Environmental effect compensation for damage detection in structures using artificial neural networks and chirplet transform

Abstract

One of the open problems to implement Structural Health Monitoring techni ques based on guided waves in real structures is the interference of the environme ntal effects in the damage diagnosis problem. This paper deals with the compensation of one of the envir onmental effects, the temperature. It is well known that the guided wave form is modified by temperature variation and causes errors in damage diagnosis. This happens because the waveform has an influence due to temperature changes of the same order tan he damage presence, which makes difficult to separate both effects in order to avoid false positives. Therefore it is necessary to quantify and compensate the temperature effect over the waveforms. There are several approaches to compensate the temperature effect such as Optimal Baseline Selection (OBS) or Baseline Signal Stretching (BSS). In this paper, the experimental data analysis consists on applying the Chirplet Transform (CT) to extract Environmental Sensitive Features (ESF) from raw data. Then, the measure of the environmental condition is related with the ESF training an ANN. The relati onship between the temperature and the ESF is captured by the ANN and then it can be use d to compensate the temperature effect in the guided wave data at a different temperat ure. When the ESF is compensated only the Damage Sensitive Feature (DSF) information is present in the experimental data acquired. Several tests were performed in a range of temperatures under damaged/undamaged conditions and used the experimental data to build and test the models. This method improves the benefits of the OBS(without the need of a big database of baselines, difficult to obtain in complex structures)with the wide range of applicability and simplicity of BSS. Another advantage of this method is its independency from structure arrangement and the type of sensors used for guided waves data acquisition because it is purely data driven. Moreover, it can be used for the simultaneous compensation of a variety of measurable environmental or operation conditions, which affects the guided wavedata acquisition, in example, temperatura and load compensation.