Fault Prediction And Diagnosis Of Compressor On Cloud Based On LSTM

With the social economic developing the demand of industry has risen rapidly.As an important power equipment,the number of compressors also increases.Its safe and efficient uninterrupted operation is very important to the production process of the whole plant.In order to avoid the heavy loss caused by the failure of the compressor,the real-time monitoring and fault prediction and diagnosis of the compressor has become an indispensable work,which is of great significance to ensure the development of the national economy.Firstly,the structure and working principle of centrifugal compressor are explained,and the common faults and detection methods are introduced.For the electric drive centrifugal compressor,a fault detection method based on motor current analysis(MCSA)is proposed.Secondly,the architecture of the compressor operation monitoring system based on cloud is designed.The design and implementation of the software and hardware design of the front-end acquisition module and the central service software are described respectively,and the system interface is displayed.Thirdly,based on the cloud platform data of the plant,the demand of intelligent monitoring is proposed.The compressor operation prediction model based on long-term and short-term memory network(LSTM)neural network can effectively highlight the frequency spectrum caused by fault by comparing difference between the actual and predicted data waveforms,extract the fault feature vectors of vibration and motor stator current signals,and then through the support vector(SVM)is used to classify fault feature vectors.Finally,the compressor detection cloud platform and intelligent monitoring function are integrated and the workflow is sorted out to take the advantages of the combination of big data of equipment operation and intelligent predictive diagnosis.This thesis takes the actually work characteristics of the compressor into a full consideration,and designs a complete compressor on cloud fault prediction and diagnosis scheme based on deep learning,which guarantees effectively that the compressor works safely,efficiently and stably.