Search Strategies for Identifying a Composite Plate Delamination Using Built-In Transducers

Abstract

Search strategies are compared that can potentially identify single, rectangular, midplane delaminations in composite plates using the signals from surface-mounted piezoceramics. The strategies are demonstrated by repeatedly assuming a single delamination and then running a delaminated plate model, also described, that predicts transducer frequency responses. An objective function compares the responses to those for a hypothetical plate with unknown damage. When the objective function is minimized, the responses agree, and the assumed delamination is the best-estimate. The chosen objective function is well-behaved but has several local minima; therefore, rardom search strategies are used to find the global minimum. The most efficient strategy starts with assumed delaminations selected from a uniform random distribution at discrete points in the parameter space (i.e., a combinatorial search). Some of these points will be close to local minima. The strategy continues by using a random walk with shrinking step size across the parameter space. This latter search is made more robust by using probabilistic hill climbing acceptance tests, also known as simulated annealing. The overall technique systematically investigates several possible local minima and identifies a best-estimate delamination to a preset accuracy.