| 摘要 |
Multiple parameters of manufactured units are continually measured until some of the units fail, where failure can be accelerated by adjusting operating conditions. Pre-failure data is then examined to find outliers or aberrant parameter values that may have contributed to the failures. The data is normalized to allow different parameters to be compared to one another. The parameters producing the highest outlier values are then used to screen subsequently manufactured units, thus significantly reducing the number of measurements that have to be taken to screen the units. Lower outlier values for these parameters are, however, used in screening subsequently manufactured units to "catch" potentially defective units.
|
