A probabilistic framework for detecting and identifying anomalies

This paper develops a probabilistic framework to detect and identify anomalies such as damage in structures. The framework is developed by introducing new terms and definitions with their corresponding mathematical formulation. An advantage of the new framework is that ill-conditioning in the identification problem is avoided and that a clear relation between measurements and modeling is established. Special results are then obtained in the form of bounds that allow for computationally efficient applications. An example application is then presented. The application is to detect and identify part-through cracks in a plate from surface strain measurements. In this application problem, the role of strain gauge size and measurement errors are considered and discussed.