|
Title |
Development of Noise Reduction Methods for a Wireless Ultrasonic Structural Monitoring System
|
|
Authors |
정현우(Hyun-Woo Jung) ; 최승국(Seung-Guk Choe) ; 김창혁(Changhyuk Kim) |
|
DOI |
https://doi.org/10.4334/JKCI.2026.38.2.155 |
|
Keywords |
초음파 센서; 다층 퍼셉트론; 스파이크 노이즈 필터링; 무선 시스템 ultrasonic sensor; multilayer perceptron; spike-noise filtering; wireless system |
|
Abstract |
In this work, a wireless miniaturized structural deformation measurement system was developed using ultrasonic sensors to overcome the limitations of conventional wired systems. Wireless communication was achieved using Wi-Fi modules, along with compact ultrasonic sensors and Arduino boards, enabling both wireless operation and hardware miniaturization. To reduce spike noise inherent in ultrasonic distance measurements, six filtering techniques, including the Kalman filter were applied. In addition, a Multilayer Perceptron (MLP) model was employed to improve prediction performance using various combinations of these filters. Bayesian optimization and K-fold cross-validation were conducted to prevent overfitting and to determine optimal hyperparameters. As the number of filter combinations was gradually expanded and compared, the MLP model combining ultrasonic data with four filters (Kalman, Exponential Moving Average, Savitzky?Golay, and Gaussian) achieved the lowest prediction error. Additional validation using a simplified test specimen produced consistent results, confirming the effectiveness and reproducibility of the proposed approach. Through this approach, spike noise was effectively reduced, and the reliability of the measurement data was improved.
|