Unknown impact force localisation and reconstruction in experimental plate structure using time-series analysis and pattern recognition

Abstract

Impact force is the load that actual structures often experience, and its localisation and time-history reconstruction are important for structural health and safety monitoring. However, classical inversion techniques for random impact-force identification are time-consuming and infeasible; therefore, they cannot be utilized for engineering applications. This study adopted multiple time-series analyses and a pattern-recognition method to identify unknown random impact forces, and this strategy is known as the pattern-recognition method combined with a similarity metric (PRMCSM). Numerical simulations of the proposed method have been conducted; however, their findings have not been experimentally verified. From an experimental perspective, this study considered suspended steel plate as the research object, impact force as the load input, and acceleration as the response output. The PRMCSM algorithm based on cosine similarity was verified, and the expected results were obtained. Moreover, various similarity metric indices are proposed to implement the feature extraction of time-domain responses, and a novel index that is not lower than cosine similarity was finally determined and was verified based on the PRMCSM strategy. In summary, the method based on pattern recognition achieves the rapid identification of random impact forces and exhibits potential for practical engineering applications.