Machine performance and condition monitoring using motor operating parameters through artificial intelligence techniques

Condit ion m onitoring (CM) of gearboxes is a necessary act ivit y due to the crucial im portance of gearboxes in power t ransm ission in m ost indust rial applicat ions. There has long been pressure to im prove m easuring techniques and develop analyt ical tools for early fault detect ion in gearboxes. This thesis develops new gearbox m onitoring m ethods by dem onst rat ing that operat ing param eters (stat ic data) obtained from m achine cont rol processes can be used, rather than param eters obtained from vibrat ion and acoust ic m easurem ents. Such a developm ent has im portant im plicat ions for the future of CM techniques because it could great ly sim plify the m easurem ent process. To m onitor the gearbox under different operat ing and fault condit ions based on the stat ic data, three art ificial intelligence (AI ) t echniques: a general regression neural network (GRNN) , a back propagat ion neural network (BPNN) , and an adapt ive neurofuzzy inference system (ANFI S) have been used successfully to capture nonlinear variat ions of the elect r ic m otor current and cont rol param eters such as load set t ings and tem peratures. The three AI system s are taught the expected values of current ; load and tem perature for the gearbox in a given condit ion, and then m easured values obtained from the gearbox with a known fault int roduced are assessed by each of the AI m odels to indicate the presence of this abnorm al condit ion. The experim ental results show that each of GRNN, BPNN and ANFI S are adequate and are able to serve as an effect ive tool for gearbox condit ion m onitoring and fault detect ion. The m ain cont r ibut ions of this study is to exam ine the perform ance of a m odel based condit ion m onitoring approach by using just operat ing param eters for fault detect ion in a two stage gearbox. A m odel for current predict ion is developed using an ANFIS, GRNN and BPNN which captures the com plicated interrelat ions between m easured variables, and uses direct com parison between the m easured and predicted values for fault detect ion. 7 DEGREE OF DOCTOR OF PHILOSOPHY (PHD) Machine Performance and Condit ion Monitor ing Using Motor Operat ing Parameters Through Art if icial I ntelligence Techniques _____________________________________________________________________________________________________