A damage detection procedure using two major signal processing techniques with the artificial neural network on a scaled jacket offshore platform

With the help of Structural Health Monitoring (SHM) methods, it is possible to identify the occurrence of damage at its early stages and prevent fatality and financial damages. Great advances in signal processing methods in combination with Machine learning tools have led to better achieve this goal. In the present paper, the two major techniques, that is, Empirical Mode Decomposition (EMD) and Discrete Wavelet Transform (DWT) are combined with Artificial Neural Network (ANN) through processing raw acceleration responses measured on a scaled jacket type offshore platform which was constructed and tested as a benchmark structure at K.N. Toosi University of Technology. In this way, ANN was trained by the signals obtained from EMD and DWT for three different conditions of the jacket platform to determine the relative damage severity. The envelope of the obtained signal's energy (ENV) as an appropriate damage index was used to determine the damage location. The results of the application of this procedure on the case study indicated that DWT, compared to EMD, is a more reliable signal processing method in damage detection due to better noise reduction.

Automated summary

With the help of Structural Health Monitoring methods, damage can be identified early on to prevent fatalities and financial losses. The combination of signal processing techniques and machine learning tools has led to significant progress in achieving this goal. Experimental results show that Discrete Wavelet Transform is a more reliable method for damage detection compared to Empirical Mode Decomposition due to superior noise reduction capabilities.