Condition Monitoring And Fault Diagnosis Of Pump System

The pump system is widely used in industrial and agricultural production, and it’soperation condition largely influences operation of the whole production system.Condition monitoring and fault diagnosis of the pump system requires monitoring ofelectric signals, as well as vibration, temperature, pressure and flow signals, causing aseries of problems such as investment increase, measuring difficulty, increase of sensorfailure rate, difficulty to accurately measure the flow signals, and signal distortionthrough analog signal transmission over long distance. Therefore, this study putsforward a new method of condition monitoring and fault diagnosis of the pump systemwhich mainly relies on electric signals and with pressure signals. With this method,vibration signals which are easy to distort, and flow signals with low measurementaccuracy could be avoided; the accuracy of fault diagnosis could be improved byimproving the quality of data; and test difficulty and system investment could bereduced by reducing signals to be measured. Thus it will help the implementation andpromotion of this new fault diagnosis technology of the pump system.First, response of characteristic parameters of the single pump system in normalstate and four fault states to be diagnosed are obtained through theoretical analysis andanalog computation. Select impeller wear, rotating components obstruction, cavitation,and main pipe blockage as fault states to be diagnosed, and build the pump systemmodel in normal state and four fault states. Take a simple single pump system as anexample. Use MATLAB programming language to analog compute its operationconditions in normal state, different kinds of fault states and different degrees of fault,in order to obtain response of characteristic parameters.Second, define response of characteristic parameters of multiple series and parallelpump systems in normal state and that of fault states to be diagnosed of a single pumpsystem. Then build models of series and parallel pump system in normal state and fourfault states. Take the simple series and parallel pump system as an example. UseMATLAB programming language to analog compute its operation conditions in normalstate, different kinds of fault states and different degrees of fault, in order to obtainresponse of characteristic parameters.Third, develop the fault diagnosis method of the pump system based on fuzzycomprehensive evaluation, with the technical objective to do fault diagnosis of thepump system that mainly relies on electric signals and is supplemented by pressuresignals. Then define the evaluation factors with the principle of ensuring the systemoperation state with large range-ability, little measurement difficulty and highmeasurement accuracy. The membership functions and weights of evaluation factors ofthe fuzzy fault diagnosis of the pump system are defined then with the method of initially defining by theoretical analysis and later optimization later. Finally, useMATLAB language to develop a program with graphical user interface for faultdiagnosis of the pump system.Fourth, define the implementation method of real-time condition monitoringsystem of the pump system. Several methods to obtain the inherent parameters of pumpsystem are provided, including test obtaining, neural network prediction and so on.Methods to test and compute the characteristic signals are provided, with samplingfrequency defined and error analysis made. The composition of real-time conditionmonitoring system of the pump system is defined, and the selection methods of sensors,data acquisition preprocessing data analysis, status display and warning device aregiven.This study shows that four given fault can be diagnosed, by measuring andanalyzing electric and pressure signals of the pump system and by using fault diagnosisthat is based on fuzzy comprehensive evaluation.