Long term aging of electronics systems & maintainability strategy for critical applications

Many electronic systems, such as computers and consumer goods, have a short life, typically two to five years. However, there are critical applications, e.g. in the case of power plants, where life may be 20 years or more. While proactive maintenance is normally practiced for mechanical components, the authors have not found references in the technical literature that address the question as to whether or not proactive maintenance for electronic systems may be warranted in cases of very long term usage. Electronics reliability models that are commonly used include MIL-HDBK 217F and Telcordia Reliability model. In the former model failures are assumed to be "random", i.e. exponentially distributed. A model such as this has no "infant mortality" or "wearout" region. The latter, Telcordia, model does add a "First year multiplier" to account for infant mortality but no factors for well-aged components. Physics of failure approaches come closest to describing the "life" of electronic products. They are useful in cases where life is limited by predictable physical mechanisms. These models work well, e.g., for the case of solder joint fatigue by thermal cycling. However, POF approaches are not adequate for large systems, with disparate part types, where the environmental conditions are benign. In the authors' applications, units are operated in a control room with carefully controlled environmental conditions. So, they conducted a large measurement program on samples of aged control circuit cards to determine what components age, and by how much, over 20 years or more. They used the results, in concert with additional analysis, to determine a maintenance strategy for older control systems.