion unit is used to fuse the initial multimodal input vector using a fusion network with two layers of feedforward network, and output a multimodal operating state representation vector.

[0147] In one embodiment of this application, the dynamic threshold acquisition module is specifically used for:

[0148] The enhanced feature vector and the multimodal operating state characterization vector are input into the constructed dynamic threshold prediction model, and the dynamic thresholds corresponding to the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature are output respectively.

[0149] The dynamic threshold prediction model is a neural network. During the training process, the dynamic threshold prediction model uses the multimodal operating state representation vector and the enhanced feature vector output by the offline self-learning model as the training set, and optimizes it with the goal of minimizing the error between the predicted threshold and the actual safety state.

[0150] In one embodiment of this application, the security monitoring and analysis module includes:

[0151] The comparison unit is used to compare the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with the corresponding dynamic thresholds to determine the current safety status of each indicator.

[0152] The current status analysis unit is used to comprehensively evaluate the overall safety status of heavy-duty trucks based on the current safety status of each indicator, and generate a safety monitoring report. The report includes the current values, dynamic thresholds, safety status, and overall safety assessment results of the wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0153] The trend prediction unit is used to combine the real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature with historical index data to form a prediction data sequence, perform time series analysis on the prediction data sequence, and predict the future trend of each index.

[0154] The trend analysis unit is used to predict the safety status of heavy-duty trucks over a future period of time based on the changing trends of various indicators, and generate a trend analysis report. The report includes historical data, changing trends, prediction results, and safety risk warnings for each indicator.

[0155] In one embodiment of this application, the aforementioned dynamic threshold includes a warning value and an alarm value, and the device further includes:

[0156] The early warning unit is used to send early warning information to the locomotive driver's cab display terminal and / or ground data analysis server when the real-time monitoring value of any one of the following indicators exceeds the early warning value: wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0157] The alarm unit is used to send alarm information to the locomotive driver's cab display terminal and / or ground data analysis server when the real-time monitoring value of any one of the following indicators exceeds the alarm value: wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature.

[0158] As can be seen from the above technical solution, the heavy-duty truck on-board safety monitoring device provided in this application overcomes the limitations of independent analysis of single sensor data in existing technologies. By integrating multi-source heterogeneous data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature, it can more accurately reflect the overall operating status of heavy-duty trucks and reduce false alarms or missed alarms. It achieves dynamic adjustment of safety thresholds, improving the accuracy and reliability of safety monitoring. It overcomes the shortcomings of using fixed safety thresholds in existing technologies, enabling dynamic adjustment of safety thresholds based on actual operating conditions, adapting to complex and changing operating environments, and reducing false alarms or missed alarms. Furthermore, this application enhances the capabilities of status quo and trend analysis, providing decision support for safety management.

[0159] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method.

[0160] This application also provides a computer-readable storage medium storing a computer program that performs the above-described methods.

[0161] This application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the above-described method.

[0162] like Figure 6 The electronic device 600 may also include: a communication module 110, an input unit 120, an audio processor 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily need to include these components. Figure 6 All components shown; in addition, electronic device 600 may also include Figure 6 The components shown can be referenced in the prior art.

[0163] like Figure 6 The central processing unit 100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device. The central processing unit 100 receives inputs and controls the operation of various components of the electronic device 600.

[0164] The memory 140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 100 may execute the program stored in the memory 140 to perform information storage or processing, etc.

[0165] Input unit 120 provides input to central processing unit 100. Input unit 120 may be, for example, a keypad or touch input device. Power supply 170 provides power to electronic device 600. Display 160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0166] The memory 140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 140 can also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142 for storing application and function programs or processes for executing the operation of the electronic device 600 via the central processing unit 100.

[0167] The memory 140 may also include a data storage unit 143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 may include various drivers for the electronic device for communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0168] The communication module 110 is a transmitter / receiver that sends and receives signals via the antenna 111. The communication module (transmitter / receiver) is coupled to the central processing unit 100 to provide input signals and receive output signals, which can be the same as in a conventional mobile communication terminal.

[0169] Based on different communication technologies, multiple communication modules 110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication modules (transmitters / receivers) are also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide audio output via the speaker 131 and receive audio input from the microphone 132, thereby enabling typical telecommunications functions. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 130 is coupled to a central processing unit 100, enabling on-device recording via the microphone 132 and on-device playback of stored audio via the speaker 131.

[0170] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0171] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0172] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0173] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0174] This application uses specific embodiments to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

[0175] As can be seen from the above technical solution, the on-board safety monitoring method for heavy-duty trucks provided in this application overcomes the limitations of independent analysis of single sensor data in existing technologies. By integrating multi-source heterogeneous data such as wheel load reduction rate, derailment coefficient, wheel axle lateral force, frame acceleration, and bearing temperature, it can more accurately reflect the overall operating status of heavy-duty trucks and reduce false alarms or missed alarms. It achieves dynamic adjustment of safety thresholds, improving the accuracy and reliability of safety monitoring. It overcomes the shortcomings of using fixed safety thresholds in existing technologies, enabling dynamic adjustment of safety thresholds based on actual operating conditions, adapting to complex and changing operating environments, and reducing false alarms or missed alarms. Furthermore, this application enhances the capabilities of status quo and trend analysis, providing decision support for safety management.

Claims

1. A method for on-board safety monitoring of heavy-duty trucks, characterized in that, The method includes: By using a wheel-rail force monitoring module installed on the axle with low power consumption, the raw monitoring data of wheel-rail force is obtained, and edge computing is performed on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel-axle lateral force. The frame acceleration and bearing temperature are obtained through a low-power vibration acceleration sensing module and a bearing temperature sensing module. The wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are fused in a multimodal manner to obtain a multimodal operating state characterization vector; Based on the enhanced feature vector obtained from the offline self-learning model and the multimodal operating state characterization vector, the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are obtained. Based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature, and the dynamic threshold, a current status analysis and trend analysis of safety monitoring are given.

2. The method for on-board safety monitoring of heavy-duty trucks as described in claim 1, characterized in that, The offline self-learning model training process is as follows: Time series analysis was performed on historical multimodal operating state vectors and operating condition data to obtain trend labels and trend directions for safety monitoring; An autoencoder feature extraction model is trained using historical multimodal running state vectors, and a low-dimensional enhanced feature vector is output. An enhanced feature vector is generated by combining the low-dimensional enhanced feature vector, the trend label, and the trend direction.

3. The method for on-board safety monitoring of heavy-duty trucks as described in claim 1, characterized in that, The process of multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature to obtain a multimodal operating state characterization vector includes: The wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are dimensionally aligned and then feature-stitched to obtain a preliminary multimodal input vector. A fusion network with two layers of feedforward network is used to fuse the initial multimodal input vector and output a multimodal operating state representation vector.

4. The method for on-board safety monitoring of heavy-duty trucks as described in claim 1, characterized in that, The enhanced feature vector obtained based on the offline self-learning model and the multimodal operating state representation vector are used to obtain the dynamic thresholds for the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature, including: The enhanced feature vector and the multimodal operating state characterization vector are input into the constructed dynamic threshold prediction model, and the dynamic thresholds corresponding to the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature are output respectively. The dynamic threshold prediction model is a neural network. During the training process, the dynamic threshold prediction model uses the multimodal operating state representation vector and the enhanced feature vector output by the offline self-learning model as the training set, and optimizes it with the goal of minimizing the error between the predicted threshold and the actual safety state.

5. The method for on-board safety monitoring of heavy-duty trucks as described in claim 1, characterized in that, The comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature, and the dynamic threshold provide a current status analysis and trend analysis of safety monitoring, including: The real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are compared with their corresponding dynamic thresholds to determine the current safety status of each indicator. Based on the current safety status of each indicator, a comprehensive assessment of the overall safety status of the heavy-duty truck is conducted, and a safety monitoring report is generated. The safety monitoring report includes the current values, dynamic thresholds, safety status, and overall safety assessment results of the wheel load reduction rate, the derailment coefficient, the lateral force of the wheel axle, the frame acceleration, and the bearing temperature. The real-time monitoring values ​​of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature are combined with historical index data to form a predictive data sequence. Time series analysis is performed on the predictive data sequence to predict the future trend of each index. Based on the changing trends of various indicators, the safety status of heavy-duty trucks in the future is predicted for a certain period of time, and a trend analysis report is generated. The report includes historical data, changing trends, prediction results, and safety risk warnings for each indicator.

6. The method for on-board safety monitoring of heavy-duty trucks as described in claim 5, characterized in that, The dynamic threshold includes a warning value and an alarm value. The method further includes: when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the warning value, sending a warning message to the locomotive driver's cab display terminal and / or the ground data analysis server; and when the real-time monitoring value of any one of the following indicators—the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, and the bearing temperature—exceeds the alarm value, sending an alarm message to the locomotive driver's cab display terminal and / or the ground data analysis server.

7. A vehicle-mounted safety monitoring device for heavy-duty trucks, characterized in that, The device includes: The wheel-rail force monitoring module is used to acquire the raw monitoring data of wheel-rail force and perform edge calculations on the raw monitoring data of wheel-rail force to obtain the wheel load reduction rate, derailment coefficient and wheel axle lateral force; Vibration acceleration sensing module, used to acquire the acceleration of the structure; Bearing temperature sensing module, used to obtain bearing temperature; The parameter fusion module is used to perform multimodal fusion of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature to obtain a multimodal operating state characterization vector; The dynamic threshold acquisition module is used to obtain the dynamic thresholds of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration and the bearing temperature based on the enhanced feature vector obtained by the offline self-learning model and the multimodal operating state characterization vector. The safety monitoring and analysis module is used to provide a current status analysis and trend analysis of safety monitoring based on the comparison results of the wheel load reduction rate, the derailment coefficient, the wheel axle lateral force, the frame acceleration, the bearing temperature and the dynamic threshold.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 6.

Images