This grant provides for the development of a science base for the integrated proactive maintenance strategy. Various fundamental issues will be studied, which include affordable in-process sensing optimization and validation with consideration of the global goal of proactive maintenance; degradation modeling, analysis, and prediction by combining in-process data and limited historical information; multi attribute decision making; uncertainty measurement and risk analysis; and global optimal maintenance policy determination and evaluation. The research will be organized into four interrelated tasks: (1) Optimizing strategies for the use of affordable in-process sensing; (2) On-line feature extraction and sensor fusion from multi-dimensional observations; (3) Developing predictive failure models combining in-process sensing data, engineering knowledge, and limited historical data; and (4) Developing a simulation testbed for evaluation of proactive maintenance policy. The central idea is to develop a methodology to achieve proactive maintenance by integrating information (with uncertainty consideration) provided by: (1) Engineering: in-process sensing and on-line modeling, diagnosis, and failure prediction; (2) Statistics; reliability models and self-learning/updating; and (3) Management Science: cost and risk analysis; uncertainty measurement, and performance evaluation. The implementation of proactive maintenance methodologies are expected to significantly improve the maintenance of manufacturing equipment and processes, which has a significant impact on manufacturing quality, productivity, and cost. The success of the research project will contribute to maintenance science by overcoming the barriers among different disciplines, and demonstrating the effectiveness of the integrated methodology on proactive maintenance. In addition, the research will also provide original contribution to each individual discipline's research in areas such as (1) affordable and optimized in- process sensin g and feature extraction techniques; (2) reliability analysis with limited historical data; and (3) Bayesian decision-making and self-learning algorithms for maintenance policy improvements. The research testbed will serve as a basic for future research and demonstration of proactive maintenance.