Support vector machine-based Grassmann manifold distance for health monitoring of viscoelastic sandwich structure with material ageing

• Traditional SVM is extended to the domain of vibration manifold analysis.
• Grassmann kernel-based SVM classifier with parameter optimization is presented.
• A novel manifold distance is designed as margin in Grassmann kernel-based SVM.
• A SHM method for ageing assessment in VSS is proposed.

Subspace analysis is an effective way for Structural Health Monitoring (SHM). In current research, linear algorithms for single-subspace analysis are commonly utilized. Nonlinearity of the structure and characteristics of subspace distribution are ignored. To overcome these shortcomings, characteristics of subspace set are analyzed and a nonlinear subspace distance is defined for SHM in this paper. To calculate this distance index, vibration response signals are firstly monitored and system subspaces are extracted by subspace analysis method. Then, subspace set is viewed as a Grassmann manifold, and the manifold is modeled by Grassmann kernel-based SVM classifier to describe its nonlinear characteristics. Finally, margin in SVM classifier modeled from Grassmann manifolds corresponding to structural normal state and abnormal state, respectively, is defined as a SHM index. This index indicates the degree of the abnormal state deviating from the normal state, and it is an effective index to reflect structural states. Effectiveness of the SHM index is validated by testing data of a Viscoelastic Sandwich Structure (VSS) with viscoelastic sandwich subjected to accelerated ageing in a thermal-oxygen ambient. Analysis result shows that the designed index is very effective to indicate health state in the VSS.

Figure optionsDownload as PowerPoint slide