Kernel Method Used for Multi-sensor Information Fusion in Self-Diagnostic Smart Structures

Smart structures is a multi-sensor architecture, and multi-sensor information fusion is a key technology used to realize the self-diagnosing damages function for smart structures. Due to a few of intrinsic flaws of traditional neural networks, the fusion algorithms based on kernel learning are progressing rapidly in recent years. Based on the approximately orthonormal property of wavelet kernel function, Least Square Wavelet Support Vector Machine (LS-WSVM) used for information fusion is proposed, and its tuning parameters are optimized by Genetic Algorithm (GA). Based on the feature-level fusion, LS-WSVM fusion algorithm is applied to self-diagnose damages for smart structures, and its performance is evaluated through some tests. The results show that LS-WSVM possesses the higher fusion accuracy, bitter dissemination ability than LS-SVM with Gaussian kernel function under the same conditions, and LS-WSVM can decrease the amount of computation in information fusion.