Robust signal processing for damaged vehicles with variability

The focus of this paper is on establishing a robust signal processing approach for damaged vehicles (i.e., cracked structures) with structural variability such as thicknesses of various components and Young's modulus variations. The approach assumes that vibration-type data is collected during the operation of a vehicle. Next, the collected data is used in a novel combined sensor selection and signal processing methodology. The new methodology resolves two key issues for complex structures with variability: (i) decides which field data channels are statistically optimal to be used, and (ii) establishes which data channels should correlate and how. The overall algorithm is based on a generalised version of the effective independence distribution vector. Also, the correlations among channels are used for noise rejection. Furthermore, the dynamics of the vehicle (i. e., a complex structure with variabilities) is modelled using Parametric Reduced Order Models (PROMs) and the concept of bilinear mode shapes introduced recently by the authors for cracked structures. PROMs are used to address the presence of variability and account for their effects on the data collected from various channels. The bilinear modes are used to capture the effects of the crack. The proposed methodology is demonstrated for a complex/realistic model of a HMMWV frame with parameter variability and a crack.