Overview: In the U.S. each year, natural hazards such as earthquakes and hurricane cause hundreds of deaths and cost tens of billions of dollars due to the destruction of buildings and infrastructures. The objectives of systems health monitoring (SHM) are to reduce the risk of loss from natural disasters by understanding systems damage/failure mechanism, predicting system performance during disasters, and designing more effective disasters-resistant structural systems, and ultimately to minimize the resulting losses by developing robust, innovative mitigation measures. However, data imperfection (e.g., uncertainty and imprecision) and information reliability are still critical issues yet to be addressed in an efficient SHM system. Ignoring these issues will result in an expensive and ineffective SHM design that adversely impacts the real-time condition assessment, damage detection, failure prevention, and hazards mitigation. The overall aim of the research is to create a comprehensive framework for data processing and decision-making in health monitoring of structural systems, considering data imperfection, information uncertainty and sensor reliability. A general methodology will be developed through the integration of soft-computing concepts (stochastic neural networks and fuzzy logic), wavelet-based signal processing, information theory, Bayesian hypothesis testing, and stochastic finite element-based structural analysis.
Intellectual Merit: The key contribution of the research is the development of Bayesian algorithms for various steps in SHM to address imperfection, uncertainty and reliability, and effective integration of both system design and data processing methodologies. Several new computational methods will be developed including: (1) combination of stochastic finite element analysis with optimization and Bayesian statistics for sensor placement design, considering sensor reliability and information uncertainty; (2) a dynamic fuzzy stochastic neural network damage detection model, which incorporates information theory and Bayesian hypothesis testing for model assessment and explicitly considers uncertainty and imprecision; (3) a Bayesian risk-based decision methodology for damage evaluation and extrapolation from the imperfectly sensed data, and (4) a multi-criteria decision-making framework for structural health monitoring based on a Bayesian decision network methodology.
Broader Impact: Multi-criteria decision-making for systems health monitoring under uncertainty using imperfectly sensed data has numerous real life applications in civil infrastructure systems as well as other fields such as aerospace industry in the context of condition assessment and damage diagnosis of aging aircrafts. The benefits will be demonstrated through two specific applications in civil and aerospace structures. The research will
- create a next generation decision-making methodology for SHM considering data imperfection, uncertainty and sensor reliability;
- reduce subjectivity by developing a quantitative approach to damage detection, and maintenance of structural safety and reliability, and
- facilitate proactive, rational decisions with respect to maintenance and repair, instead of reactive, subjective decisions based on external, visual symptoms.
As a result, the research will help in significantly reducing inspection and repair costs, and thus life cycle costs, while significantly helping to increase the longevity of infrastructure systems.