C05-29 Quantitative Assessment of Uncertainties in Health Monitoring of Electronics
Objective:
- Identify uncertainties in the CALCE health monitoring methodology for electronics.
- Assess the impact of identified uncertainties in sensing, recording, and physics of failure modeling on the remaining life predictions.
Background:
For the past five years, CALCE has been conducting research on health monitoring of electronic systems, addressing areas that include Built-in Test (BIT), in-situ semiconductor health monitors, and Life Consumption Monitoring (LCM). CALCE¡¯s LCM methodology is a six-step process, figure 1, that combines systematic study of the different failure modes and mechanisms of the product under consideration, monitoring of relevant environmental and/or operational parameters, and use of physics of failure models to asses the damage and ultimately predict remaining life. This methodology has been successfully demonstrated for an electronic board operated in an automotive underhood environment.
For completeness, the uncertainties associated in the measurement and procedures within the prognostic and health monitoring process should be investigated. Uncertainties may arise from input data (due to measurement errors, physical randomness, use of sampled information) and models (due to limitations in analyst¡¯s knowledge of phenomenon, deliberate simplifications introduced for modeling, assumptions in forecasting procedures and damage accumulation methods). The assessments of these uncertainties assist in effective scheduling of maintenance and cost analysis
In C04-19 and its continuation projects, CALCE is working on integrating hardware with an embedded software solution that can enable monitoring and assessment of electronic products in their application environment. As outlined in CALCE¡¯s health monitoring roadmap, figure 2, over the next few years, the capability of the health and usage monitoring system (HUMS) prototype to predict the remaining life of solder joints due to failure under thermal cycling and vibration loading conditions will be validated. Methods for humidity and contamination assessment will also be explored. In addition, in-situ ¡°canary¡± (consumable) devices (P05-P1) will be studied and calibrated for supporting prognostic health monitoring. Their variability in predicting remaining life will be assessed, and their accuracy compared to that of LCM. In parallel with these efforts, potential sources of uncertainty in life predictions of electronics, derived using health monitoring methodologies, will be analyzed and quantified.Approach:
The proposed project will initiate a two-year effort to assess uncertainties in the implementation of prognostics and health management systems in electronic equipment. The first year effort focuses on identifying potential sources of uncertainties based on both current and past prognostic and health management activities (C04-19, C05-28). Sources may include inaccuracies in measurement, in terms of sensor limitations, measurement technique, errors in signal conditioning and data collection, inherent limitations in the PoF models, and inaccuracies in damage accumulation and inference. The identified sources of uncertainty will be reviewed, and their potential impact examined. Uncertainty analysis will incorporate standard techniques for uncertainty modeling, including probabilistic tools, possibility theory, evidence theory, and fuzzy set theories. The primary output of the project is expected to define the boundary of uncertainty and its impact on prognostics and health management systems.
In the second year, methods for handling uncertainties when implementing prognostic and health management systems will be developed. These methods will be demonstrated with respect to on-going prognostic and health management projects (C05-28).Reports:
Final Project Report