Study And Designing Of On-Line Detection System To Test Roller's Vibration

With the persistent advancement and increasing development of science and technology,especially the technology of signal measurement,digital signal analysis and processing,computer and software concerned,along with the artificial intelligence such as neural networks and so on,automation technology is gradually plunged into application in many industry manufacturing aspects, more and more proficient,providing powerful tools for machinery's fault diagnosis.In the meanwhile,virtual instrument (Ⅵ) emerges as the times require,thus it combines measurement and control system(MACSYM) with computer software as well as hardware,creating a convenient method for the exploiting MACSYM.Presently,the virtual instrument has been comprehensively used in different industries to control theirs producing process.Roller,the key apparatus unit in spinning machine,widely used mainly to take draft function on fibre strand.Whether it works well or not will determine yarn's quality.However,it still remains the primary stage that roller's faults are mostly diagnosed by operator's hand-touching with theirs experience in textile mill nowadays,which is a process short of impersonal judgment,and in certain extend,holding some limitation.Therefore,it's very significant to develop roller's on-line detection and diagnosis system so as to meet the coming need of textile industry.This subject carries out vibration signal's testing with an eddy current type transducer on old-style evenness tester by simulating its common malfunctions on base of PC-DAQ virtual instrument, and transferring the signal to computer through multifunction data acquisition board.Then,design the control panel and do manifold analysis and processing from time domain to frequency domain to the signal in virtue of software Labview,which is famousⅥ-developing platform possessing so much superiority in programming such as visualization,openness,compatibility etc that it can realize combination with Matlab to expand its powerful operation capability.By comparative analysis,find out each malfunction's characteristic parameters or frequencies and choose several of them advisably to form fault samples for the use of constructing RBF NN with the aid of invoking function Newrbe in Matlab NN toolbox,afterward,train the RBF NN and test its performance with acquired data.The experiment result indicates that the virtual system can be applied to roller unit for theirs on-line inspection and measurement, and also,the constructed RBF NN trained on base of acquired data is able to identify all kinds of roller's fault modes correctly in some accuracy,which,from another point of view,show that RBF neural networks' uniquely advantage in fault diagnosis field.