A Non-Parametric Approach toward Structural Health Monitoring for Processing Big Data Collected from the Sensor Network

Abstract

Advanced sensor network techniques recently allow collecting large amounts of data for structural health monitoring and damage detection, while how to effectively interpret these complex sensor data to technical information poses many challenges. This paper presents a machine learning algorithm for processing of big data collected from the sensor networks of civil structure. The proposed approach consists of training and monitoring phases. The training phase was focused on the extracting statistical features and conducting Moving Kernel Principal Component Analysis (MKPCA) in order to derive the damage sensitive indices. The monitoring phase included tracking of errors associated with the derived models. The main goal was to analyze the efficiency of the developed system for health monitoring of the benchmark experimental data with the 17 different damage scenarios. In this paper KPCA has been implemented in a new form as Moving KPCA (MKPCA) for effectively segmenting large data and for determining the changes, as data are continuously collected. Numerical results revealed that, the proposed health monitoring system has a satisfactory performance for the detection of the damage scenarios in three-story frame aluminum structure. Furthermore, enhanced version of KPCA methods exhibited significantly improvement in sensitivity, accuracy and effectiveness over conventional methods.