Research On Fault Diagnosis And Performance Prediction System Of Air Compressor Based On Data-driven

In recent years,with the increasing speed of modern industrial development,artificial intelligence and industrial Internet have gradually become important technical means for the development of equipment management under the background of the Industry 4.0 era.With the advancement of smart pipeline construction,"full data perception" has become a basic requirement for smart pipelines.The compressor unit is the core power equipment of the gas pipeline.As an important auxiliary system of the compressor unit,the air compressor's operational reliability will directly affect the gas supply stability of the compressor station.How to improve the operating reliability of the air compressor,find the operating failure of the air compressor timely and accurately,and reduce the energy consumption and maintenance cost of the air compressor is an urgent problem in the engineering field.This article regards air compressor as the key equipment research object,focusing on its fault diagnosis and performance prediction research.The main contents of the paper are as follows:(1)First of all,summarizes the operating mechanism,mechanical structure,failure types and corresponding solutions of the oil-injected screw air compressor.Give a detailed introduction to the research background,research significance,research content,and current status of the subject at home and abroad,mainly analyze and introduce air compressor remote monitoring technology,deep learning algorithm and development trend.(2)Secondly,design the air compressor fault diagnosis and performance prediction system scheme.First,the air compressor data acquisition and transmission scheme was determined.After comparing the advantages and disadvantages of the two schemes of adding a serial port hub and wireless transmission,combined with the actual project,finally chose to use wireless The transmission scheme not only ensures the data transmission,but also does not affect the existing communication link and the remote control of the air compressor.The air compressor remote monitoring and fault diagnosis system was designed from the two aspects of the system function and structure,and finally the three main functions of the system were determined: data collection,transmission and storage,data monitoring and release,operation data diagnosis and prediction.In terms of system structure,the software structure and hardware structure are introduced in detail,and finally the interface of the system's live operation is displayed.(3)Used the C# programming environment to realize the remote centralized monitoring and fault diagnosis system of the air compressor.This system is composed of relevant hardware and software,SCADA network,neural network algorithm and man-machine interface and other modules,which can perform real-time remote centralized monitoring of air compressors.At the same time,used the artificial intelligence technology to conduct in-depth mining of the operating data of the air compressor to realize the comprehensive alarm,fault diagnosis and performance prediction of each operating index of the air compressor,and to grasp the operating energy consumption and health status of the air compressor in time.Discover the potential failures of the air compressor in advance,give maintenance solutions and energy-saving suggestions,reduce the operation and maintenance costs of the air compressor,improved the reliability of air compressor operation.(4)Analyze the current BP neural network's shortcomings in fault diagnosis and prediction,give a neural network model based on the LM-BP algorithm,and introduce a comprehensive health-based air compressor state evaluation method,and combine it with the neural network Machine for combined diagnosis.Firstly,comprehensively evaluate the operating state of the air compressor through the comprehensive health evaluation model and give the evaluation value,and then perform the next step of neural network diagnosis for the air compressor below the hidden danger threshold,and finally give a detailed diagnosis result.Compared with the diagnostic invalidity that may be caused by the neural network model's one-by-one diagnosis of the data and the unity of the comprehensive health evaluation model,the combined result is more direct,accurate and efficient.(5)Established an ARIMA time series forecasting model with an oil-injected screw type as the main body.Finally,the established model is verified by an example,and the error of the actual operating value of the comparison system is controlled below 8%,and the prediction effect is good.