Gearbox internal fault monitoring method

Abstract

The invention discloses a gearbox internal fault monitoring method, and mainly relates to the field of fault monitoring. Comprising a vibration sensor, a temperature sensor and an acoustic sensor and is used for collecting vibration signals, temperature signals and acoustic signals in the operation process of the gearbox. The system has the beneficial effects that various types of sensors are mounted at multiple key parts of the gear box, various operation data such as vibration, temperature and acoustics are comprehensively acquired in real time, and subtle changes generated by initial faults in the gear box can be sensitively captured by applying advanced signal processing and feature extraction technologies. According to the monitoring method, the characteristics can be accurately extracted and identified, so that early warning can be timely given out when the fault is in the germination stage, further deterioration of the fault and sudden shutdown of equipment are effectively avoided, and the maintenance cost and the production loss caused by shutdown are greatly reduced.

Description

Technical Field

[0001] The invention relates to the field of fault monitoring, and in particular to a method for monitoring internal faults of a gear box. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] During the long-term operation of the gearbox, various faults are very likely to occur due to the combined effects of various factors such as complex and variable working loads, impacts caused by high-speed operation, friction and wear, poor lubrication, material fatigue, assembly process defects, and harsh working environment. For example; tooth surface wear, tooth root fatigue cracks, tooth surface bonding, broken teeth, bearing wear, and shaft deformation and fracture. These faults will not only lead to reduced transmission accuracy, increased vibration and noise of the gearbox, but also if they are not discovered and handled in time, the faults will further deteriorate, which may cause the entire equipment to shut down, causing huge economic losses, and even in some special application scenarios, may endanger the safety of personnel lives.

[0004] Traditional gearbox fault monitoring methods mainly rely on regular manual inspections and post-event maintenance. Manual inspections usually use methods such as listening to sounds, measuring temperature, and observing appearance to preliminarily judge the operating status of the gearbox, but this method has obvious limitations. It is difficult to capture early signs of faults in real time and accurately, and some internal hidden faults are almost undetectable. Post-event maintenance is to perform maintenance only after the fault has occurred and caused the equipment to shut down, which will undoubtedly increase the equipment's downtime and maintenance costs, and at the same time cannot meet the modern industry's requirements for high reliability and continuous operation of equipment.

[0005] With the rapid development of sensor technology, signal processing technology, computer technology and artificial intelligence technology, it is possible to realize real-time, online and intelligent monitoring of gearbox internal faults. By installing various sensors at key parts of the gearbox, such as vibration sensors, temperature sensors, acoustic sensors, oil sensors, etc., various parameter information reflecting the operating status of the gearbox can be collected in real time. Then, these raw data are analyzed and processed using advanced signal processing algorithms to extract the changing trends of sensitive indicators that can characterize the fault characteristics. Combined with intelligent diagnosis algorithms, such as fault diagnosis models based on machine learning and deep learning, the extracted fault characteristics are identified and classified, thereby achieving accurate diagnosis and early warning of internal faults in the gearbox, providing a strong basis for timely and effective maintenance measures, and ensuring the safe, stable and efficient operation of the gearbox and even the entire equipment system. Summary of the invention

[0006] The purpose of the present invention is to provide a method for monitoring internal faults of a gearbox, which can collect various operating data such as vibration, temperature, acoustics, etc. comprehensively and in real time by installing various types of sensors at multiple key positions of the gearbox, and can keenly capture the subtle changes caused by the initial faults inside the gearbox by using advanced signal processing and feature extraction technology. This monitoring method can accurately extract and identify these features, so as to issue an early warning in time when the fault is in its infancy, effectively avoiding further deterioration of the fault and sudden shutdown of the equipment, and greatly reducing maintenance costs and production losses caused by shutdowns.

[0007] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0008] A method for monitoring internal faults of a gearbox comprises the following steps;

[0009] S1. Installing a sensor group at multiple key locations of the gearbox, the sensor group includes a vibration sensor, a temperature sensor and an acoustic sensor for collecting vibration signals, temperature signals and acoustic signals during the operation of the gearbox;

[0010] S2. The vibration signal, temperature signal and acoustic signal collected are amplified, filtered and analog-to-digital converted by a signal conditioning circuit to obtain initial monitoring data in digital form;

[0011] S3. Preprocessing the initial monitoring data, removing noise interference by wavelet transform, and processing the data into a unified dimension and numerical range by data normalization method;

[0012] S4. Extract fault characteristic parameters from the preprocessed monitoring data, including the peak value, effective value, and kurtosis of the time domain characteristic parameters of the vibration signal, the spectrum amplitude and frequency component of the frequency domain characteristic parameters, the rate of change and gradient of the temperature signal, and the sound pressure level and frequency distribution characteristics of the acoustic signal;

[0013] S5. Input the extracted fault characteristic parameters into a pre-trained fault diagnosis model, wherein the fault diagnosis model is constructed based on a deep learning algorithm and adopts a multi-layer neural network structure, and learns and trains a large number of gearbox monitoring data samples containing different fault types and degrees to achieve identification and judgment of fault types and fault degrees;

[0014] S6. Determine whether there is a gearbox fault and the type and severity of the fault according to the output of the fault diagnosis model;

[0015] S7. When it is determined that the gearbox is faulty, the fault information is sent to the remote monitoring center through the wireless communication module, and the fault information includes the fault type, fault location, fault severity and corresponding monitoring data;

[0016] S8. After receiving the fault information, the remote monitoring center generates maintenance suggestions based on the preset fault handling strategy and feeds back the maintenance suggestions to the on-site operators. The maintenance suggestions include maintenance methods, parts to be replaced, and maintenance schedule.

[0017] Preferably, the sensor group also includes an oil sensor for detecting the oil in the gearbox. After the data collected by the oil sensor is subjected to signal conditioning and preprocessing, the extracted oil-related fault characteristic parameters are input into the fault diagnosis model to help improve the accuracy of fault diagnosis.

[0018] The training process of the fault diagnosis model includes: using a cross-validation method to divide the sample data into a training set, a validation set and a test set, using the validation set to evaluate and adjust the performance of the model, and when the accuracy of the model on the test set reaches a preset threshold, the model training is completed and the trained model parameters are saved.

[0019] The wireless communication module adopts communication technology or industrial Internet of Things communication protocol to ensure that fault information can be transmitted to the remote monitoring center quickly and stably, and has a data encryption function to ensure the security of data transmission.

[0020] The remote monitoring center is provided with a visual interface for displaying the operating status information, fault alarm information and trend charts of historical monitoring data of each gearbox in real time, so as to facilitate operators to intuitively understand the overall operating conditions and fault occurrence patterns of the gearbox.

[0021] Before preprocessing the monitoring data, a data verification step is also included. The integrity and rationality of the collected initial monitoring data are checked through a data verification algorithm, abnormal data are eliminated, and missing data are supplemented by interpolation or statistical laws based on historical data.

[0022] In the fault diagnosis model of the multi-layer neural network structure, the number of hidden layers and the number of neurons in each hidden layer are optimized and adjusted according to the complexity of the gearbox and the dimension of the monitoring data, and an adaptive learning rate adjustment strategy and regularization technology are used to prevent model overfitting and improve the generalization ability of the model.

[0023] The on-site operator can interact with the remote monitoring center through the mobile terminal device, receive maintenance suggestions and upload on-site maintenance conditions and operating status information after the equipment is restarted, so as to achieve full tracking and closed-loop management of the gearbox fault maintenance process.

[0024] Compared with the prior art, the beneficial effects of the present invention are:

[0025] Achieve accurate fault diagnosis and early warning: By installing various types of sensors at multiple key locations of the gearbox, comprehensive and real-time collection of various operating data such as vibration, temperature, acoustics, etc., and using advanced signal processing and feature extraction technology, it is possible to keenly capture the subtle changes caused by the initial faults inside the gearbox. This monitoring method can accurately extract and identify these features, thereby issuing a timely warning when the fault is in its infancy, effectively avoiding further deterioration of the fault and sudden shutdown of the equipment, and significantly reducing maintenance costs and production losses caused by shutdowns.

[0026] Improve the accuracy and reliability of fault diagnosis: The fault diagnosis model built based on the deep learning algorithm is used. Through in-depth learning and training of a large number of historical monitoring data samples covering various fault types and working conditions, the model can accurately identify complex and changeable fault modes, overcoming the limitations of traditional diagnostic methods that rely on manual experience and a single diagnostic indicator. Whether it is fatigue cracks of gears, tooth surface bonding, or wear and burns of bearings, etc., it can accurately distinguish and judge their severity, and the diagnostic accuracy is significantly improved, providing a reliable basis for maintenance decisions, reducing the risk of misjudgment and missed judgment, and ensuring the stable operation of equipment and the continuity of the production process.

[0027] Possess real-time online monitoring capability: The entire monitoring system can collect, process and analyze the gearbox operation data in real time, and realize uninterrupted monitoring of the gearbox working status. On-site operators and remote monitoring centers can obtain the latest equipment operation information at any time, and timely grasp the health status of the equipment. Once an abnormal situation occurs, they can respond quickly and take corresponding measures. This real-time online monitoring function greatly enhances the timeliness and initiative of equipment management, effectively improves the equipment maintenance efficiency and production safety, and is especially suitable for industrial production fields with extremely high requirements for equipment reliability.

[0028] Support remote monitoring and intelligent management: With the help of wireless communication modules, fault information is transmitted to the remote monitoring center in real time, breaking through geographical restrictions and facilitating professional technicians to centrally manage and uniformly dispatch multiple gearboxes distributed in different locations. The remote monitoring center can not only display the operating parameters and fault alarm information of each gearbox in real time, but also use big data analysis technology to conduct in-depth mining of historical monitoring data, summarize the laws and trends of fault occurrence, and provide strong support for preventive maintenance. At the same time, the maintenance suggestions generated according to the preset fault handling strategy can be fed back to the site in a timely manner, realizing the intelligent and information-based equipment maintenance, optimizing the configuration of equipment maintenance resources, reducing maintenance costs and management difficulties, and improving the overall operational efficiency of the enterprise.

[0029] Multi-parameter fusion improves comprehensiveness of diagnosis: In addition to common monitoring parameters such as vibration and temperature, acoustic signals and oil quality parameters such as viscosity, pH, and metal particle content are also introduced for comprehensive analysis. Different monitoring parameters reflect the operating status of the gearbox from multiple angles. The parameters complement and verify each other, avoiding possible misjudgments in single parameter monitoring. Acoustic signals will also have abnormal characteristics. Through the fusion analysis of these multi-source data, the health status of the gearbox can be judged more comprehensively and accurately, potential fault hazards can be discovered in a timely manner, and the safe and reliable operation of the equipment can be ensured.

[0030] Optimize equipment maintenance strategies and reduce costs: Based on accurate fault diagnosis results and intelligent remote monitoring management, enterprises can formulate more scientific and reasonable equipment maintenance plans according to the actual operating conditions of the gearbox. Instead of relying on traditional regular maintenance methods, they switch to preventive maintenance and precise repairs based on the actual health status of the equipment, avoiding waste of resources and unnecessary equipment downtime caused by excessive maintenance. At the same time, by early detection and timely repair of faults, the need to replace expensive parts due to serious damage to the equipment is reduced, effectively reducing the full life cycle cost of the equipment and improving the economic benefits and market competitiveness of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Attached Figure 1 It is a step diagram of a gearbox internal fault monitoring method of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the application equally.

[0033] The invention discloses a method for monitoring internal faults of a gearbox. The main structure of the method comprises a sensor module, a signal conditioning circuit, a data acquisition card, a microprocessor, a wireless communication module and a remote monitoring center.

[0034] Sensor module: Install high-precision vibration sensors on the surface of the gearbox near the gear meshing part, bearing seat and other key locations to capture the vibration of the gears and bearings during operation; install temperature sensors and oil quality sensors in the oil circulation system of the gearbox. The temperature sensor monitors the oil temperature in real time, and the oil quality sensor detects the oil viscosity, pH, metal particle content and other parameters; at the same time, arrange acoustic sensors at appropriate locations on the gearbox housing to collect sound signals generated during operation. These sensors have high sensitivity, wide-band response and good anti-interference ability, and can accurately collect various physical quantities reflecting the internal state of the gearbox.

[0035] Signal conditioning circuit: The original signal collected by the sensor is usually weak and may contain noise interference. The signal conditioning circuit performs pre-processing operations such as amplification and filtering. For example, for the weak electrical signal output by the vibration sensor, the amplifier circuit amplifies it to a suitable amplitude range for subsequent analog-to-digital conversion; the combination of low-pass, high-pass and band-pass filters is used to remove high-frequency noise and low-frequency drift in the signal, extract the effective signal frequency band, and ensure the signal quality transmitted to the data acquisition card.

[0036] Data acquisition card: After signal conditioning, the analog signal is converted into a digital signal by the data acquisition card. The data acquisition card has a high sampling rate and resolution, which can meet the needs of fast and accurate acquisition of various sensor signals. For example, for vibration signals, the sampling rate is set to several thousand hertz or even higher to capture the high-frequency components in the vibration signal, so as to accurately extract the fault characteristics.

[0037] Microprocessor: A high-performance microprocessor is selected to further process, analyze and calculate the collected digital signals. The microprocessor runs the specially developed fault monitoring algorithm software to realize data preprocessing, feature extraction, and the operation and result judgment of the fault diagnosis model. At the same time, the microprocessor also controls and coordinates the hardware modules of the entire monitoring system to ensure the stable operation of the system.

[0038] Wireless communication module: The microprocessor sends the fault diagnosis results and key monitoring data to the remote monitoring center through the wireless communication module. The wireless communication module can use 5G communication technology or a dedicated wireless communication protocol based on the industrial Internet of Things to ensure high-speed, stable and reliable data transmission. In addition, the communication module has a data encryption function to ensure the security of data during transmission and prevent data from being stolen or tampered with.

[0039] Fault monitoring method steps:

[0040] Data acquisition and preprocessing: After the system is initialized, each sensor starts to collect the gearbox operation data in real time, including vibration signals, temperature signals, acoustic signals and oil quality parameters, and transmits these analog signals to the signal conditioning circuit.

[0041] The signal conditioning circuit amplifies and filters the input analog signal, removes noise and interference components, and then outputs the conditioned analog signal to the data acquisition card.

[0042] The data acquisition card samples the analog signal at a set sampling frequency, converts it into a digital signal, and transmits it to the microprocessor.

[0043] After receiving the digital signal, the microprocessor first checks the data for integrity and rationality, and removes obviously abnormal data points through the data verification algorithm. For a small amount of missing data, linear interpolation or statistical laws based on historical data are used to supplement it to ensure the continuity and integrity of the data.

[0044] Wavelet transform is used to denoise the verified and supplemented data, remove random noise in the signal, and highlight the fault feature information. Then, the denoised data is normalized to unify the dimensions and numerical range of data from different sensors. For example, the vibration amplitude, temperature value, acoustic signal intensity, etc. are all normalized to the [0,1] interval for subsequent feature extraction and analysis.

[0045] Fault feature extraction:

[0046] Vibration signal feature extraction: Perform time domain analysis on the pre-processed vibration signal and calculate its time domain feature parameters such as peak value, effective value, and kurtosis. The peak value reflects the maximum instantaneous amplitude in the vibration signal, the effective value reflects the magnitude of the vibration energy, and the kurtosis is more sensitive to the impact component in the signal, which can effectively detect early local damage faults of gears and bearings.

[0047] The vibration signal is transformed into the frequency domain by fast Fourier transform (FFT) to obtain frequency domain characteristic parameters such as spectrum amplitude and frequency components. By analyzing the specific frequencies in the spectrum and their amplitude changes, it is possible to determine whether there are abnormalities in the meshing frequency of the gear and the fault characteristic frequency of the bearing, thereby diagnosing the wear, tooth surface bonding, broken teeth of the gear and the wear, fatigue peeling and other fault types of the bearing. For example, when the gear tooth surface is worn, the amplitude of the meshing frequency and its harmonics will increase, and may be accompanied by the appearance of sidebands; when the inner ring of the bearing fails, the spectrum amplitude at its fault characteristic frequency will increase significantly.

[0048] Temperature signal feature extraction: Calculate the rate of change and gradient of the temperature signal to monitor the rising or falling trend of the oil temperature. If the temperature rises too quickly, it may indicate excessive friction, poor lubrication or heat dissipation problems inside the gearbox, such as severe wear of the bearings leading to increased frictional heat generation, or a cooling system failure leading to poor heat dissipation.

[0049] Analyze the fluctuation of temperature signals. During normal operation, the temperature fluctuation is relatively small. If there is a large fluctuation, it may indicate that there is an intermittent fault source inside the gearbox, such as uneven wear of gears leading to uneven load distribution, which in turn causes periodic changes in temperature.

[0050] Acoustic signal feature extraction: Extract the sound pressure level of the acoustic signal. A sudden increase in the sound pressure level may indicate a serious fault inside the gearbox, such as gear breakage, bearing jamming, etc., which will generate a lot of noise.

[0051] Perform spectrum analysis on the acoustic signal to obtain its frequency distribution characteristics. Different fault types will generate noise with specific frequency components. For example, when the gear is slightly worn, the energy of the acoustic signal in the high frequency band will increase; and when the bearing fails, there will be obvious acoustic energy concentration at the specific bearing fault frequency and its harmonics.

[0052] Oil quality parameter feature extraction: According to the oil viscosity changes detected by the oil quality sensor, the oil aging degree and the decline in lubrication performance are judged. The abnormal increase in oil viscosity may be caused by oil contamination or oxidation, which will affect the lubrication effect of gears and bearings and increase the risk of wear; the decrease in viscosity may be caused by the mixing of other low-viscosity substances, which will also have an adverse effect on lubrication.

[0053] Analyze the changes in the pH of the oil. A pH outside the normal range may indicate a chemical reaction in the oil. This may be due to local overheating, metal corrosion and other problems inside the gearbox, which will accelerate the damage of components.

[0054] Monitor the metal particle content in the oil. The number and size of metal particles are important indicators for judging the degree of wear of gears and bearings. By analyzing the composition of metal particles, it is also possible to determine whether the gear or bearing is worn, and the severity of the wear.

[0055] Fault diagnosis and early warning: The extracted fault feature parameters are combined into feature vectors and input into a pre-trained fault diagnosis model based on a deep learning algorithm. The model adopts a multi-layer neural network structure, including an input layer, multiple hidden layers, and an output layer. The number of neurons in the input layer corresponds to the number of input fault feature parameters. The hidden layer performs deep feature learning and pattern recognition on the input features through weight connections and nonlinear activation functions between neurons. The output layer outputs the diagnosis results of the fault type and fault degree.

[0056] The training process of the fault diagnosis model is as follows:

[0057] A large number of gearbox monitoring data samples under different working conditions, including various known fault types and degrees, are collected, and the sample data are preprocessed and feature extracted to form training data sets and test data sets.

[0058] The neural network is trained using a suitable deep learning algorithm. The weights and biases between the neurons in the network are continuously adjusted to minimize the error between the model's prediction results on the training data set and the actual fault label. During the training process, the training data set is further divided into training subsets and validation subsets using the cross-validation method. The validation subset is used to evaluate and adjust the performance of the model to prevent overfitting. When the accuracy of the model on the test data set reaches the preset threshold, the model training is considered complete and the trained model parameters are saved.

[0059] The microprocessor determines whether the gearbox is faulty and the type and severity of the fault based on the output of the fault diagnosis model. If the diagnosis result indicates that there is a fault, the microprocessor sends the fault information to the remote monitoring center through the wireless communication module. At the same time, the sound and light alarm device will be triggered on the local monitoring system to remind the on-site operators to pay attention to the equipment status in time.

[0060] Remote monitoring and maintenance decision-making: After receiving fault information from multiple gearboxes, the remote monitoring center centrally manages and visualizes this information through a specially developed monitoring software platform. The monitoring platform presents the real-time operating status of each gearbox, fault alarm information, and trend changes of historical monitoring data in the form of intuitive charts and graphs, making it easy for technicians to quickly understand the overall operation of the equipment and the laws of fault occurrence.

[0061] Based on the preset fault handling strategy and expert experience database, the software system of the monitoring center automatically generates maintenance suggestions for each fault, including maintenance methods, a list of parts to be replaced, and maintenance schedule, etc. Maintenance suggestions are sent to on-site operators and related maintenance personnel in a timely manner via SMS, email, or system messages, so that they can quickly take effective maintenance measures.

[0062] After completing the fault repair, the maintenance personnel upload the maintenance situation and the running status information of the equipment after restart to the remote monitoring center through the mobile terminal device, so as to realize the whole tracking and closed-loop management of the gearbox fault repair process. The remote monitoring center re-evaluates and records the health status of the equipment based on the feedback information, provides a reference basis for subsequent equipment maintenance and management, and also helps to continuously optimize the fault diagnosis model and maintenance strategy, and improve the performance and reliability of the entire fault monitoring and maintenance system.

Claims

1. A method for monitoring internal faults of a gearbox, characterized in that: The steps include: S1. Installing a sensor group at multiple key locations of the gearbox, the sensor group includes a vibration sensor, a temperature sensor and an acoustic sensor for collecting vibration signals, temperature signals and acoustic signals during the operation of the gearbox; S2. The vibration signal, temperature signal and acoustic signal collected are amplified, filtered and analog-to-digital converted by a signal conditioning circuit to obtain initial monitoring data in digital form; S3. Preprocessing the initial monitoring data, removing noise interference by wavelet transform, and processing the data into a unified dimension and numerical range by data normalization method; S4. Extract fault characteristic parameters from the preprocessed monitoring data, including the peak value, effective value, and kurtosis of the time domain characteristic parameters of the vibration signal, the spectrum amplitude and frequency component of the frequency domain characteristic parameters, the rate of change and gradient of the temperature signal, and the sound pressure level and frequency distribution characteristics of the acoustic signal; S5. Input the extracted fault characteristic parameters into a pre-trained fault diagnosis model, wherein the fault diagnosis model is constructed based on a deep learning algorithm and adopts a multi-layer neural network structure, and learns and trains a large number of gearbox monitoring data samples containing different fault types and degrees to achieve identification and judgment of fault types and fault degrees; S6. Determine whether there is a gearbox fault and the type and severity of the fault according to the output of the fault diagnosis model; S7. When it is determined that the gearbox is faulty, the fault information is sent to the remote monitoring center through the wireless communication module, and the fault information includes the fault type, fault location, fault severity and corresponding monitoring data; S8. After receiving the fault information, the remote monitoring center generates maintenance suggestions based on the preset fault handling strategy and feeds back the maintenance suggestions to the on-site operators. The maintenance suggestions include maintenance methods, parts to be replaced, and maintenance schedule.

2. A gearbox internal fault monitoring method according to claim 1, characterized in that: The sensor group also includes an oil sensor for detecting the oil in the gearbox. After signal conditioning and preprocessing, the data collected by the oil sensor is extracted, and the oil-related fault characteristic parameters are input into the fault diagnosis model to help improve the accuracy of fault diagnosis.

3. A gearbox internal fault monitoring method according to claim 1, characterized in that: The training process of the fault diagnosis model includes: using a cross-validation method to divide the sample data into a training set, a validation set and a test set, using the validation set to evaluate and adjust the performance of the model, and when the accuracy of the model on the test set reaches a preset threshold, the model training is completed and the trained model parameters are saved.

4. A gearbox internal fault monitoring method according to claim 1, characterized in that: The wireless communication module adopts communication technology or industrial Internet of Things communication protocol to ensure that fault information can be transmitted to the remote monitoring center quickly and stably, and has a data encryption function to ensure the security of data transmission.

5. The method for monitoring internal faults of a gearbox according to claim 1, characterized in that: The remote monitoring center is provided with a visual interface for displaying the operating status information, fault alarm information and trend charts of historical monitoring data of each gearbox in real time, so as to facilitate operators to intuitively understand the overall operating conditions and fault occurrence patterns of the gearbox.

6. A gearbox internal fault monitoring method according to claim 1, characterized in that: Before preprocessing the monitoring data, a data verification step is also included. The integrity and rationality of the collected initial monitoring data are checked through a data verification algorithm, abnormal data are eliminated, and missing data are supplemented by interpolation or statistical laws based on historical data.

7. A gearbox internal fault monitoring method according to claim 1, characterized in that: In the fault diagnosis model of the multi-layer neural network structure, the number of hidden layers and the number of neurons in each hidden layer are optimized and adjusted according to the complexity of the gearbox and the dimension of the monitoring data, and an adaptive learning rate adjustment strategy and regularization technology are used to prevent model overfitting and improve the generalization ability of the model.

8. The method for monitoring internal faults of a gearbox according to claim 1, characterized in that: The on-site operator can interact with the remote monitoring center through the mobile terminal device, receive maintenance suggestions and upload on-site maintenance conditions and operating status information after the equipment is restarted, so as to achieve full tracking and closed-loop management of the gearbox fault maintenance process.

Images