Research And Design Of Motor Fault Diagnosis System For Pipeline Robot

The motor of the pipeline robot is its important power source,and it is necessary to monitor the health of the motor when it is working itself because it is prone to different types of failures due to the unsatisfactory working environment or long working wear and tear,which leads to the failure of the motor to work properly and brings unnecessary losses.The research and design of the pipeline robot motor fault diagnosis system is carried out for the problem that the early stage of fault cannot be warned in time and the experience and ability of human perception of the fault are required high.The system process is mainly divided into motor signal pre-processing,fault diagnosis algorithm training,deployment in embedded devices and real-time display of diagnosis results in the host computer interface.The main work is as follows:(1)Motor signal pre-processing.The health condition of the pipeline robot motor is obtained by analyzing the vibration signal and the sound signal,and the wavelet threshold denoising algorithm is used to improve the signal-to-noise ratio of the motor signal for the problem of noise interference in the motor signal.On the basis of signal denoising,for the original one-dimensional time domain signal only reflects the characteristics of the signal from the time dimension,there is a problem of incomplete characterization of signal features and difficult to distinguish the differences between different signal features,the short-time Fourier transform and wavelet transform are used to transform the motor signal from one-dimensional time domain to two-dimensional time-frequency diagram,combining the characteristics of the signal in both time and frequency domains to effectively improve the fault characteristics of the The ability to express the fault features is effectively improved.Finally,the time-frequency map is divided into a training set,a validation set and a test set to prepare for the subsequent input of the time-frequency map into the convolutional neural network for further extraction of effective features.(2)Fault diagnosis algorithm training.On the basis of signal preprocessing,migration learning is used to speed up the training speed of fault diagnosis models and improve the recognition accuracy of the algorithm in response to the problems of deep learning requiring large amounts of training data and long training time;the pretraining model uses VGG16 and ResNet50 to compare the accuracy of short-time Fourier transform time-frequency maps and wavelet transform time-frequency maps as input to convolutional networks;on the basis of analyzing sound signals and vibration signals,the sound-vibration signal fusion diagnosis algorithm is proposed to enhance the credibility of motor fault detection for pipeline robots,in response to the problem that analyzing the motor condition by a single type of measurement signal may lead to inadequate analysis.The fusion algorithm consists of two convolutional network models,each of which takes the time-frequency maps of vibration and sound as input for feature extraction,and concatenates the sound feature vector and vibration feature vector into a single acoustic-vibration feature vector before output,and connects them to the same classification layer for final prediction.(3)Deploy on embedded devices and display the diagnosis results in real time on the upper computer page.Save the trained fault diagnosis algorithm model and deploy the model on the embedded hardware Raspberry Pi,build a real-time monitoring platform using Raspberry Pi,save the data logs in MySQL database,and develop the upper computer interface to display the health status of the pipeline robot motor in real time.In this thesis,a flexible and efficient motor fault diagnosis system is built.After the overall test,the system can meet the requirements of pipeline robot motor health monitoring,and the recognition accuracy reaches 96.15%,which is conducive to improving the reliability and stability of the pipeline robot.