Optimal sensor distance for damage detection considering wavelet sensitivity and uncertainties

10th Asia Pacific Structural Engineering and Construction Conference 2018

工业技术（总论）

Abdulkareem, Muyideen^1 ; Kasali, Mujedu^2^3 ; Martins, Esan^2^4 ; Nathaniel, Olukotun^2^5 ; Abd Majid, Muhd Z.^1

Construction Research Centre, Universiti Teknologi Malaysia, Johor, Malaysia^1

School of Civil Engineering, Universiti Teknologi Malaysia, Johor, Malaysia^2

Department of Civil Engineering Federal Polytechnic, Osun State, Ede, Nigeria^3

Department of Building Technology, Federal Polytechnic, Osun State, Ede, Nigeria^4

Department of Building Technology, Kogi State Polytechnic, Kogi State, Nigeria^5

关键词: Computational time;    Continuous wavelet transforms;    Damage detectability;    Effect of noise;    False detections;    Inadequate data;    Sensor placement;    Time and frequency domains;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/513/1/012018/pdf
DOI  :  10.1088/1757-899X/513/1/012018

学科分类：工业工程学

来源: IOP

【 摘 要 】

Wavelet Transform (WT) is an efficient signal processing tool used extensively to detect damage in various types of structures. It is capable of describing signals in both time and frequency domains. However, improper sensor distance may result to false detection of damage, thus affecting the reliability of the method. A small sensor distance leads to large number of sensors, thus increases the financial cost and causes higher computational time, while large sensor distance may provide inadequate data for accurate damage detection. In this paper, the optimum sensor placement for damage detection is investigated by considering the effect of noise and sensitivity of WT. This involves using Continuous Wavelet Transform (CWT) to decompose the mode shape difference of a plate numerical model. Different measurement distances and different levels of uncertainties are added to the mode shape data to evaluate the optimum sensor distance that provide the best damage detection result. A numerical plate model with all four sides fixed is used as an example. The results indicate that increase of noise reduced the detectability of damage. It is also observed that excessive sensor distance increment significantly effects damage detectability.