Inverse identification of impact locations using multilayer perceptron with effective time-domain feature

• Main Authors: Hossain, Md Sazzad, Ong, Zhi Chao, Ng, Siew Cheok, Ismail, Zubaidah, Khoo, Shin Yee
• Format: Article
• Published: Taylor & Francis 2018
• Subjects: R Medicine; TA Engineering (General). Civil engineering (General); TJ Mechanical engineering and machinery
• Online Access: http://eprints.um.edu.my/21868/; https://doi.org/10.1080/17415977.2017.1316496

This study presents a novel time-domain feature to identify impact locations using Multilayer Perceptron. This feature is based on the minimum arrival time (MAT) of surface wave at a particular location of an object due to impact. Trial impact forces were made on a Perspex plate structure and corresponding acceleration responses were acquired from six locations. Two other conventional time-domain features–peak arrival time (PAT) and threshold crossing (TC), were compared with the proposed feature. Each feature was used separately as network inputs to identify 15 fixed impact locations. The results showed that impact localization with MAT feature resulted in the highest accuracy, making a relative decrease in error of 42.06 and 81.04% compared to PAT and TC cases, respectively. A consistency measurement scheme has been developed as well, which indicated that MAT is more consistent than PAT and TC for a particular pair of impact-sensor location, and hence a more accurate localization of impact can be obtained.