| Created: 10/24/95 | Updated: 4/18/97 |
| Objective | Background | Work Accomplished |
Assignable causes can be clearly analyzed in terms of failure mechanisms, failure modes and failure sites. Non-assignable causes are described as can-not-duplicate CND, retest OK, cannot verify, or unknown causes, and are responsible for up to 70 percent of failures, primary reason being the lack of knowledge about the field conditions. The result is that the potential failure mechanisms may be overlooked in the design process. In terms of root cause failure analysis, incompatibility of the testing conditions and working conditions may make failure detection impossible. The detection and diagnosis of potential non- assignable, as well as assignable, failures in electronic systems demands (1) defect growth characterization, (2) on-line, nondestructive health management techniques, and (3) failure models from which to assess remaining life. Health management will characterize degradation and on-line diagnostic techniques in conjunction with the physics-of-failure approach predict the reliability of the system.
Previous work (C95-20) developed a Health Management and Early Detection
Fuse (HMEDF) for conductive filament formation (CFF). To assess feasibility
of the approach for failures in electronic systems, critical failures and
their symptoms were identified. Failures were classified and ranked according
to the severity and frequency of their occurrence. Following is the approach
for achieving successful health management of the system.
