Fault Monitoring Method for Transmission Devices of Wheeled Transport Platforms Based on Multi-Source Data Fusion under Complex Road Conditions

Abstract

This invention relates to the field of wheeled transport platform technology, and particularly to a fault monitoring method for the transmission device of a wheeled transport platform based on multi-source data fusion under complex road conditions. The method includes: determining the bump level of the road surface on which the wheeled transport platform travels; determining a deep learning diagnostic model based on the bump level; collecting fault signals of the transmission device through sensors; preprocessing the collected fault signals; and completing the fault monitoring of the wheeled transport platform's transmission device. This invention can more comprehensively capture changes in the operating status of the transmission device, promptly detect potential faults, reduce downtime caused by faults, and improve equipment utilization efficiency. Furthermore, it addresses the problem of abnormal vibration frequencies caused by complex road conditions, especially muddy and potholed sections, for wheeled transport platforms. This patent detects and compares road conditions to determine the road condition, thereby switching to the appropriate model for fault prediction and improving the accuracy of fault prediction.

Description

Technical Field

[0001] This invention relates to the field of wheeled transport platform technology, and in particular to a method for fault monitoring of the transmission device of a wheeled transport platform based on multi-source data fusion under complex road conditions. Background Technology

[0002] Wheeled transport platforms are a major combat asset of modern armies, integrating protection, mobility, and firepower. They possess powerful off-road mobility, protective capabilities, and firepower systems. Mobility is the most fundamental performance characteristic of wheeled transport platforms, and the power transmission system is the core of this mobility. Wheeled transport platforms typically operate in extremely harsh environments, and their power transmission systems bear a wide range of loads; for example, the resistance experienced in complex terrain is approximately 10 to 15 times that on flat terrain.

[0003] The transmission system of a wheeled transport platform consists of multiple complex components, including flexible couplings, a transmission box, a main clutch, a gearbox, a fan drive, and a planetary steering gear. These characteristics require the power transmission system of a specialized wheeled transport platform to withstand greater load impacts than that of a typical power transmission system. Fault data analysis shows that the most frequent failures in wheeled transport platforms are in the power unit, followed by electrical control, fire control, steering, and transmission systems. Failures in the power and transmission units account for more than half of all vehicle failures, approximately 51.21%. With the advent of the information age, equipment is becoming increasingly larger, more powerful, and more complex. The transmission system of a wheeled transport platform is no exception, and its complexity makes fault diagnosis more difficult.

[0004] Current research on the diagnosis of wheeled transport platforms in my country mainly focuses on individual components under stable operating conditions, with limited research on the overall transmission system under complex operating conditions and little attention paid to the failure mechanisms of the transmission system. Consequently, the accuracy and reliability are not ideal. Traditional methods for monitoring transmission system failures in wheeled transport platforms often rely on periodic manual inspections or analysis of single sensor data. These methods suffer from long monitoring cycles, slow response times, low fault identification rates, and difficulties in monitoring, making them unsuitable for meeting the requirements of rapid response and high reliability in modern warfare. Summary of the Invention

[0005] This invention discloses a fault monitoring method for the transmission device of a wheeled transport platform based on multi-source data fusion under complex road conditions. The specific method is as follows:

[0006] Determine the bump level of the road surface on which the wheeled transport platform travels;

[0007] Determine the deep learning diagnostic model based on the level of road bumpiness;

[0008] Fault signals of the transmission device are collected using sensors;

[0009] Preprocess the collected fault signals;

[0010] The preprocessed fault signals are input into a defined deep learning diagnostic model to complete the fault monitoring of the transmission device of the wheeled transport platform.

[0011] Furthermore, the bumpiness level of the road surface on which the wheeled transport platform travels is determined using the following method:

[0012] The distance between the road surface and the bottom of the wheeled transport platform is measured in real time using a laser rangefinder sensor, generating a distance data sequence.

[0013] Set up a sliding window, calculate the variance of the distance data sequence for each sliding window, and construct a variance value sequence;

[0014] Smooth the variance sequence;

[0015] Determine the bump level of the road surface based on the variance value sequence.

[0016] Furthermore, the bumpiness level of the driving road surface is divided into five levels, namely, smooth level, level one bumpiness, level two bumpiness, level three bumpiness, and abnormal bumpiness.

[0017] Furthermore, the fault signals include: transmission device vibration signals, transmission device operating temperature signals, transmission device operating noise signals, and external wear image data of the transmission device.

[0018] Furthermore, the preprocessing methods for the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal are as follows:

[0019] Exclude outliers, duplicates, and missing values ​​from the statistical characteristics of the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal;

[0020] Smoothing is performed on the vibration signal, operating temperature signal, and operating noise signal of the transmission device.

[0021] Statistical features of the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal are extracted. The statistical features include mean features, variance features, and peak value features.

[0022] Synchronize the extracted statistical features over time.

[0023] Furthermore, the vibration signal, operating temperature signal, and operating noise signal of the transmission device are smoothed. The specific methods are as follows:

[0024] Determine the range of the smoothing coefficient, which is between 0 and 1. The larger the smoothing coefficient, the closer the smoothed data is to the original data; the smaller the smoothing coefficient, the smoother the data.

[0025] Within the range of smoothing coefficients, a candidate set is defined;

[0026] For each smoothing coefficient in the candidate set, the smoothing exponent is predicted using the training set data, and the mean squared error and mean absolute error of the prediction results on the test set are calculated.

[0027] The optimal smoothing coefficient is selected based on the mean square error and the mean absolute error.

[0028] Furthermore, the preprocessing method for the external wear image data of the transmission device is as follows:

[0029] The external wear image data of the transmission device was adjusted to a fixed size using interpolation.

[0030] Image feature extraction is performed using a convolutional neural network.

[0031] Furthermore, there are three deep learning diagnostic models;

[0032] The first deep learning model is used for diagnosis of smooth driving road surfaces;

[0033] The second deep learning model is used for road surface diagnosis of Level 1, Level 2, and Level 3 bumpy driving conditions.

[0034] The third deep learning model is used for diagnosing abnormally bumpy road surfaces.

[0035] Furthermore, deep learning diagnostic models include an input layer, a hidden layer, and an output layer;

[0036] The input layers of the first and second deep learning models take the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal as inputs; the third deep learning model takes the transmission device operating temperature signal, transmission device operating noise signal, and external wear image data of the transmission device as inputs; the outputs of the deep learning diagnostic models are all probability predictions of fault types.

[0037] Due to the adoption of the above technical solutions, the present invention has the following beneficial effects:

[0038] 1. This invention can more comprehensively capture changes in the operating status of the transmission device, promptly detect potential faults, reduce downtime caused by faults, and improve the efficiency of equipment use.

[0039] 2. This patent addresses the issue of abnormal vibration frequencies caused by complex road conditions, especially muddy and potholed sections, in wheeled transport platforms. It detects and compares road conditions to determine the appropriate model for fault prediction, thereby improving the accuracy of fault prediction.

[0040] 3. By utilizing advanced predictive model algorithms, this invention can not only monitor current faults, but also predict future faults based on historical and real-time data, thereby achieving predictive maintenance, avoiding unplanned downtime, and reducing maintenance costs.

[0041] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0042] The accompanying drawings of this invention are described below.

[0043] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0044] Figure 2 This is a schematic diagram of the deep learning diagnostic model process. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] A fault monitoring method for the transmission device of a wheeled transport platform based on multi-source data fusion under complex road conditions, such as... Figure 1 As shown, the specific steps are as follows:

[0047] S1. Determine the bump level of the road surface on which the wheeled transport platform travels.

[0048] Specifically, the method for determining the bumpiness level of the road surface on which the wheeled transport platform travels is as follows:

[0049] S11. The distance between the road surface and the bottom of the wheeled transport platform is measured in real time using a laser rangefinder sensor, generating a distance data sequence;

[0050] S12. Set up a sliding window, calculate the variance of the distance data sequence for each sliding window, and construct a variance value sequence;

[0051] S13. Smooth the variance sequence;

[0052] S14. Determine the bump level of the road surface based on the variance value sequence.

[0053] S2. Determine the deep learning diagnostic model based on the road surface bump level.

[0054] In this embodiment, the bumpiness level of the driving road surface is divided into five levels: smooth, level 1 bumpiness, level 2 bumpiness, level 3 bumpiness, and abnormal bumpiness. Three deep learning diagnostic models are used: the first deep learning model is used for diagnosing smooth driving road surfaces; the second deep learning model is used for diagnosing level 1, level 2, and level 3 bumpy driving road surfaces; and the third deep learning model is used for diagnosing abnormal bumpy driving road surfaces.

[0055] S3. Collect fault signals of the transmission device through sensors.

[0056] In this embodiment, the fault signals include: transmission device vibration signal, transmission device operating temperature signal, transmission device operating noise signal, and external wear image data of the transmission device.

[0057] Vibration signals from transmission devices can reflect internal faults such as wear, loosening, and imbalance.

[0058] Temperature signals from the transmission system, especially from easily worn parts such as bearings and gears. An abnormally high temperature may indicate malfunctions such as poor lubrication, overload, or excessive wear.

[0059] Noise signals from the transmission system, especially cylinder combustion noise and gear meshing noise. Noise data can reflect abnormal friction, impact, and other faults within the transmission system.

[0060] S4. Preprocess the collected fault signals.

[0061] The specific preprocessing methods for transmission device vibration signals, transmission device operating temperature signals, and transmission device operating noise signals are as follows:

[0062] S41. Remove outliers, duplicates, and missing values ​​from the statistical characteristics of the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal.

[0063] S42. Smooth the vibration signal, operating temperature signal, and operating noise signal of the transmission device.

[0064] In step S42, smoothing is performed, and the specific method is as follows:

[0065] S421. Determine the range of the smoothing coefficient. The smoothing coefficient ranges from 0 to 1. The larger the smoothing coefficient, the closer the smoothed data is to the original data. The smaller the smoothing coefficient, the smoother the data is.

[0066] S422. Within the range of smoothing coefficients, set a candidate set;

[0067] S423. For each smoothing coefficient in the candidate set, use the training set data to predict the smoothing exponent, and calculate the mean square error and mean absolute error of the prediction results on the test set.

[0068] S424. Select the optimal smoothing coefficient based on the mean square error and the mean absolute error.

[0069] In this embodiment, the evaluation index for the optimal smoothing coefficient is specifically formulated as follows:

[0070] S t =α.X t +(1-α).S t-1

[0071] Among them, X t S is the original data value at time t. t-1 It is the smoothed value at time t-1, and α is the smoothing coefficient.

[0072] S43. Extract the statistical characteristics of the transmission device vibration signal, transmission device operating temperature signal and transmission device operating noise signal. The statistical characteristics include mean characteristics, variance characteristics and peak characteristics.

[0073] S44. Synchronize the extracted statistical features over time.

[0074] In this embodiment, the external wear image data of the transmission device is adjusted to a fixed size by interpolation; then, image features are extracted by a convolutional neural network. The external wear image data processing model of the transmission device includes: an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.

[0075] Input layer: Receives preprocessed image data.

[0076] Convolutional layer 1: Uses multiple convolutional kernels to perform convolution operations and extract low-level features of the image.

[0077] Pooling layer 1: The output of convolutional layer 1 is pooled to reduce the size of the feature map.

[0078] Convolutional layer 2: Uses more convolutional kernels to perform convolution operations and extracts intermediate features of the image.

[0079] Pooling layer 2: Performs pooling operation on the output of convolutional layer 2.

[0080] Convolutional layer 3 and pooling layer 3: further extract high-level features of the image.

[0081] Fully connected layer: Maps the features extracted by the convolutional layer to the output space to obtain the feature vector of the image.

[0082] S5. Input the preprocessed fault signal into the determined deep learning diagnostic model to complete the fault monitoring of the transmission device of the wheeled transport platform.

[0083] In this embodiment, the workflow of the deep learning diagnostic model is as follows: Figure 2 As shown, a deep learning model includes an input layer, a hidden layer, and an output layer;

[0084] The input layers of the first and second deep learning models take the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal as inputs; the third deep learning model takes the transmission device operating temperature signal, transmission device operating noise signal, and external wear image data of the transmission device as inputs; the outputs of the deep learning diagnostic models are all probability predictions of fault types.

[0085] In this embodiment, the hidden layer is designed as follows:

[0086] First hidden layer

[0087] Number of neurons: Determined based on the dimension of the input features and the complexity of the model, and is usually set to a multiple of the dimension of the input features.

[0088] Activation function: ReLU function, which adds non-linearity.

[0089] Regularization: L1 regularization can be introduced to prevent overfitting.

[0090] Second hidden layer

[0091] Number of neurons: fewer than in the first layer, to progressively extract high-level features.

[0092] Activation function: ReLU function.

[0093] Dropout layer: Set the dropout ratio to 0.5 to prevent overfitting.

[0094] Third hidden layer

[0095] Activation function: ReLU function

[0096] Output layer: Use the softmax function to output the probability distribution of fault types.

[0097] The output layer predicts the failure probability of the transmission device.

[0098] The output layer is a regression layer.

[0099] The fused data is output in a unified format to facilitate subsequent data analysis and fault prediction. The trained model is then deployed to the real-time monitoring system. Through these steps, a deep learning-based fault prediction model for the transmission system of wheeled transport platforms can be constructed. This model can monitor the failure probability of the transmission system in real time and accurately, providing strong support for the maintenance and upkeep of wheeled transport platforms.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for fault monitoring of the transmission device of a wheeled transport platform based on multi-source data fusion under complex road conditions, characterized in that, The specific method is as follows: Determine the bump level of the road surface on which the wheeled transport platform travels; Determine the deep learning diagnostic model based on the level of road bumpiness; Fault signals of the transmission device are collected using sensors; Preprocess the collected fault signals; The preprocessed fault signals are input into a defined deep learning diagnostic model to complete the fault monitoring of the transmission device of the wheeled transport platform. The road surface is classified into five levels of bumpiness: smooth, level 1 bumpiness, level 2 bumpiness, level 3 bumpiness, and abnormal bumpiness. The fault signals include: transmission device vibration signal, transmission device operating temperature signal, transmission device operating noise signal, and external wear image data of the transmission device; The preprocessing method for external wear image data of the transmission device is as follows: The external wear image data of the transmission device was adjusted to a fixed size using interpolation. Image feature extraction using convolutional neural networks; There are three deep learning diagnostic models; The first deep learning model is used for diagnosis of smooth driving road surfaces; The second deep learning model is used for road surface diagnosis of Level 1, Level 2, and Level 3 bumpy driving conditions. The third deep learning model is used for diagnosing abnormally bumpy road surfaces. Deep learning diagnostic models consist of an input layer, hidden layers, and an output layer. The input layers of the first and second deep learning models take the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal as inputs; the third deep learning model takes the transmission device operating temperature signal, transmission device operating noise signal, and external wear image data of the transmission device as inputs; the outputs of the deep learning diagnostic models are all probability predictions of fault types.

2. The method for fault monitoring of wheeled transport platform transmission device based on multi-source data fusion under complex road conditions as described in claim 1, characterized in that, The specific method for determining the bumpiness level of the road surface on which a wheeled transport platform travels is as follows: The distance between the road surface and the bottom of the wheeled transport platform is measured in real time using a laser rangefinder sensor, generating a distance data sequence. Set up a sliding window, calculate the variance of the distance data sequence for each sliding window, and construct a variance value sequence; Smooth the variance sequence; Determine the bump level of the road surface based on the variance value sequence.

3. The method for fault monitoring of wheeled transport platform transmission device based on multi-source data fusion under complex road conditions as described in claim 1, characterized in that, The preprocessing methods for transmission device vibration signals, transmission device operating temperature signals, and transmission device operating noise signals are as follows: Exclude outliers, duplicates, and missing values ​​from the statistical characteristics of the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal; Smoothing is performed on the vibration signal, operating temperature signal, and operating noise signal of the transmission device. Statistical features of the transmission device vibration signal, transmission device operating temperature signal, and transmission device operating noise signal are extracted. The statistical features include mean features, variance features, and peak value features. Synchronize the extracted statistical features over time.

4. The method for fault monitoring of wheeled transport platform transmission device based on multi-source data fusion under complex road conditions as described in claim 3, characterized in that, The vibration signal, operating temperature signal, and operating noise signal of the transmission device are smoothed using the following methods: Determine the range of the smoothing coefficient, which is between 0 and 1. The larger the smoothing coefficient, the closer the smoothed data is to the original data; the smaller the smoothing coefficient, the smoother the data. Within the range of smoothing coefficients, a candidate set is defined; For each smoothing coefficient in the candidate set, the smoothing exponent is predicted using the training set data, and the mean squared error and mean absolute error of the prediction results on the test set are calculated. The optimal smoothing coefficient is selected based on the mean square error and the mean absolute error.

Images