Method and system for remotely diagnosing power system fault of tractor

Abstract

The invention relates to the technical field of tractor power system fault diagnosis, and discloses a method and system for remotely diagnosing a power system fault of a tractor, and the method comprises the steps: monitoring a vibration signal, a sound signal and a temperature signal of a transmission gear of the tractor; extracting a frequency domain feature, a Mel frequency cepstrum coefficient feature and a temperature change trend feature of the vibration signal; carrying out weighted fusion on the signal features by adopting a dynamic weighting mechanism; constructing a fault prediction model based on an XGBoost algorithm, and outputting the gear wear degree and the fault occurrence time; and generating a fault early warning and transmitting the fault early warning to a mobile terminal APP and a cloud platform through an LTE or 5G network. Compared with a traditional fault diagnosis method in the prior art, particularly under the condition of dynamic working conditions, the technical problem that efficient and accurate fault diagnosis of the tractor power system is difficult to realize is solved, and the accuracy and response efficiency of fault diagnosis are improved by combining the convolutional neural network and the XGBoost algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fault diagnosis of power systems of tractors, and in particular to a method and system for remotely diagnosing faults of power systems of tractors. BACKGROUND

[0002] Currently, the fault diagnosis technology of power systems of tractors is facing a series of challenges. Traditional fault diagnosis methods mainly rely on manual feature extraction, empirical rules or simple models. These methods usually cannot fully capture the complex information in the signals, and are very sensitive to noise, environmental changes and working condition fluctuations. For example, under dynamic working conditions such as high-speed driving, load changes or long-time running, traditional methods are difficult to adapt to the rapid changes of these signals, which can easily lead to a decrease in diagnostic accuracy, thereby affecting the reliability and safety of the system. In addition, existing technologies rely on the monitoring of a single signal source (such as vibration or temperature), ignoring the importance of multi-modal data fusion. This single signal processing method cannot provide comprehensive diagnostic results when facing complex fault patterns. For example, vibration signals may not accurately reflect slight gear wear, while temperature and sound signals may need to be analyzed in combination to accurately determine the type and severity of the fault. Therefore, there is an urgent need for a method that can achieve efficient and accurate fault diagnosis under complex dynamic working conditions and multi-signal fusion. Especially in intelligent manufacturing equipment such as tractors, the fault diagnosis of the power system needs to be monitored in real time and warned in advance to avoid major safety accidents caused by faults. SUMMARY

[0003] In view of the above technical deficiencies, the purpose of the present application is to provide a method for remotely diagnosing faults of power systems of tractors, which aims to solve the technical problem that traditional fault diagnosis methods, especially under dynamic working conditions, are difficult to achieve efficient and accurate fault diagnosis of power systems of tractors.

[0004] To solve the above technical problems, the present application adopts the following technical solution: The present application provides a method for remotely diagnosing faults of power systems of tractors,

[0005] The method for remotely diagnosing faults of power systems of tractors comprises:

[0006] Step S10: The physical signals of the transmission gear in the power system of the tractor at time t are monitored by acceleration sensors, sound sensors and temperature sensors respectively to obtain first vibration signals, first sound signals and first temperature signals;

[0007] Step S20: The first vibration signals, first sound signals and first temperature signals of the transmission gear are first standardized, and vibration signal frequency domain features f v(t), a mel-frequency cepstral coefficient feature f s (t) and a temperature change trend feature f T (t);

[0008] Step S30: using a dynamic weighting mechanism to weight and fuse the vibration signal frequency domain features f v (t), a mel-frequency cepstral coefficient feature f s (t) and a temperature change trend feature f T (t) to generate a comprehensive feature vector f(t); using a pre-set convolutional neural network to perform time domain feature extraction on the comprehensive feature vector f(t) to obtain a gear comprehensive feature vector;

[0009] Step S40: constructing a transmission gear fault prediction model based on an XGBoost algorithm, inputting the gear comprehensive feature vector into the pre-trained transmission gear fault prediction model, and outputting a transmission gear fault prediction result containing a gear wear degree and a fault occurrence time;

[0010] Step S50: when the gear wear degree is greater than a pre-set gear wear threshold, automatically generating a fault warning information and transmitting it to a mobile terminal APP and a cloud monitoring platform through an LTE network or a 5G network.

[0011] Preferably, in step S30, f(t) = w v (t)·f v (t) + w s (t)·f s (t) + w T (t)·f t (t), wherein w v (t), w s (t) and w T (t) are weight coefficients dynamically adjusted according to different working conditions of the automobile.

[0012] Preferably, in step S20, the standardization processing includes noise removal processing, abnormal value filtering processing and normalization processing; wherein the noise removal processing uses a low-pass filtering method; the abnormal value filtering processing uses a Z-score abnormal value detection method to identify and process abnormal values, and fills or deletes the abnormal values through a linear interpolation method; the normalization processing uses a minimum-maximum normalization method.

[0013] Preferably, in step S20, the first vibration signal, the first sound signal and the first temperature signal of the transmission gear are first subjected to standardization processing, and features are obtained respectively to obtain the vibration signal frequency domain feature f v (t), the mel-frequency cepstral coefficient feature f s (t) and the temperature change trend feature f TThe step (t) specifically comprises: first, normalizing the first vibration signal, the first sound signal and the first temperature signal of the transmission gear to obtain a second vibration signal, a second sound signal and a second temperature signal; and then, extracting the second vibration signal by fast Fourier transform (FFT) to obtain a vibration signal frequency domain feature f v (t); extracting the second sound signal by mel-frequency cepstral coefficient (MFCC) to obtain a mel-frequency cepstral coefficient feature f s (t); and calculating the second temperature signal by regression analysis to obtain a temperature change trend feature f T (t).

[0014] Preferably, in the step S30, a preset convolutional neural network is used to extract the time domain features of the comprehensive feature vector f(t) to obtain a gear comprehensive feature vector, and the step specifically comprises: presetting the convolutional neural network, wherein the convolutional neural network comprises an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a full connection layer and an output layer.

[0015] Preferably, in the step S30, the input layer is used to receive the comprehensive feature vector f(t) in the step S20.

[0016] The first convolutional layer is used to extract basic time domain features from the input layer, the size of the convolution kernel of the first convolutional layer is set according to the characteristics of the comprehensive feature vector f(t), and the comprehensive feature vector f(t) is processed by a sliding window to detect local signal patterns, including gear vibration signal patterns and sound change signal patterns; and the first convolutional layer outputs a gear feature set.

[0017] The first pooling layer is used to down-sample the output of the first convolutional layer to reduce the dimension of the features and retain important feature information, a maximum pooling method is used, and a 2×2 pooling window is used to pool the gear feature set output by the first convolutional layer to output a pooled gear feature set.

[0018] The second convolutional layer is used to extract a more complex gear time domain feature set from the gear feature set of the first convolutional layer to capture complex vibration pattern features, sound features and temperature change features generated in the gear wear process.

[0019] The second pooling layer is used to down-sample the output of the second convolutional layer, a maximum pooling method is used, and a 2×2 pooling window is used to pool the complex vibration pattern features, sound features and temperature change features output by the second convolutional layer to output pooled complex vibration pattern features, sound features and temperature change features.

[0020] The full connection layer is used to integrate the features after the convolution and pooling layers extract the features, and a gear comprehensive feature vector is generated by a nonlinear activation function ReLU.

[0021] Output layer: used for outputting the gear comprehensive feature vector.

[0022] Preferably, in step S40, the pre-training process of the transmission gear fault prediction model specifically comprises: obtaining a historical data set, training the transmission gear fault prediction model based on the historical data set using an XGBoost algorithm, including: training set division: dividing the historical data set into 80% training set and 20% validation set; hyperparameter tuning: tuning the hyperparameters of the XGBoost algorithm through grid search or random search, including the depth of the XGBoost algorithm tree, the learning rate, and the minimum sample split number; cross-validation: using k-fold cross-validation to evaluate the performance score of the transmission gear fault prediction model.

[0023] The application also provides a system for remotely diagnosing faults of a power system of a tractor, comprising:

[0024] A sensor data acquisition module is configured to monitor physical signals of a transmission gear in the power system of the tractor at time t through an acceleration sensor, a sound sensor, and a temperature sensor, to obtain a first vibration signal, a first sound signal, and a first temperature signal.

[0025] A signal processing and feature extraction module is configured to perform standardization processing on the first vibration signal, the first sound signal, and the first temperature signal of the transmission gear, and to extract features to obtain vibration signal frequency domain features f v (t), mel-frequency cepstral coefficient features f s (t), and temperature change trend features f T (t).

[0026] A feature weighting fusion and feature extraction module is configured to perform weighted fusion on the vibration signal frequency domain features f v (t), the mel-frequency cepstral coefficient features f s (t), and the temperature change trend features f T (t) using a dynamic weighting mechanism to generate a comprehensive feature vector f(t); and to perform time domain feature extraction on the comprehensive feature vector f(t) using a preset convolutional neural network to obtain a gear comprehensive feature vector.

[0027] A transmission gear fault prediction module is configured to construct a transmission gear fault prediction model based on an XGBoost algorithm, to input the gear comprehensive feature vector into the pre-trained transmission gear fault prediction model, and to output a transmission gear fault prediction result containing a gear wear degree and a time of fault occurrence.

[0028] The fault early warning information generation and transmission module is configured to automatically generate fault early warning information and transmit the information to a mobile terminal APP and a cloud monitoring platform through an LTE network or a 5G network when the gear wear degree is greater than a preset gear wear threshold.

[0029] The application further provides a computer program product comprising a program for remotely diagnosing a fault of a power system of a tractor, and the program for remotely diagnosing a fault of a power system of a tractor is executed by a processor to implement the method for remotely diagnosing a fault of a power system of a tractor.

[0030] The application has the beneficial effects that the application can effectively improve the accuracy and response efficiency of fault diagnosis of a power system of a tractor by combining a convolutional neural network and an XGBoost algorithm, and can accurately identify the wear degree of a gear of a transmission and predict the time of fault occurrence, especially under dynamic working conditions, thereby improving the intelligent level of fault diagnosis.

[0031] The application provides an efficient and accurate fault diagnosis method for the intelligent manufacturing equipment industry, can realize real-time monitoring and fault early warning, and promotes the technological innovation and development of the intelligent manufacturing equipment industry. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0033] Figure 1 The flowchart of the first embodiment of the method for remotely diagnosing a fault of a power system of a tractor.

[0034] Figure 2 The device schematic diagram of the method for remotely diagnosing a fault of a power system of a tractor. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0036] Embodiment one: as Figure 1As shown, it is a flow chart of the first embodiment of the method for remotely diagnosing the power system fault of the tractor of the present application, and the first embodiment of the method for remotely diagnosing the power system fault of the tractor of the present application is proposed.

[0037] In the first embodiment, the method for remotely diagnosing the power system fault of the tractor comprises:

[0038] Step S10: The physical signals of the transmission gear in the power system of the tractor at time t are monitored by the acceleration sensor, the sound sensor and the temperature sensor respectively to obtain the first vibration signal, the first sound signal and the first temperature signal;

[0039] It should be noted that the acceleration sensor is installed at the key position of the transmission gear for real-time monitoring of the vibration signal of the gear. The acceleration sensor can capture the vibration mode of the gear under different working conditions, including the vibration caused by gear wear, damage or load change. The sound sensor is used to monitor the noise change of the gear during operation, especially when the gear wears and fails, the friction between the gears will produce a noise mode different from normal operation. The temperature sensor is installed near the gear to monitor the temperature change of the gear in real time. Since the gear generates more friction heat when it wears, the temperature sensor can capture the rising trend of the temperature. By monitoring the temperature change, combined with other signals, the wear degree of the gear can be effectively identified.

[0040] It can be understood that it is necessary to ensure that the data of different sensors are collected synchronously at the same time and calibrated to eliminate the time deviation between different sensors. For this purpose, the time stamp synchronization technology is adopted to ensure that the collected data can accurately correspond to the physical state at the same time, avoiding errors caused by synchronization problems. Through the cooperative work of these sensors, the present application can comprehensively monitor the state of the transmission gear, providing accurate raw data support for subsequent fault diagnosis and early warning.

[0041] Step S20: The first vibration signal, the first sound signal and the first temperature signal of the transmission gear are first standardized, and the features are obtained respectively to obtain the vibration signal frequency domain feature f v (t), the mel frequency cepstral coefficient feature f s (t) and the temperature change trend feature f T (t);

[0042] It should be noted that standardization not only helps to eliminate the dimensional difference between different sensors, but also improves the accuracy and robustness of the subsequent processing steps. Especially in the machine learning algorithm, standardization can ensure the comparability between feature values, thereby avoiding the influence of a certain feature on the model being too large, leading to bias.

[0043] It can be understood that the noise and unnecessary fluctuations in the standardized signal are removed, and the remaining signal will be more representative. Through further feature extraction, such as frequency domain features, Mel-frequency cepstral coefficient (MFCC) features, and temperature change trend features, the potential regularities in the signal can be further analyzed to capture details related to gear wear. Each feature extraction method has its unique advantages and can provide important information for subsequent fault prediction. For example, the frequency domain features extracted from the vibration signal through Fast Fourier Transform (FFT) can help identify the frequency changes of the gear during the wear stage, such as the vibration pattern or abnormal frequency components of the gear. The MFCC features of the sound signal can capture the sound changes caused by wear, especially high-frequency noise and subtle differences in the signal. The temperature signal change trend is extracted through regression analysis or difference method, revealing the heat change due to friction, especially during the temperature rise caused by gear wear, which can capture subtle temperature fluctuations.

[0044] Step S30: using a dynamic weighting mechanism to weight and fuse the vibration signal frequency domain features f v (t), Mel-frequency cepstral coefficient features f s (t), and temperature change trend features f T (t) to generate a comprehensive feature vector f(t); using a pre-set convolutional neural network to extract time domain features from the comprehensive feature vector f(t) to obtain a gear comprehensive feature vector;

[0045] It should be noted that since the vibration signal, sound signal, and temperature signal may provide different information and importance in gear wear diagnosis, a dynamic weighting mechanism is used for feature weighting and fusion to ensure that different signal features can be reasonably weighted according to their actual importance in the current working condition.

[0046] It can be understood that through weighted fusion, the signal features collected by different sensors (such as vibration, sound, and temperature) have different roles in gear wear analysis. After weighted fusion, the model can effectively focus on the key features that affect wear diagnosis, improving the diagnosis accuracy. For example, during high-speed driving, the vibration signal may be more decisive, while during low-speed operation or long-time working condition, the temperature signal may better reflect the wear condition of the gear.

[0047] It should be understood that after the weighted fusion of the signals is completed, the comprehensive feature vector will be input into a convolutional neural network (CNN) for further time domain feature extraction. The CNN effectively extracts the local time domain features of the signals and their inherent laws through multiple layers of convolution and pooling operations. The advantage of this process is that valuable features can be extracted from complex signals through automatic feature learning without relying on manually designed features. After CNN processing, the obtained gear comprehensive feature vector contains high-level feature representations of the multi-modal signals, providing deep and accurate data support for subsequent fault prediction.

[0048] Step S40: Based on the XGBoost algorithm, a gearbox gear fault prediction model is constructed, the gear comprehensive feature vector is input into the pre-trained gearbox gear fault prediction model, and the gearbox gear fault prediction result containing the gear wear degree and the time of fault occurrence is output.

[0049] It should be noted that XGBoost (Extreme Gradient Boosting) is an ensemble learning algorithm based on gradient boosting trees (GBDT), widely used in regression and classification problems, and can effectively handle complex nonlinear relationships and high-dimensional feature data. In the present application, the XGBoost algorithm is used to construct a gearbox gear fault prediction model, which predicts the wear degree and the time of fault occurrence by inputting the gear comprehensive feature vector.

[0050] It can be understood that XGBoost makes predictions through the integration of multiple decision trees, and this integration method can capture complex patterns and diverse features in the data. Compared with traditional single decision tree models, XGBoost can continuously optimize the error of each tree through gradient boosting techniques, making the model more robust and accurate in prediction. It has strong advantages in handling high-dimensional features, nonlinear relationships, and missing data, and is suitable for the multi-modal signal data in the present application. The output of the XGBoost model can be divided into regression tasks and classification tasks according to the nature of the task. In the present application, if the goal is to predict the wear degree of the gear, the model output is a continuous value (such as wear percentage); if the goal is to predict the time of fault occurrence, the output is a specific time point (such as hours or days). If the task involves multiple categories of fault types, XGBoost can also be used as a multi-classification model.

[0051] Step S50: When the gear wear degree is greater than the preset gear wear threshold, automatically generate fault warning information and transmit it to the mobile terminal APP and the cloud monitoring platform through the LTE network or the 5G network.

[0052] It should be noted that when the wear degree exceeds the threshold value, a fault warning information containing the following contents will be generated: wear degree: indicating the current wear condition of the gear (such as percentage), providing data support for subsequent decision-making; predicted failure occurrence time: according to the failure prediction model, estimating the time point of failure occurrence, facilitating early preparation and maintenance arrangement; fault type or severity. The application of LTE or 5G network enables the real-time transmission of fault warning information to mobile APP and cloud monitoring platform. This not only improves the speed and stability of information transmission, but also ensures that maintenance personnel and monitoring platform can obtain the latest fault status in the shortest time and respond in time.

[0053] It can be understood that when the fault warning information is transmitted to the mobile APP or the cloud monitoring platform, the user or the operation and maintenance personnel can real-time view the running state and potential failure risk of the tractor and make corresponding maintenance or repair decisions. This network-based remote monitoring and diagnosis system can greatly improve the running safety of the tractor, especially in the absence of professional technical support, through mobile devices for remote fault monitoring and management, reducing the risk of accidents.

[0054] Embodiment two: In addition, the system for remotely diagnosing the failure of the power system of the tractor provided by the present application adopts the method for remotely diagnosing the failure of the power system of the tractor in the above embodiment, which can solve the technical problem of remotely diagnosing the failure of the power system of the tractor. Compared with the prior art, the beneficial effects of the system for remotely diagnosing the failure of the power system of the tractor provided by the present application are the same as those of the method for remotely diagnosing the failure of the power system of the tractor provided by the above embodiment, and other technical features of the system for remotely diagnosing the failure of the power system of the tractor are the same as those disclosed in the above embodiment method, which will not be repeated here.

[0055] Embodiment three: The present application provides a device for remotely diagnosing the failure of the power system of the tractor, please refer to Figure 2An apparatus for remotely diagnosing a failure of a powertrain of a tractor includes at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a method for remotely diagnosing a failure of a powertrain of a tractor according to the embodiment 1. The apparatus for remotely diagnosing a failure of a powertrain of a tractor according to the embodiment of the present application can include, but is not limited to, a mobile terminal such as a mobile phone, a notebook, a digital broadcasting receiver, a PDA (Personal Digital Assistant), a PAD (Portable Application Description), a PMP (Portable Media Player), a vehicle terminal (e.g., a vehicle navigation terminal), and the like, and a stationary terminal such as a digital TV, a desktop computer, and the like. The apparatus for remotely diagnosing a failure of a powertrain of a tractor is only one example and should not limit the function and the use range of the embodiment of the present application. The apparatus for remotely diagnosing a failure of a powertrain of a tractor can include a processing device 1001 (e.g., a central processing unit, a graphic processing unit, and the like) which can perform various appropriate actions and processes according to a program stored in a ROM (Read Only Memory) 1002 or a program loaded from a storage device 1003 to a RAM (Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the apparatus for remotely diagnosing a failure of a powertrain of a tractor are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An I / O (Input / Output) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, and the like; an output device 1008 including, for example, an LCD (Liquid Crystal Display), a speaker, a vibrator, and the like; the storage device 1003 including, for example, a magnetic tape, a hard disk, and the like; and a communication device 1009. The communication device 1009 can allow the apparatus for remotely diagnosing a failure of a powertrain of a tractor to communicate with other devices wirelessly or wiredly to exchange data. Although the apparatus for remotely diagnosing a failure of a powertrain of a tractor having various systems is illustrated, it is understood that all of the illustrated systems are not required to be implemented or provided. More or less systems can be alternatively implemented or provided.

[0056] Embodiment four: the application further provides a computer program product comprising a computer program, which, when executed by a processor, implements the steps of a method for remotely diagnosing a power system fault of a tractor as described above. The computer program product provided by the application can solve the technical problem of remotely diagnosing a power system fault of a tractor. Compared with the prior art, the beneficial effects of the computer program product provided by the application are the same as those of the method for remotely diagnosing a power system fault of a tractor provided by the above-described embodiments, and are not described here in detail.

[0057] In particular, according to the embodiments disclosed by the application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments disclosed by the application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, the above-mentioned functions defined in the method of the embodiments disclosed by the application are executed.

[0058] It should be understood that various parts of the application disclosed can be realized in hardware, software, firmware or a combination thereof. In the description of the above-described embodiments, specific features, structures, materials or characteristics can be combined in any appropriate manner in any one or more embodiments or examples.

[0059] Obviously, those skilled in the art can make various modifications and variations to the application without departing from the spirit and scope of the application. Thus, if these modifications and variations of the application fall within the scope of the claims of the application and their equivalent technologies, the application also intends to include these modifications and variations.

Claims

1. A method for remotely diagnosing a power system failure of a tractor, characterized in that: Methods include: Step S10: monitoring the physical signals of the transmission gear in the power system of the tractor at time t by using an acceleration sensor, a sound sensor, and a temperature sensor, respectively, to obtain a first vibration signal, a first sound signal, and a first temperature signal; Step S20: The first vibration signal, the first sound signal and the first temperature signal of the transmission gear are first normalized and characterized to obtain the frequency domain feature f of the vibration signal. v (t), Mel frequency cepstral coefficient feature f s (t) and temperature change trend characteristics f T (t); Step S30: Use dynamic weighting mechanism to transform the vibration signal frequency domain feature f v (t), Mel frequency cepstral coefficient feature f s (t) and temperature change trend characteristics f T (t) is weightedly fused to generate a comprehensive feature vector f(t); a preset convolutional neural network is used to extract time domain features from the comprehensive feature vector f(t) to obtain a comprehensive feature vector of the gear; Step S40: constructing a transmission gear fault prediction model based on the XGBoost algorithm, inputting the gear comprehensive feature vector into the pre-trained transmission gear fault prediction model, and outputting a transmission gear fault prediction result including the gear wear degree and the time when the fault occurs; Step S50: When the gear wear degree is greater than the preset gear wear threshold, fault warning information is automatically generated and transmitted to the mobile APP and cloud monitoring platform via the LTE network or 5G network.

2. A method for remotely diagnosing a power system failure of a tractor according to claim 1, characterized in that: In step S30, f(t)=w v (t)·f v (t)+w s (t)·f s (t)+w T (t)·f t (t), where w v (t),w s (t),w T (t) is the weight coefficient that is dynamically adjusted according to the different working conditions of the vehicle.

3. The method for remotely diagnosing a power system failure of a tractor according to claim 1, characterized in that: In step S20, the standardization process includes noise removal, outlier filtering and normalization. The noise removal process uses a low-pass filtering method. The outlier filtering process uses a Z-score outlier detection method to identify and process outliers, and fills or deletes outliers through linear interpolation. The normalization process uses a minimum-maximum normalization method.

4. The method for remotely diagnosing a power system failure of a tractor according to claim 1, wherein: In step S20, the first vibration signal, the first sound signal and the first temperature signal of the transmission gear are first standardized, and the characteristics are respectively obtained to obtain the frequency domain characteristics f of the vibration signal. v (t), Mel frequency cepstral coefficient feature f s (t) and temperature change trend characteristics f T The step (t) specifically includes: firstly performing normalization processing on the first vibration signal, the first sound signal and the first temperature signal of the transmission gear to obtain the second vibration signal, the second sound signal and the second temperature signal; and performing fast Fourier transform (FFT) extraction on the second vibration signal to obtain the frequency domain feature f of the vibration signal. v (t); Mel frequency cepstral coefficient MFCC extraction is performed on the second sound signal to obtain the Mel frequency cepstral coefficient feature f s (t); For the second temperature signal, regression analysis is used to calculate the temperature change trend characteristic f T (t).

5. The method for remotely diagnosing a power system failure of a tractor according to claim 1, wherein: In step S30, a preset convolutional neural network is used to perform time domain feature extraction on the comprehensive feature vector f(t) to obtain a comprehensive feature vector of the gear, which specifically includes: presetting a convolutional neural network, the convolutional neural network includes an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a fully connected layer and an output layer.

6. A method for remotely diagnosing a power system failure of a tractor as claimed in claim 5, characterized in that: In step S30, the input layer is used to receive the comprehensive feature vector f(t) in step S20; The first convolutional layer is used to extract basic time-domain features from the input layer. The convolution kernel size of the first convolutional layer is set according to the characteristics of the comprehensive feature vector f(t). The comprehensive feature vector f(t) is processed through a sliding window to detect local signal patterns, including gear vibration signal patterns and sound change signal patterns; and the gear feature set is output. The first pooling layer is used to downsample the output of the first convolutional layer, reduce the dimension of the features and retain important feature information. The maximum pooling method is used to pool the gear feature set output by the first convolutional layer using a 2×2 pooling window, and the pooled gear feature set is output; The second convolutional layer is used to extract a more complex gear time domain feature set from the gear feature set in the first convolutional layer, capturing the complex vibration mode characteristics, sound characteristics, and temperature change characteristics generated during gear wear. The second pooling layer is used to downsample the output of the second convolutional layer. It uses the maximum pooling method and a 2×2 pooling window to pool the complex vibration pattern features, sound features, and temperature change features output by the second convolutional layer, and outputs the pooled complex vibration pattern features, sound features, and temperature change features. Fully connected layer: After the convolution and pooling layers have extracted features, the features will be integrated through the fully connected layer and the gear comprehensive feature vector will be generated through the nonlinear activation function ReLU; Output layer: used to output the comprehensive feature vector of gears.

7. The method for remotely diagnosing a power system failure of a tractor according to claim 1, wherein: In step S40, the pre-training process of the transmission gear fault prediction model specifically includes: obtaining a historical data set, and training the transmission gear fault prediction model based on the historical data set using the XGBoost algorithm, including: training set partitioning: dividing the historical data set into an 80% training set and a 20% validation set; hyperparameter tuning: tuning the hyperparameters of the XGBoost algorithm through grid search or random search, including the depth of the XGBoost algorithm tree, the learning rate, and the minimum number of sample splits; cross-validation: using k-fold cross-validation to evaluate the performance score of the transmission gear fault prediction model.

8. A system for remotely diagnosing a power system failure of a tractor, applied to a method for remotely diagnosing a power system failure of a tractor according to any one of claims 1 to 7, characterized in that: The system for remotely diagnosing power system failures of a tractor includes: a sensor data acquisition module, configured to monitor physical signals of a transmission gear in a power system of the tractor at time t through an acceleration sensor, a sound sensor, and a temperature sensor, respectively, to obtain a first vibration signal, a first sound signal, and a first temperature signal; The signal processing and feature extraction module is used to first perform normalization processing on the first vibration signal, the first sound signal and the first temperature signal of the transmission gear, and perform feature extraction to obtain the frequency domain feature f of the vibration signal. v (t), Mel frequency cepstral coefficient feature f s (t) and temperature change trend characteristics f T (t); The feature weighted fusion and feature extraction module is used to use a dynamic weighting mechanism to transform the vibration signal frequency domain features f v (t), Mel frequency cepstral coefficient feature f s (t) and temperature change trend characteristics f T (t) is weightedly fused to generate a comprehensive feature vector f(t); a preset convolutional neural network is used to extract time domain features from the comprehensive feature vector f(t) to obtain a comprehensive feature vector of the gear; The transmission gear fault prediction module is used to build a transmission gear fault prediction model based on the XGBoost algorithm, input the gear comprehensive feature vector into the pre-trained transmission gear fault prediction model, and output the transmission gear fault prediction result including the gear wear degree and the time when the fault occurs; The fault warning information generation and transmission module is used to automatically generate fault warning information when the gear wear degree exceeds the preset gear wear threshold and transmit it to the mobile APP and cloud monitoring platform via the LTE network or 5G network.

9. A device for remotely diagnosing power system failures of a tractor, characterized in that: The device for remotely diagnosing a power system failure of a tractor includes: a memory, a processor, and a program for remotely diagnosing a power system failure of a tractor stored in the memory and executable on the processor. When the program for remotely diagnosing a power system failure of a tractor is executed by the processor, a method for remotely diagnosing a power system failure of a tractor according to any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that The computer program product includes a program for remotely diagnosing a power system failure of a tractor. When the program for remotely diagnosing a power system failure of a tractor is executed by a processor, the method for remotely diagnosing a power system failure of a tractor according to any one of claims 1 to 7 is implemented.

Images