CN117628005A - A fusion signal hydraulic motor fault diagnosis method and system - Google Patents

Abstract

The invention belongs to the technical field of fault diagnosis of a fusion signal hydraulic motor, and discloses a fault diagnosis method of the fusion signal hydraulic motor. According to the method for diagnosing the faults of the hydraulic motor with the fusion signal, 10 characteristic values such as root mean square, standard deviation and the like with the largest influence factors in the sound and vibration signals are selected by collecting the vibration and the sound signals of the hydraulic motor, and the fusion characteristic vector with higher and more obvious fault information content is constructed by adopting a similar label splicing method. And a fault diagnosis model of the plunger abrasion of the hydraulic motor is constructed by combining the fusion feature vector and the LightGBM algorithm, so that the rapid and accurate identification of the plunger abrasion of the hydraulic motor with 3 degrees is realized. Finally, the hydraulic motor plunger abrasion recognition results between different algorithms and different signals are compared and analyzed, and the superiority of the method is verified.

Description

A fusion signal hydraulic motor fault diagnosis method and system

Technical Field

The invention belongs to the technical field of fusion signal hydraulic motor fault diagnosis, and in particular relates to a fusion signal hydraulic motor fault diagnosis method.

Background Art

Hydraulic steering gear is a commonly used mechanical equipment in aerospace, submarines and ships, and plays a vital role in steering and maintaining stability. As the actuator of the steering gear hydraulic system, the working performance of the hydraulic motor determines the safe and stable operation of the entire steering gear system. Therefore, the use of simple and effective fault diagnosis methods is of great significance to the normal operation of the hydraulic system. Traditional hydraulic motor fault diagnosis methods rely heavily on empirical knowledge and have the problem of difficulty in extracting fault features. Some researchers have collected vibration signals or sound signals during the operation of hydraulic motors and combined them with machine learning for fault diagnosis. Vibration signals have a high signal-to-noise ratio and sensitivity, sound sensors can be collected non-contact, and sound signals have a higher bandwidth and a wider frequency range, which are often used in fault diagnosis methods. Zhu and Jiang et al. monitored the vibration signals of hydraulic piston pumps, and used wavelet transform to decompose and reconstruct the vibration signals to extract fault features such as time domain and frequency domain, and combined with convolutional neural networks to realize the classification of different faults of hydraulic piston pumps. Wang and Liu used empirical mode decomposition to obtain the modal function components of the vibration signal, and then calculated the entropy of the time-frequency matrix. Finally, they used random forests to diagnose the faults of centrifugal pumps. The sensitivity of the acceleration sensor is a fixed value, which is not accurate enough for early fault identification. The vibration signal has strong nonlinearity and non-stationarity, which has a great impact on the sensor acquisition signal. The sound sensor has the advantages of non-destructive installation and non-contact measurement, and has higher sensitivity. It is used by most scholars in data acquisition for fault diagnosis. Tang et al. reviewed mechanical fault diagnosis based on audio signal analysis, analyzed the advantages of using audio signals, and summarized the development prospects of fault diagnosis based on audio signals. Mian et al. collected the sound signals of bearings under different fault states and extracted the sound quality characteristics of the signals. Finally, they realized bearing fault diagnosis using the support vector machine algorithm. The sound signal contains a lot of interference noise in the environment. To improve the signal-to-noise ratio of the sound signal, it is necessary to perform decomposition and noise reduction processing. It is difficult to achieve denoising effect with a small number of decomposition layers. A large number of decomposition layers will reduce the useful information in the signal and affect the accuracy of diagnosis.

Due to the uncertainty and hidden nature of hydraulic system failures, the analysis of fault points is complex, and it is difficult to accurately identify hydraulic system failures using only a single signal source or fault feature. In order to ensure the stable operation of the inner curve radial piston hydraulic motor and realize timely and accurate fault monitoring, some scholars have proposed a fault diagnosis method based on multi-source sensor fusion features, and achieved relatively ideal results. Li et al. proposed a multi-feature fusion fault diagnosis model based on the Fourier descriptor and graphic features of the indicator graph to detect the fault type of the rod pump system, which enhanced the robustness of the features and verified that the multi-input feature fusion model has a higher diagnostic accuracy than the single-input feature model. Tang et al. used the vibration signal, pressure signal and sound signal of the hydraulic piston pump and combined it with the convolutional neural network (CNN) to propose a hydraulic piston pump fault diagnosis method with an adaptive learning rate. The three original signals were converted into two-dimensional time-frequency images by continuous wavelet transform, and different fault types were identified using the CNN model. In order to reduce the wear and failure risk of worm gearboxes, Karabacak et al. [9] processed the vibration, sound and thermal image data of normal and faulty rotors, and used artificial neural networks (ANN) and support vector machines (SVM) to extract singular, dual or triple forms of time domain, frequency domain and thermal image features, thus realizing wear fault diagnosis of worm gearboxes. Long et al. proposed a multi-sensor information-driven motor fault diagnosis method based on AdaBoost. The frequency domain features of the collected motor current, magnetic and vibration signals were extracted using Hilbert transform and Fourier transform, and the AdaBoost model was used for training and diagnosis. This fault diagnosis method has high robustness and generalization ability. Through the analysis of the above literature, it can be seen that the use of multi-signal fusion obviously has a higher diagnostic recognition accuracy than a single signal, but the processing of the original signal is more complicated, and the fault diagnosis method combined with traditional machine learning such as neural network and AdaBoost algorithm has a large memory usage and a long operation time, and the fault recognition rate is not high.

Traditional machine learning algorithms such as GBDT and AdaBoost are implementations of Boosting algorithms, which have the disadvantages of poor robustness and low diagnostic efficiency. XGBoost and LightGBM algorithms are based on the improvement of GBDT algorithm. The gradient boosting tree algorithm solves the problems of long training time and low fault recognition rate of traditional algorithms, and is currently the most widely used fault diagnosis method. Xiang and Wang et al. used fast Fourier transform and support vector machine algorithms to process the original data, respectively, and then combined the XGBoost model to sort and diagnose the feature importance, thereby improving the accuracy of rolling bearing fault diagnosis. Wu and Zhang et al. established an XGBoost fault recognition model to improve the accuracy and reliability of wind turbine fault diagnosis, and compared the diagnosis results with support vector machine and Adaboost algorithms. The results showed that the XGBoost algorithm has a higher classification accuracy. The research of the above scholars shows that the XGBoost algorithm has a higher fault diagnosis accuracy than the traditional machine learning algorithm, and is widely used in the fault diagnosis of various mechanical equipment. However, the way the XGBoost algorithm traverses data limits the speed of calculation and reduces the efficiency of fault diagnosis. The LightGBM algorithm can improve the efficiency of data calculation while ensuring accuracy, and is widely used in fault diagnosis of gearboxes, rolling bearings and hydraulic motors. Tang et al. studied the application of the LightGBM algorithm in wind turbine gearbox fault diagnosis in response to the low efficiency and accuracy of traditional machine learning algorithms in wind turbine gearbox fault diagnosis. It was further confirmed that the LightGBM algorithm has higher detection accuracy than traditional algorithms and has lower false alarm rate and missed detection rate. Guo et al. combined the wavelet denoising algorithm with the LightGBM model to diagnose the loosening fault of the hydraulic motor base bolt and compared it with the traditional machine learning algorithm. The results showed that the LightGBM algorithm has higher diagnostic efficiency. In order to solve the problems of long training time and low diagnostic efficiency of traditional fault diagnosis models based on deep learning algorithms, Xu et al. proposed a rolling bearing fault diagnosis method combining convolutional neural network and LightGBM. And by constructing a data set, it was verified that the diagnostic efficiency and accuracy of the LightGBM model are better than other models.

From the above literature, we can know that the sensitivity of vibration sensor is a fixed value, which is not accurate enough for early fault identification. In addition, vibration signal has strong nonlinearity and non-stationarity, which has a great impact on the sensor acquisition signal. Sound signal contains a lot of interference noise in the environment. To improve the signal-to-noise ratio of sound signal, decomposition and noise reduction processing is required. It is difficult to achieve denoising effect with fewer decomposition layers, and more decomposition layers will reduce the useful information in the signal and affect the diagnosis accuracy. The fault diagnosis recognition rate of single signal is not high, while the signal fusion method can combine the advantages of the two sensors and use the fault-sensitive frequency bands of different sensors to improve the fault recognition rate. For the existing multi-source sensor fusion method, some literatures have confirmed that it has a higher diagnostic effect than the single signal feature model. However, it is more troublesome to convert the original data and extract the features, and most of them are based on the traditional GBDT and AdaBoost algorithms for fault identification. This method takes a long time to train the fault diagnosis model and has low diagnostic efficiency.

Through the above analysis, the problems and defects of the prior art are as follows:

The fault of hydraulic motor is identified only through a single signal source or fault feature. Due to the uncertainty and hidden nature of the fault in hydraulic motor system, the fault identification result is inaccurate.

A single signal source combined with machine learning algorithms such as GBDT and AdaBoost is used to carry out hydraulic motor fault identification, but the robustness and efficiency of the diagnosis process are low.

Summary of the invention

In view of the problems existing in the prior art, the present invention provides a fusion signal hydraulic motor fault diagnosis method.

The present invention is implemented in this way: a fusion signal hydraulic motor fault diagnosis method comprises:

Step 1: simulate the piston failure by using the inner curve radial piston hydraulic motor test bench, and collect the vibration signal and sound signal of the motor when the piston is normal and three different degrees of wear;

Step 2: Use the wavelet soft threshold algorithm to perform noise reduction, and extract the time domain and frequency domain eigenvalues for feature fusion based on data correlation;

Step 3: Build a LightGBM motor plunger fault diagnosis model and divide the fused features into a data set containing normal and three types of fault signal labels for training and diagnosis;

Step 4: The superiority of this method is verified by comparing different algorithms and different signals.

Further, the method for simulating plunger failure by using the inner curve radial plunger hydraulic motor test bench is as follows:

The structure of the inner curve radial piston hydraulic motor is mainly composed of a stator, a rotor, an output shaft, a plunger and a ball. The stator curve consists of 6 evenly distributed action arcs, and 10 plungers are evenly spaced around the rotor. When the motor is working, high-pressure oil enters the rotor high-pressure chamber through the oil distribution shaft, generating a large pressure to push the plunger outward along the shaft radial direction, so that the ball contacts the inner curve guide rail to generate an interaction force, in which the tangential component drives the rotor to rotate continuously and output torque. At the same time, the hydraulic oil in the plunger chamber in the low-pressure area is discharged through the low-pressure oil port of the oil distribution shaft, and the motor is continuously operated by continuously introducing high-pressure oil and discharging low-pressure oil.

The vibration between the stator and the rotor is mainly caused by the plunger assembly. From the plunger motor structure diagram, we can see that there are two plungers in the same position. The typical vibration frequency is mainly:

Where n is the speed of the hydraulic motor, p is the number of inner curves, and q is the number of plungers. The frequency of one of the plungers can be expressed as

Furthermore, the method for extracting time domain and frequency domain eigenvalues based on data correlation to perform feature fusion is as follows:

(1) Feature selection

The denoised signals were visualized in Python for data correlation. Six time-domain features, including mean, root mean square, peak index, standard deviation, waveform index, and pulse index, and six frequency-domain features, including amplitude spectrum mean, amplitude spectrum standard deviation, frequency center of gravity, spectrum amplitude skewness, frequency standard deviation, and frequency skewness, were selected from the vibration signal and sound signal data according to feature importance.

(2) Model evaluation method

The confusion matrix is a commonly used indicator for evaluating model performance. It can reflect the relationship between the predicted results and the actual situation. Four indicators, including accuracy, precision, recall and f1-score, are often used for model evaluation.

The expressions of each evaluation index are:

PR curve ROC curve is a technique and tool for evaluating the performance of a classifier. It can calculate the accuracy and reliability of the prediction results. In the PR curve, recall and precision are plotted on the same curve. The ROC curve can visually compare the performance of the classifier by plotting the true positive rate (TPR) and false positive rate (FPR) of the classifier on a curve. TPR and FPR can be expressed as:

(3) Diagnostic method flow

Test bench construction and signal collection: Build an inner curve radial fault diagnosis test bench, wear the plunger at three different depths, and use vibration acceleration sensors and PCB sound sensors to collect vibration and sound signals of the motor when there are different faults;

Signal denoising and feature extraction: The vibration signal and sound signal are denoised using wavelet algorithm, and time domain and frequency domain features are extracted using mathematical methods;

Signal feature fusion: The feature fusion method is used to fuse the feature values of the vibration signal and the sound signal;

Model training and fault diagnosis: The vibration signal data, sound signal data, and fusion signal feature data sets are divided into training sets and test sets, respectively, and input into the fault diagnosis model for training and diagnosis, and the maximum depth is set to prevent overfitting.

Furthermore, the method for collecting the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear is as follows:

1) Experimental methods;

2) Data processing;

3) Feature selection.

Further, the experimental method:

The present invention builds an internal curve radial piston hydraulic motor fault diagnosis experimental platform to simulate the fault types of different wear depths of the piston structure. The test bench consists of a fault simulation system and a data acquisition system:

The plunger fault simulation test bench uses an inner curve radial plunger hydraulic motor to simulate the fault. The PCB sound sensor is installed 25cm outside the hydraulic motor. The vibration acceleration sensor is magnetically installed on the rear end cover of the hydraulic motor. The X-axis is the same as the radial direction of the motor. The speed sensor is installed 4mm outside the axial direction of the hydraulic motor.

Data acquisition system, PC end connects to data acquisition card via Ethernet to receive and store data; the speed display can display the speed of the hydraulic motor in real time to ensure that the speed is the same for each experiment, eliminate the interference of irrelevant factors, and improve the accuracy of the experimental results; during the experiment, the servo control system controls the hydraulic oil source to supply a stable flow rate to ensure that the speed of the hydraulic motor is stable at 300r/min; the sampling frequency of the data acquisition system is set to 2000Hz, and the vibration and sound signals of the plunger are collected when it is normal and when there are three different degrees of wear failures. Each set of experimental data is collected for 6 minutes.

Furthermore, the data processing method:

The experimental data obtained by vibration sensors and sound sensors usually contain interference from noise signals. In particular, the detection frequency of sound sensors is low and they are more sensitive to the sounds in the environment. Therefore, according to the characteristics of plunger wear fault signals, the wavelet soft threshold algorithm is used to decompose and reconstruct the vibration and sound signals to achieve the denoising effect of the original signals. The soft threshold function can be expressed as:

The signal after wavelet soft threshold denoising is a time domain signal. In order to better observe the characteristics of the signal and extract the frequency domain features, the fast Fourier transform (FFT) is selected to transform the signal into a frequency domain representation. The Fourier transform and its inverse transform formulas are:

Through frequency domain comparison, it can be seen that the burrs in the signal are reduced after wavelet soft threshold denoising, which reflects that the cleanliness of the signal has been improved and the denoising effect has been achieved; the present invention adopts an inner curve radial piston hydraulic motor, the number of pistons is 10, the inner curve of the shell is 6, and the experimental speed is 300r/min; from the comparison of Figure 6, it can be seen that the vibration sensor is more sensitive to the half frequency of the hydraulic motor speed, and the sound sensor is more sensitive to the one and two times the frequency of the hydraulic motor speed, and the frequency response effect of a single plunger is better; considering that the vibration sensor and the sound sensor have different response frequencies to plunger faults, the present invention combines the respective advantages of vibration and sound signals, and fuses the characteristic values of the two signals to improve the diagnostic accuracy of hydraulic motor plunger faults.

Furthermore, the feature selection method:

The inner curve radial piston hydraulic motor test bench was used to simulate the piston failure and collect the vibration and sound signals of the piston in different states. The acquisition frequency was set to 2000Hz, and each set of data was collected for 6 minutes. Each set of data had 720,000 sampling points. 4000 sampling points were selected in turn to generate a sample, with a total of 180 samples. The number of samples in the training set and the test set were 80% and 20% respectively.

Time domain and frequency domain characteristics are commonly used indicators to reflect the operating status of hydraulic motors. Vibration and sound sensors are used to collect signals of normal and different wear of the plunger. After wavelet soft threshold denoising, mathematical methods are used to extract 11 categories of time domain and frequency domain eigenvalues of each group of data. Vibration signal and sound signal eigenvalue data sets are generated, and correlation analysis of the data sets is performed:

The data correlation heat map reflects the correlation information between the data. The larger the positive correlation and negative correlation values in the figure, the more it can reflect the operating status information of the hydraulic motor; based on the heat map, 10 types of vibration signal characteristic values are selected, including root mean square value, root square amplitude, standard deviation, peak index, waveform index, spectrum amplitude mean, spectrum amplitude standard deviation, spectrum amplitude skewness, spectrum amplitude kurtosis, and frequency skewness; 10 types of sound signal characteristic values are selected, including root mean square value, root square amplitude, peak value, standard deviation, peak index, pulse index, spectrum amplitude mean, spectrum amplitude standard deviation, frequency center of gravity, and frequency standard deviation; finally, each group of normal and fault characteristic values are horizontally fused according to the corresponding data labels, without changing the number of samples, only increasing the length of the data set.

Furthermore, the fusion features are divided into a data set containing normal and three types of fault signal labels for training and diagnosis:

(1) Data pre-classification and visualization;

(2) Model evaluation.

Furthermore, the data pre-classification and visualization method:

In order to observe the classification effect of the fused signal data relative to the vibration and sound signals in advance, the T-SNE algorithm is used to visualize the data. It can be seen from the figure that the classification of vibration signals has a certain degree of clustering, but there are too many clustering points, and samples of the same category are not clustered together. The classification of sound signals is more discrete, and different types of data points are fused together, and the classification effect is poor. For the classification of fused signals, it can be seen that different types of data points are almost completely separated, there are fewer intersections between samples, and the classification effect is better.

Further, the model evaluation method:

The AdaBoost algorithm is an improvement on the Boosting algorithm, and has high effectiveness and practicality; the GBDT algorithm is a gradient boosting iterative decision tree algorithm, and the XGBoost and LightGBM algorithms are improvements based on GBDT, and are also the most popular decision tree boosting algorithms, with higher learning efficiency; the present invention verifies the efficiency of the LightGBM algorithm and the classification effect of the fusion signal data by comparing the classification accuracy of the four algorithms for the fusion signal, vibration signal and sound signal respectively;

The diagnostic results of different models show that the traditional AdaBoost algorithm has poor classification effect, and its diagnostic accuracy for the three data sets is more than 10% lower than that of the other three algorithms; the classification effects of GBDT, XGBoost and LightGBM algorithms on the three data sets are not much different, but the diagnostic accuracy of the LightGBM algorithm is 0.17%-0.33% higher, and the diagnostic results of the three algorithms for the fusion data set are more accurate. The training time of the AdaBoost algorithm is the shortest, but its diagnostic accuracy is lower; the diagnostic accuracy of the GBDT, XGBoost and LightGBM algorithms is similar, but the training time of LightGBM is 6 times and 3 times shorter than that of GBDT and XGBoost algorithms, respectively. Therefore, the LightGBM algorithm has higher diagnostic accuracy and higher efficiency;

The vibration signal and sound signal are denoised using the wavelet soft threshold algorithm, and the eigenvalues are extracted and fused according to the corresponding data labels to generate a fused signal data set. Each data set contains normal and three fault types, but the number of samples does not change. The LightGBM model is used to diagnose the faults of the fused signal, vibration signal and sound signal, and introduces accuracy, precision, recall and f1-score as evaluation indicators of classification performance. The LightGBM model is used to train and diagnose the three data sets. By comparison, it can be seen that the diagnostic accuracy of the fused data set is 4.86% and 14.59% higher than that of the vibration and sound signals respectively. The diagnostic results of the precision, recall and f1-score of the fused data set are also 4.5%-13.25% higher than those of the vibration and sound data sets.

ROC and PR curves are used to evaluate the classification performance of machine learning algorithms for a given data set. Each data set contains a fixed number of positive samples and negative samples. The larger the area under the curve, the better the classification performance of the machine learning algorithm for the data set. FPR refers to the probability that a sample that is actually a negative sample is predicted to be a positive sample, and TPR refers to the probability that a sample that is actually a positive sample is predicted to be a negative sample. Precision refers to the proportion of samples predicted to be positive to positive samples, and Recall refers to the proportion of samples predicted to be positive to positive samples. Based on the classification effect of the LightGBM algorithm on vibration, sound, and fusion signal data sets, the LightGBM algorithm has the worst classification effect on the sound data set, and the classification effect on the fusion data set is much higher than that of the vibration and sound data sets.

Another object of the present invention is to provide a fusion signal hydraulic motor fault diagnosis system, comprising:

Fault simulation component: including an inner curve radial piston hydraulic motor test bench, which is used to simulate the normal state and different degrees of wear faults of the piston, and collect the corresponding vibration and sound signals;

Signal processing component module: Use the wavelet soft threshold algorithm to reduce the noise of the collected signal and improve the signal quality for subsequent analysis;

Feature extraction and fusion module: Based on data correlation analysis, feature values are extracted from the processed time domain and frequency domain signals, and feature fusion is performed to form a comprehensive feature set for fault diagnosis;

Fault diagnosis model module: Build a motor plunger fault diagnosis model based on the LightGBM algorithm, and use a fusion feature dataset containing normal and three fault signal labels for model training and diagnosis;

Performance Verification Component Module: Contains tools for evaluating and verifying the performance of diagnostic models, such as confusion matrix, PR curve, and ROC curve, as well as comparative analysis of different algorithms and signal types to verify the effectiveness and superiority of the proposed method.

Furthermore, it includes an inner curve radial piston hydraulic motor test bench for simulating normal and different degrees of wear failure of the piston, and is equipped with a vibration and sound signal collection device for collecting corresponding signals of the motor in various states.

Furthermore, a signal processing module is included, which uses a wavelet soft threshold algorithm to perform noise reduction processing on the collected vibration and sound signals, thereby improving the signal quality so as to facilitate more accurate fault diagnosis.

Furthermore, a feature extraction and fusion module is included for extracting time domain and frequency domain feature values from the vibration and sound signals after noise reduction based on data correlation analysis, and performing feature fusion to form a comprehensive feature set for fault diagnosis.

Furthermore, a motor plunger fault diagnosis model building module based on the LightGBM algorithm is included to train and diagnose a fused feature dataset containing normal and three types of fault signal labels, thereby achieving accurate diagnosis of hydraulic motor plunger faults.

Furthermore, it includes a model evaluation and validation module, which uses evaluation tools such as confusion matrix, PR curve and ROC curve to evaluate and validate the performance of the diagnostic model, including the calculation of indicators such as accuracy, precision, recall and f1-score, as well as comparative analysis of different algorithms and signal types to verify the superiority of the diagnostic method.

In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

First, a fusion signal hydraulic motor fault diagnosis method provided by the present invention. The plunger fault is simulated by an inner curve radial piston hydraulic motor test bench, and the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear are collected. The wavelet soft threshold algorithm is used for noise reduction processing, and feature selection is performed based on data correlation and feature importance. The time domain and frequency domain eigenvalues are extracted for feature fusion. A LightGBM motor plunger fault diagnosis model is constructed, and the fusion features are divided into a data set containing normal and three fault signal labels for training and diagnosis. Finally, the superiority of this method is verified by comparing different algorithms and different signals. The fusion feature fault diagnosis method proposed in the present invention solves the problem of low diagnostic accuracy when using a single sensor to collect data. By using two different signals for diagnosis, the fault information of different sensors can be combined, and the fault features can be increased to improve the fault recognition rate. The fault diagnosis model is established using LightGBM. Under the premise of ensuring classification accuracy, it not only reduces the running memory usage but also increases the model training time, greatly improving the fault diagnosis efficiency.

The results show that the diagnostic accuracy of the LightGBM algorithm is 0.17%-0.33% higher than that of the AdaBoost, GBDT and XGBoost algorithms; the diagnostic accuracy of the fusion signal under the LightGBM model is 4.86% and 14.59% higher than that of the vibration and sound signals, respectively.

Second, the present invention provides a fusion signal hydraulic motor fault diagnosis method. The plunger fault is simulated by an inner curve radial piston hydraulic motor test bench, and the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear are collected. The wavelet soft threshold algorithm is used for noise reduction processing, and feature selection is performed based on data correlation and feature importance. The time domain and frequency domain eigenvalues are extracted for feature fusion. A LightGBM motor plunger fault diagnosis model is constructed, and the fusion features are divided into a data set containing normal and three fault signal labels for training and diagnosis. Finally, the superiority of this method is verified by comparing different algorithms and different signals. The fusion feature fault diagnosis method proposed in the present invention solves the problem of low diagnostic accuracy when using a single sensor to collect data. By using two different signals for diagnosis, the fault information of different sensors can be combined, and the fault features can be increased to improve the fault recognition rate. The fault diagnosis model is established using LightGBM. Under the premise of ensuring classification accuracy, it not only reduces the running memory usage but also increases the model training time, greatly improving the fault diagnosis efficiency.

The results show that the diagnostic accuracy of the LightGBM algorithm is 0.17%-0.33% higher than that of the AdaBoost, GBDT and XGBoost algorithms; the diagnostic accuracy of the fusion signal under the LightGBM model is 4.86% and 14.59% higher than that of the vibration and sound signals, respectively.

Third, the expected benefits and commercial value of the technical solution of the present invention after transformation are as follows: a) Due to the uncertainty and hidden nature of hydraulic motor system failures, it is difficult to accurately identify hydraulic motor failures only through a single signal source or fault feature. The fusion signal fault diagnosis method combines the advantages of sound and vibration sensors, contains more fault information, and makes feature extraction and fusion simpler, and the diagnosis results more reliable; b) 10 eigenvalues such as the root mean square and standard deviation with the largest influencing factors in sound and vibration signals are selected, and feature fusion is performed using similar labels to form a new feature vector. The sample vector has higher and more obvious fault information content; c) The fusion sample vector is combined with the LightGBM algorithm model to perform multi-threaded parallel computing, saving computer memory usage, greatly reducing diagnosis time, and improving the fault identification performance of the plunger motor.

The technical solution of the present invention fills the technical gaps in the industry at home and abroad: a) The present invention provides a method for diagnosing plunger wear faults of inner-curve plunger hydraulic motors based on the LightGBM algorithm to identify sound-vibration fusion signal characteristics, opening up a feasible technical route for fault diagnosis of plunger hydraulic motors; b) Fault diagnosis methods based on machine learning are widely used in rotating machinery such as bearings and gears, but there is no application in the diagnosis of plunger motors, especially inner-curve radial plunger hydraulic motors, in existing domestic and foreign literature.

Fourth, the significant technical progress of the fusion signal hydraulic motor fault diagnosis method provided by the present invention can be summarized as follows:

1. Accurate acquisition of fault simulation and experimental data

Authenticity of fault simulation: By simulating plunger faults through the inner curve radial piston hydraulic motor test bench, the fault conditions that occur in actual work can be accurately simulated, thereby obtaining test data that is closer to the actual situation.

Multi-dimensional data collection: Collect vibration signals and sound signals, and combine two different signal types to improve the accuracy and reliability of fault diagnosis.

2. Efficient signal processing and feature fusion

Application of denoising algorithm: Use wavelet soft threshold algorithm for noise reduction processing to effectively remove noise in the signal and improve the quality of the signal.

Innovation in feature fusion: Using time domain and frequency domain eigenvalues for feature fusion enhances the diagnostic capability of the model and ensures the comprehensiveness and depth of fault diagnosis.

3. Leverage advanced machine learning models

Application of LightGBM model: Build a motor plunger fault diagnosis model based on LightGBM, which has the characteristics of high efficiency and high precision and is suitable for processing large-scale data.

Optimization of model training and diagnosis: The fusion features are divided into a data set containing normal and three types of fault signal labels for training, which improves the generalization ability and diagnostic accuracy of the model.

4. Comprehensive verification of system performance

Comprehensive algorithm comparison: By comparing different algorithms, the superiority of the proposed method is verified, ensuring the advancement and effectiveness of the method.

Comprehensive analysis of signal types: Comparative analysis between different signals further verified the importance of feature fusion in improving diagnostic accuracy.

The significant technical progress of the fusion signal hydraulic motor fault diagnosis method provided by the present invention is mainly reflected in improving the accuracy, reliability and efficiency of fault diagnosis. By comprehensively utilizing multiple signals, advanced signal processing methods and machine learning models, the performance of hydraulic motor fault diagnosis is significantly improved.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a fusion signal hydraulic motor fault diagnosis method provided in an embodiment of the present invention.

FIG. 2 is a structural diagram of an inner-curve radial piston motor provided in an embodiment of the present invention.

FIG3 is a schematic diagram of the LightGBM algorithm provided in an embodiment of the present invention.

FIG. 4 is a flow chart of a method provided by an embodiment of the present invention.

FIG. 5 is a diagram of an experimental device provided in an embodiment of the present invention.

FIG. 6 is a diagram showing the degree of wear of the plunger provided in an embodiment of the present invention.

FIG. 7 is a comparison diagram of signal denoising effects provided by an embodiment of the present invention.

FIG8 is a data correlation heat map provided by an embodiment of the present invention.

Fig. 9 is a T-SNE clustering diagram provided by an embodiment of the present invention. (a) vibration signal (b) sound signal (c) fusion signal.

Fig. 10 is a confusion matrix diagram provided by an embodiment of the present invention. (a) vibration signal (b) sound signal (c) fusion signal.

FIG. 11 is a diagram of diagnosis results of different models provided by an embodiment of the present invention.

FIG12 is a diagram of training durations for different models provided in an embodiment of the present invention.

FIG13 is a diagram of the LightGBM prediction accuracy provided by an embodiment of the present invention.

FIG14 is a diagram showing the classification effect of the vibration, sound and fusion signal data sets based on the LightGBM algorithm provided by an embodiment of the present invention. (a) ROC curve; (b) PR curve.

In Figure 2: 1, ball; 2, plunger; 3, rotor; 4, output shaft; 5, stator.

DETAILED DESCRIPTION

In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

Based on the fusion signal hydraulic motor fault diagnosis method provided by the present invention, the following are two specific embodiments and their implementation schemes:

Example 1: Diagnosis of minor wear faults

1) Fault simulation: On the inner curve radial piston hydraulic motor test bench, simulate the plunger fault with slight wear.

2) Data collection: Use vibration acceleration sensors and PCB sound sensors to collect vibration signals and sound signals under normal conditions and slight wear conditions.

3) Signal processing: The collected signal is denoised using the wavelet soft threshold algorithm, and then the eigenvalues are extracted from the time domain and frequency domain.

4) Feature fusion and model construction: Fusion of time domain and frequency domain feature values to build a motor plunger fault diagnosis model based on LightGBM.

5) Model training and diagnosis: The model is trained and diagnosed using a dataset containing labels of normal and minor wear signals.

6) Performance evaluation: The accuracy and reliability of the model are evaluated through confusion matrix and PR curve ROC curve.

Example 2: Diagnosis of severe wear fault

1) Fault simulation: Severely worn plunger failure is simulated on the same test bench.

2) Data collection: Vibration and sound signals under normal and severe wear conditions are also collected.

3) Signal processing: Perform the same noise reduction processing on the signal and extract the corresponding time domain and frequency domain eigenvalues.

4) Feature fusion and model optimization: Fuse the extracted features and optimize the LightGBM model to adapt to the fault characteristics of heavy wear.

5) Model training and diagnosis: The model is trained and diagnosed using a dataset containing labels of heavy wear signals.

6) Performance verification: Model evaluation indicators are used to verify the performance of the model to ensure high accuracy even under severe wear conditions.

The two embodiments provided by the present invention show how to apply the fusion signal hydraulic motor fault diagnosis method for different degrees of wear faults (minor and severe), provide practical operation processes and steps, and ensure high accuracy and effectiveness of fault diagnosis.

As shown in FIG1 , the present invention provides a fusion signal hydraulic motor fault diagnosis method comprising the following steps:

S101, simulate the piston failure through the inner curve radial piston hydraulic motor test bench, and collect the vibration and sound signals of the motor when the piston is normal and has three different degrees of wear;

S102, using a wavelet soft threshold algorithm to perform noise reduction processing, and extracting time domain and frequency domain feature values for feature fusion based on data correlation;

S103, build the LightGBM motor plunger fault diagnosis model, divide the fusion features into a data set containing normal and three types of fault signal labels for training and diagnosis;

S104, the superiority of this method is verified by comparing different algorithms and different signals.

The method for simulating plunger failure by using an inner curve radial plunger hydraulic motor test bench provided by the present invention is as follows:

The working principle of the inner curve radial piston hydraulic motor is shown in Figure 2. Its structure mainly consists of a stator, a rotor, an output shaft, a plunger and a ball. The stator curve consists of 6 evenly distributed action arcs, and 10 plungers are evenly spaced around the rotor. When the motor is working, high-pressure oil enters the rotor high-pressure chamber through the oil distribution shaft, generating a large pressure to push the plunger outward along the shaft radially, so that the ball contacts the inner curve guide rail to generate an interaction force, in which the tangential component drives the rotor to rotate continuously and output torque. At the same time, the hydraulic oil in the plunger chamber in the low-pressure area is discharged through the low-pressure oil port of the oil distribution shaft, and the motor is continuously operated by continuously introducing high-pressure oil and discharging low-pressure oil.

The vibration between the stator and the rotor is mainly caused by the plunger assembly. From the plunger motor structure diagram, we can see that there are two plungers in the same position. The typical vibration frequency is mainly:

Where n is the speed of the hydraulic motor, p is the number of inner curves, and q is the number of plungers. The frequency of one of the plungers can be expressed as

LightGBM Algorithm Principle

Both XGBoost and LightGBM algorithms are improvements based on the GBDT algorithm. Both GBDT and XGBoost models use the Level-wise growth strategy. By traversing the data once, the leaves of the same layer can be split at the same time, reducing overfitting and making it easier to optimize for multiple threads. However, this method requires traversing the data layer by layer, which will generate many unnecessary searches or splits, thereby consuming more running memory, increasing the computing time and reducing efficiency. The LightGBM model integrates the histogram algorithm, which is based on the principle of dividing the original data of the feature into k discrete features and constructing k boxes for statistical features. When searching, there is no need to traverse the data, only k boxes need to be traversed to find the best split point, which greatly shortens the computing time. The splitting method of the LightGBM model adopts the Leaf-wise growth strategy. This method searches for the leaf with the largest information gain in the current leaf. Under the same number of splits, the Leaf-wise growth strategy has lower error than the Level-wise growth strategy, occupies less running memory, and improves the accuracy and efficiency of the algorithm. The disadvantage of the LightGBM model is that it is prone to overfitting. Therefore, a maximum depth parameter is added to limit it during model training.

The LightGBM model also integrates two algorithms: Gradient-based One-Side Sampling (GOSS) and Exclusive Feature Bundling (EFB), the principle of which is shown in Figure 3. GOSS uses gradient information to sample samples, and samples with large gradients have a greater impact on information gain. When sampling, samples with large gradients are retained, and samples with small gradients are randomly sampled, which does not lose model accuracy but can reduce memory usage and training time, greatly improving the diagnostic efficiency of the model. The feature space in the data is often mutually exclusive. The EFB algorithm bundles multiple mutually exclusive features together from the perspective of reducing feature dimensions, reducing the number of features and improving the running speed of the model.

The method for extracting time domain and frequency domain eigenvalues and performing feature fusion based on data correlation provided by the present invention is as follows:

(1) Feature selection

The denoised signal is visualized in Python for data correlation. Six time domain features, including mean, root mean square, peak index, standard deviation, waveform index, and pulse index, and six frequency domain features, including amplitude spectrum mean, amplitude spectrum standard deviation, frequency center of gravity, spectrum amplitude skewness, frequency standard deviation, and frequency skewness, are selected from the vibration signal and sound signal data according to the feature importance. These features can best reflect the working status of the inner curve plunger motor and can timely reflect the fault information of the motor. The mathematical formulas of time domain and frequency domain features are shown in Table 1:

Table 1 Statistical characteristics of time domain and frequency domain

In the table, x(n) (n=1, 2, ..., N) represents the original signal, N represents the number of data points; s(k) (k=1, 2, ..., M) represents the signal spectrum amplitude, M represents the number of spectrum lines; f(k) represents the amplitude of the frequency of the kth spectrum line;

(2) Model evaluation method

The confusion matrix is a common indicator for evaluating model performance, as shown in Table 2. It can reflect the relationship between the predicted results and the actual situation. Four indicators, including accuracy, precision, recall and f1-score, are often used for model evaluation.

Table 2 Confusion matrix

The expressions of each evaluation index are:

PR curve ROC curve is a technique and tool for evaluating the performance of a classifier. It can calculate the accuracy and reliability of the prediction results. In the PR curve, recall and precision are plotted on the same curve. The ROC curve can visually compare the performance of the classifier by plotting the true positive rate (TPR) and false positive rate (FPR) of the classifier on a curve. TPR and FPR can be expressed as:

(3) Diagnostic method flow

Test bench construction and signal collection: Build an inner curve radial fault diagnosis test bench, wear the plunger at three different depths, and use vibration acceleration sensors and PCB sound sensors to collect vibration and sound signals of the motor when there are different faults;

Signal denoising and feature extraction: The vibration signal and sound signal are denoised using wavelet algorithm, and time domain and frequency domain features are extracted using mathematical methods;

Signal feature fusion: The feature fusion method is used to fuse the feature values of the vibration signal and the sound signal;

Model training and fault diagnosis: The vibration signal data, sound signal data and fusion signal feature data set are divided into training sets and test sets respectively, and input into the fault diagnosis model for training and diagnosis, and the maximum depth is set to prevent overfitting.

The diagnostic method flow is shown in Figure 4.

The method provided by the present invention for collecting the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear is as follows:

1) Experimental methods;

2) Data processing;

3) Feature selection.

The experimental method provided by the present invention is:

The present invention builds an internal curve radial piston hydraulic motor fault diagnosis experimental platform to simulate the fault types of different wear depths of the piston structure. The test bench consists of a fault simulation system and a data acquisition system. The experimental equipment and sensor installation positions are shown in Figure 5:

The plunger fault simulation test bench is shown in Figure 5(a). An inner curve radial plunger hydraulic motor is used to simulate the fault. The PCB sound sensor is installed 25 cm outside the hydraulic motor. The vibration acceleration sensor is magnetically installed on the rear end cover of the hydraulic motor. The X-axis is the same as the radial direction of the motor. The speed sensor is installed 4 mm outside the axial direction of the hydraulic motor.

The data acquisition system is shown in Figure 5(b). The PC is connected to the data acquisition card via Ethernet to receive and store data. The speed display can display the speed of the hydraulic motor in real time to ensure that the speed of each experiment is the same, eliminate the interference of irrelevant factors, and improve the accuracy of the experimental results. During the experiment, the servo control system controls the hydraulic oil source to supply a stable flow rate to ensure that the speed of the hydraulic motor is stable at 300r/min. The sampling frequency of the data acquisition system is set to 2000Hz, and the vibration and sound signals of the plunger are collected when it is normal and when there are three different degrees of wear failures. Each set of experimental data is collected for 6 minutes. The models and parameters of each experimental equipment are shown in Table 3:

Table 3 Experimental equipment models and parameters

Device Name model parameter Vibration Sensor 1C302 Frequency range: 20Hz- 12KHz Sound Sensor PCB378B02 Frequency range: 3.75Hz- 20KHz Piston hydraulic motor 1001-0.1 Maximum speed: 360r/min Data Acquisition Card FK2012 16 channels Speed sensor ZSM12-CS-01 Output current: 50 mA Speed display WR5135-FR-N Measuring range: 0.2Hz-100KHz

In order to achieve better experimental results, the fault is created along the axis of the plunger by using the transverse wear method. According to the wear depth, the plunger fault is divided into light wear, moderate wear and deep wear. The specific wear effect is shown in Figure 6.

The data processing method provided by the present invention:

The experimental data obtained by vibration sensors and sound sensors usually contain interference from noise signals. In particular, the detection frequency of sound sensors is low and they are more sensitive to the sounds in the environment. Therefore, according to the characteristics of plunger wear fault signals, the wavelet soft threshold algorithm is used to decompose and reconstruct the vibration and sound signals to achieve the denoising effect of the original signals. The soft threshold function can be expressed as:

The signal after wavelet soft threshold denoising is a time domain signal. In order to better observe the characteristics of the signal and extract the frequency domain features, the fast Fourier transform (FFT) is selected to transform the signal into a frequency domain representation. The Fourier transform and its inverse transform formulas are:

FIG7 is a comparison diagram of the denoising effects of vibration signals and sound signals, respectively. It can be seen from the frequency domain comparison that the burrs in the signal are reduced after wavelet soft threshold denoising, which reflects that the cleanliness of the signal has been improved and the denoising effect has been achieved. The present invention adopts an inner curve radial piston hydraulic motor with 10 pistons, 6 inner curves of the housing, and an experimental rotation speed of 300 r/min. It can be seen from the comparison in FIG7 that the vibration sensor is more sensitive to the half frequency of the hydraulic motor rotation speed, and the sound sensor is more sensitive to the one and two frequencies of the hydraulic motor rotation speed, and has a better frequency response effect to a single plunger. Taking into account that the vibration sensor and the sound sensor have different response frequencies to plunger failures, the present invention combines the respective advantages of vibration and sound signals and fuses the characteristic values of the two signals to improve the diagnostic accuracy of hydraulic motor plunger failures.

The feature selection method provided by the present invention:

The inner curve radial piston hydraulic motor test bench is used to simulate the piston failure and collect the vibration and sound signals of the piston in different states; the acquisition frequency is set to 2000Hz, each set of data is collected for 6 minutes, each set of data has a total of 720000 sampling points, 4000 sampling points are selected in turn to generate a sample, a total of 180 samples, the number of samples in the training set and the test set are 80% and 20% respectively; the piston type and the number of samples are shown in Table 4:

Table 4 Dataset information

state Training set/test set Label normal 144/36 0 Fault A 144/36 1 Fault B 144/36 2 Fault C 144/36 3

Time domain and frequency domain characteristics are commonly used indicators to reflect the operating status of hydraulic motors. Vibration and sound sensors are used to collect signals of plungers in normal and different wear states. After wavelet soft threshold denoising, mathematical methods are used to extract 11 categories of time domain and frequency domain eigenvalues of each group of data. Vibration signal and sound signal eigenvalue data sets are generated, and data set correlation analysis is performed. The information correlation heat map between data sets is shown in Figure 8:

The data correlation heat map in Figure 8 reflects the correlation information between the data. The larger the positive correlation and negative correlation values in the figure, the more they can reflect the operating status information of the hydraulic motor; based on the heat map, 10 types of vibration signal feature values are selected, including root mean square value, root square amplitude, standard deviation, peak index, waveform index, spectrum amplitude mean, spectrum amplitude standard deviation, spectrum amplitude skewness, spectrum amplitude kurtosis, and frequency skewness; 10 types of sound signal feature values are selected, including root mean square value, root square amplitude, peak value, standard deviation, peak index, pulse index, spectrum amplitude mean, spectrum amplitude standard deviation, frequency center of gravity, and frequency standard deviation; finally, each group of normal and fault feature values are horizontally spliced according to the corresponding data labels to form a new feature vector without changing the number of samples;

The present invention provides a method for training and diagnosing by dividing the fusion features into a data set containing normal and three types of fault signal labels:

(1) Data pre-classification and visualization;

(2) Model evaluation.

The data pre-classification and visualization method provided by the present invention:

In order to observe the classification effect of the fused signal data relative to the vibration and sound signals in advance, the T-SNE algorithm is used to visualize the data. The T-SNE cluster diagram is shown in Figure 9. It can be seen from the figure that the classification of vibration signals has a certain degree of clustering, but there are too many clustering points, and samples of the same category are not clustered together; the classification of sound signals is highly discrete, and different types of data points are fused together, and the classification effect is poor; for the classification of fused signals, it can be seen that different types of data points are almost completely separated, there are fewer intersections between samples, and the classification effect is better;

Figure 10 is the four-classification confusion matrix of the vibration signal, sound signal and fusion signal data sets, which is used to visualize the classification effects of the three data sets; the number of samples with labels 0-3 in the test set are 28, 32, 44 and 40 respectively; from the vibration signal confusion matrix in Figure 10(a), it can be seen that there are 3 samples predicted as faults when they are actually normal, and 9 samples are misclassified among the three faults; from the sound signal confusion matrix in Figure 10(b), it can be seen that all the samples that are actually normal are predicted correctly, but there are 23 samples that are misclassified among the three faults; from the fusion signal confusion matrix in Figure 10(c), it can be seen that all the samples that are actually normal are predicted correctly, and there are only 5 samples that are misclassified between fault B and fault C among the three faults; the classification accuracy of the fusion signal data set is much higher than that of the vibration signal and sound signal.

The model evaluation method provided by the present invention:

The AdaBoost algorithm is an improvement on the Boosting algorithm, and has high effectiveness and practicality; the GBDT algorithm is a gradient boosting iterative decision tree algorithm, and the XGBoost and LightGBM algorithms are improvements based on GBDT, and are also the most popular decision tree boosting algorithms, with higher learning efficiency; the present invention verifies the efficiency of the LightGBM algorithm and the classification effect of the fusion signal data by comparing the classification accuracy of the four algorithms for the fusion signal, vibration signal and sound signal respectively;

The diagnosis results of different models are shown in Figure 11. The traditional AdaBoost algorithm has poor classification effect, and its diagnostic accuracy for the three data sets is more than 10% lower than that of the other three algorithms. The classification effects of GBDT, XGBoost and LightGBM algorithms on the three data sets are not much different, but the diagnostic accuracy of the LightGBM algorithm is 0.17%-0.33% higher, and the diagnostic results of the three algorithms for the fusion data set are more accurate. Combined with the training time of different models in Figure 12, it can be seen that the training time of the AdaBoost algorithm is the shortest, but its diagnostic accuracy is low. The diagnostic accuracy of the GBDT, XGBoost and LightGBM algorithms is similar, but the training time of LightGBM is 6 times and 3 times shorter than that of the GBDT and XGBoost algorithms, respectively. Therefore, the LightGBM algorithm has higher diagnostic accuracy and higher efficiency.

The vibration signal and sound signal are denoised using the wavelet soft threshold algorithm, and the eigenvalues are extracted and fused according to the corresponding data labels to generate a fused signal data set. Each data set contains normal and three fault types, but the number of samples does not change; Figure 13 shows the fault diagnosis results of the LightGBM model for the fused signal, vibration signal and sound signal, and introduces accuracy, precision, recall and f1-score as evaluation indicators of classification performance; the LightGBM model is used to train and diagnose the three data sets. By comparison, it can be seen that the diagnostic accuracy of the fused data set is 4.86% and 14.59% higher than that of the vibration and sound signals respectively; the diagnostic results of the precision, recall and f1-score of the fused data set are also 4.5%-13.25% higher than those of the vibration and sound data sets;

ROC and PR curves are used to evaluate the classification performance of machine learning algorithms for a given data set. Each data set contains a fixed number of positive samples and negative samples. The larger the area under the curve, the better the classification performance of the machine learning algorithm for the data set. FPR refers to the probability that a negative sample is predicted to be a positive sample, and TPR refers to the probability that a positive sample is predicted to be a negative sample. Precision refers to the proportion of samples predicted to be positive to positive samples, and Recall refers to the proportion of samples predicted to be positive to positive samples. The classification effect of the vibration, sound and fusion signal data sets based on the LightGBM algorithm is shown in Figure 14. As can be seen from the figure, the classification effect of the LightGBM algorithm on the sound data set is the worst, and the classification effect on the fusion data set is much higher than that of the vibration and sound data sets.

The present invention provides a fusion signal hydraulic motor fault diagnosis method. The plunger fault is simulated by an inner curve radial piston hydraulic motor test bench, and the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear are collected. The wavelet soft threshold algorithm is used for noise reduction processing, and feature selection is performed based on data correlation and feature importance. The time domain and frequency domain eigenvalues are extracted for feature fusion. A LightGBM motor plunger fault diagnosis model is constructed, and the fusion features are divided into a data set containing normal and three fault signal labels for training and diagnosis. Finally, the superiority of this method is verified by comparing different algorithms and different signals. The fusion feature fault diagnosis method proposed in the present invention solves the problem of low diagnostic accuracy when using a single sensor to collect data. By using two different signals for diagnosis, the fault information of different sensors can be combined, and the fault features can be increased to improve the fault recognition rate. The fault diagnosis model is established using LightGBM. Under the premise of ensuring classification accuracy, it not only reduces the running memory usage but also increases the model training time, greatly improving the fault diagnosis efficiency.

The fusion signal hydraulic motor fault diagnosis system provided by the present invention includes an inner curve radial piston hydraulic motor test bench for simulating normal and different degrees of wear faults of the piston, and is equipped with a vibration and sound signal acquisition device for acquiring the corresponding signals of the motor in various states. It includes a signal processing module, which uses a wavelet soft threshold algorithm to perform noise reduction processing on the collected vibration and sound signals to improve the signal quality for more accurate fault diagnosis. It includes a feature extraction and fusion module, which is used to extract time domain and frequency domain feature values from the noise-reduced vibration and sound signals based on data correlation analysis, and perform feature fusion to form a comprehensive feature set for fault diagnosis. It includes a motor plunger fault diagnosis model construction module based on the LightGBM algorithm, which is used to train and diagnose a fused feature data set containing normal and three fault signal labels, thereby realizing accurate diagnosis of hydraulic motor plunger faults. It includes a model evaluation and verification module, which uses evaluation tools such as confusion matrix, PR curve and ROC curve to evaluate and verify the performance of the diagnostic model, including the calculation of indicators such as accuracy, precision, recall rate and f1-score, as well as comparative analysis of different algorithms and signal types to verify the superiority of the diagnostic method.

The connection relationship and working principle of the fusion signal hydraulic motor fault diagnosis system provided by the present invention are as follows:

1) Fault simulation component: This component includes an inner curve radial piston hydraulic motor test bench, which is used to simulate the normal state and different degrees of wear failure of the piston. It also collects the corresponding vibration and sound signals of the motor in various states.

2) Signal Processing Component Module: This module receives the vibration and sound signals from the fault simulation component. It then uses the wavelet soft threshold algorithm to perform noise reduction on these signals to improve the signal quality and make it more suitable for subsequent fault diagnosis.

3) Feature extraction and fusion module: This module receives the processed signals from the signal processing component module. Based on data correlation analysis, it extracts feature values from the processed time domain and frequency domain signals and performs feature fusion to form a comprehensive feature set for fault diagnosis.

4) Fault diagnosis model module: This module receives the comprehensive feature set from the feature extraction and fusion module. Then, it builds a motor plunger fault diagnosis model based on the LightGBM algorithm, using the fused feature dataset containing normal and three fault signal labels for model training and diagnosis.

5) Performance Verification Component Module: This module is used to evaluate and verify the performance of the diagnostic model. It uses tools such as confusion matrix, PR curve and ROC curve to calculate the accuracy, precision, recall and f1-score indicators of the diagnostic model. In addition, it also verifies the superiority of the diagnostic method through comparative analysis of different algorithms and signal types.

The workflow of the whole system is as follows: First, the fault simulation component simulates and collects the vibration and sound signals of the hydraulic motor. Then, the signal processing component module performs noise reduction processing on these signals. Next, the feature extraction and fusion module extracts and fuses features from the processed signals. After that, the fault diagnosis model module uses these features to train and diagnose the model. Finally, the performance verification component module evaluates and verifies the performance of the diagnosis model and verifies the superiority of the diagnosis method. It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person skilled in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and/or contained in a processor control code, for example, such codes are provided on a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., and can also be implemented by software executed by various types of processors, or by a combination of the above hardware circuits and software, such as firmware.

In order to solve the problem of low accuracy in fault diagnosis using single signal data, improve the diagnostic efficiency of inner curve radial piston hydraulic motor plunger wear fault, and ensure the stable operation of the hydraulic system, a hydraulic motor plunger wear fault diagnosis method based on LightGBM to identify the fusion features of vibration and sound signals is proposed; by collecting the vibration signals and sound signals of the motor when the plunger of the fault diagnosis test bench is normal and three different degrees of wear, feature selection is performed based on data correlation and feature importance, and time domain and frequency domain eigenvalues are extracted for feature fusion; a LightGBM motor plunger fault diagnosis model is constructed, and the fused features are divided into a data set containing normal and three fault signal labels for training and diagnosis. Finally, the superiority of this method is verified by comparing different algorithms and different data sets; the specific conclusions are as follows:

Build an inner curve radial piston hydraulic motor plunger fault test bench to collect the vibration and sound signals of the plunger when the motor is running normally and in three different degrees of wear state;

The wavelet algorithm is used to reduce the noise of the two signals, and the time domain and frequency domain eigenvalues of the signals are extracted to fuse the same category data to generate a fused data set;

A LightGBM fault diagnosis model was built and compared with the AdaBoost, GBDT, and XGBoost models. The results showed that the fault diagnosis accuracy of the LightGBM model was 0.17%-0.33% higher, and the training time was shortened by 3-5 times, which verified that the LightGBM model has higher diagnostic efficiency.

The comparison shows that the diagnostic accuracy of the LightGBM model for the fusion signal dataset is 4.86% and 14.59% higher than that of the vibration signal and sound signal, respectively. The evaluation indicators such as precision, recall and f1-score are also 4.5%-13.25% higher than those of the vibration and sound datasets.

Through ROC and PR curve analysis, it can be seen that the classification effect of the LightGBM algorithm on the fusion data set is better than that on the vibration and sound data;

The method proposed in the present invention has achieved the expected effect in the fault diagnosis of plunger hydraulic motors, but there are still some shortcomings. Future research should also consider the following issues: 1) When collecting sound signals, the sound sensor is directly exposed to the environment, and a soundproof cover can be added to reduce the interference of environmental noise; 2) In the experiment, each group of signals was collected for 6 minutes, and there were not many data points. The number of samples can be increased to improve the accuracy of fault diagnosis.

The present invention provides a fusion signal hydraulic motor fault diagnosis method. The plunger fault is simulated by an inner curve radial piston hydraulic motor test bench, and the vibration signal and sound signal of the motor when the plunger is normal and three different degrees of wear are collected. The wavelet soft threshold algorithm is used for noise reduction processing, and feature selection is performed based on data correlation and feature importance. The time domain and frequency domain eigenvalues are extracted for feature fusion. A LightGBM motor plunger fault diagnosis model is constructed, and the fusion features are divided into a data set containing normal and three fault signal labels for training and diagnosis. Finally, the superiority of this method is verified by comparing different algorithms and different signals. The fusion feature fault diagnosis method proposed in the present invention solves the problem of low diagnostic accuracy when using a single sensor to collect data. By using two different signals for diagnosis, the fault information of different sensors can be combined, and the fault features can be increased to improve the fault recognition rate. The fault diagnosis model is established using LightGBM. Under the premise of ensuring classification accuracy, it not only reduces the running memory usage but also increases the model training time, greatly improving the fault diagnosis efficiency.

The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims (7)

1. A fusion signal hydraulic motor fault diagnosis method, which is characterized by: first, simulating a plunger fault through an inner curve radial plunger hydraulic motor test bench, and collecting the vibration and sound signals of the motor under normal conditions and different degrees of wear. ; Secondly, the wavelet soft threshold algorithm is used to denoise these signals, and then based on data correlation analysis, feature values are extracted from the time domain and frequency domain, and feature fusion is performed; then, a motor plunger fault diagnosis based on LightGBM is constructed model, and use a data set containing normal and three fault signal labels for model training and diagnosis; finally, through comparative analysis of different algorithms and different signal types, the effectiveness and superiority of the method are verified. 2. The fusion signal hydraulic motor fault diagnosis method according to claim 1, characterized in that it includes an inner curve radial plunger hydraulic motor test bench for simulating plunger faults, including normal conditions and different degrees of wear conditions, Collect relevant vibration and sound signals to provide original data for subsequent fault diagnosis. 3. The fusion signal hydraulic motor fault diagnosis method according to claim 1, characterized in that the collected vibration and sound signals are subjected to noise reduction processing through a wavelet soft threshold algorithm to effectively remove noise interference in the signal. 4. The fusion signal hydraulic motor fault diagnosis method as claimed in claim 1, characterized in that, based on data correlation analysis, time domain and frequency domain eigenvalues are extracted from the processed vibration and sound signals, and these eigenvalues are Fusion to form a comprehensive feature set for more accurate fault diagnosis. 5. The fusion signal hydraulic motor fault diagnosis method as claimed in claim 3, characterized in that a motor plunger fault diagnosis model based on the LightGBM algorithm is constructed, and a data set containing normal and three fault signal labels is used for model training and diagnosis, and through comparative analysis of different algorithms and different signal types, the effectiveness and superiority of the method are verified. 6. A fusion signal hydraulic motor fault diagnosis system based on the method of claim 1, characterized in that it includes: Fault simulation component: includes an inner curve radial piston hydraulic motor test bench, which is used to simulate the normal state of the plunger and different degrees of wear failures, and collect corresponding vibration and sound signals; Signal processing component module: Use wavelet soft threshold algorithm to perform noise reduction processing on the collected signals to improve signal quality for subsequent analysis; Feature extraction and fusion module: Based on data correlation analysis, feature values are extracted from the processed time domain and frequency domain signals, and feature fusion is performed to form a comprehensive feature set for fault diagnosis; Fault diagnosis model module: Construct a motor plunger fault diagnosis model based on the LightGBM algorithm, and use a fusion feature data set containing normal and three fault signal labels for model training and diagnosis; Performance verification component module: Contains tools for evaluating and verifying the performance of diagnostic models, including confusion matrices, PR curves and ROC curves, as well as comparative analysis of different algorithms and signal types to verify the effectiveness and superiority of the proposed method. 7. A fusion signal hydraulic motor fault diagnosis system based on the method of claim 1, characterized in that it includes: An inner curve radial piston hydraulic motor test bench is used to simulate normal and varying degrees of wear failure of the plunger. It is also equipped with a vibration and sound signal acquisition device to collect the corresponding signals of the motor in various states; A signal processing module uses wavelet soft threshold algorithm to perform noise reduction processing on the collected vibration and sound signals to improve signal quality for more accurate fault diagnosis; A feature extraction and fusion module for extracting time domain and frequency domain feature values from denoised vibration and sound signals based on data correlation analysis, and performing feature fusion to form a comprehensive feature set for fault diagnosis; A motor plunger fault diagnosis model building module based on the LightGBM algorithm, used to train and diagnose a fused feature data set containing normal and three fault signal labels, thereby achieving accurate diagnosis of hydraulic motor plunger faults; A model evaluation and verification module that uses confusion matrix, PR curve and ROC curve evaluation tools to evaluate and verify the performance of the diagnostic model, including the calculation of accuracy, precision, recall and f-score indicators, as well as through different algorithms Comparative analysis with signal types to verify the superiority of the diagnostic method.