Fault Diagnosis And Prognosis For Diesel Generator Sets Based On GRU Model

Fault diagnosis and prognosis technology has improved to a certain extent the insufficiency of regular maintenance of industrial equipment and diagnosis only after faults occur.Considering that diesel generator sets have a wide range of industrial application,and many sensors have been used,but based on the threshold of the sensor signal to determine whether the equipment is malfunctioning,there are limitations such as the inability to effectively determine the cause of the malfunction and identify certain faults.In addition,the sensor signal is time-dependent and contains predictability.Taking appropriate methods to analyze such data can improve the accuracy and efficiency of fault diagnosis and prognosis methods.In view of the above reasons and the rapid growth of deep learning,this article conducts research based on the actual work background of the Master of Engineering.First,on the basis of studying the status quo of diagnosis and prediction technology,a recurrent neural network(RNN)with good classification and prediction capabilities for time series data is selected,and the working principle and performance of the diesel generator sets selected in the experiment are understood to establish the equipment Mathematical model,then use Simscape to establish Simulink simulation model of main system of the diesel genset.Then,according to the common failure of the experimental equipment,some typical faults are set in the simulation model,and a dataset for deep learning model training and test is generated.In the data analysis and processing stage,the PyTorch machine learning library is used to design and verify the Gated Recurrent Unit(GRU).Compared with other fault diagnosis and prediction methods,it has the advantages of high accuracy and simple model.On the premise of verifying the good effectiveness of GRU for diagnosis and prognosis,this research result is applied to actual equipment.By installing an electronic control module that supports industrial bus and wireless communication,remotely obtain sensor data and operating parameters,and performing fault diagnosis and prediction,the failure rate of the equipment is reduced to a certain extent,and the operating efficiency and service life are improved.In addition,this is a data-driven method,which has good applicability,and can be expanded to other equipment or fields.