An improved D-S evidence fusion algorithm for sub-area collaborative guided wave damage monitoring of large-scale structures

Abstract

Due to the development of new materials and advanced manufacturing technologies, the application of large-scale composite structures has become increasingly widespread. Ensuring the safe and stable operation of such structures presents new challenges across various application domains. Addressing the limitations of existing guided wave structural health monitoring methods for online damage monitoring in large-scale structures, such as cumbersome equipment setup, insufficient signals coverage, and difficulties in processing massive data, a method for sub-area collaborative guided-wave-based structural damage monitoring and severity assessment based on sparse sensing is proposed. By employing a sparse sensing array layout, the structure is divided into multiple monitoring sub-areas with arranged sensing arrays to reduce overall complexity. The characteristic responses of the guided wave signals from different sub-areas are extracted to construct feature sub-spaces. Support vector machines are adopted to construct evaluation sub-networks in each feature sub-space, enabling regional monitoring. Additionally, an improved D-S evidence fusion algorithm is applied to fuse the decision-layer inputs from each evaluation sub-network, effectively utilizing the feature information from multiple sub-areas and enhancing the accuracy of damage severity assessment for large-scale structures. Experimental results on typical composite structure specimens demonstrate that by fusing the support vector machine evaluation results from each sub-area, the accuracy of damage severity assessment reaches 97.5%, with uncertainties in the severity assessment below 5%.