This paper describes how a bearing skidding defect has been eliminated successfully as a result of adequate selection of bearing components by the use of discriminant analysis and other multivariate procedures. In the first period of engine production, although a high percentage of defects has been reported, we were not sure how and what sort of attributes of the parts to be controlled under the technical viewpoints. Some six dimensions, clearance, and other technical values were measured and controlled with no results. It was assumed that the defects were caused by some attributes related mutually and complexly. It seemed that relying only on inherent technology based on past experience could not resolve these complexly interrelated situations. So, we decided to apply experimentally some of the more statistical and reliable techniques, as follows: Principal component analysis and outliers test for data stratification, selection of variables by prediction sum of squares, discriminant analysis and construction of bivariate control chart to forecast the bearing skidding from assembly data. Consequently, we succeeded to prevent the bearing skidding defects.
Reliability improvement on aircraft engine bearing by discriminant analysis
Zuverlaessigkeitsverbesserung bei Flugmotorlagern durch Diskriminanten-Analyse
Ishikawajima-Harima Engineering Review ; 20 , 1 ; 29-34
1980
6 Seiten, 11 Bilder, 2 Tabellen, 6 Quellen
Aufsatz (Zeitschrift)
Japanisch