Classification of materials by modal analysis and neural network

Abstract

Modal analysis is the study of dynamic characteristic of structures induced by vibrational excitation. Under modal excitation, three important parameters namely natural frequency, damping ratio and mode shape associated with the structural properties are acquired. These modal parameters are used as the extracted features for classification on artificial neural network. This paper presents an experimental investigation of two different kinds of materials by implementation of modal analysis along with the integration of neural network for materials classification. The experimental modal analysis is done using the LMS instruments and software where Fast Fourier Transform (FFT) and Frequency Response Function (FRF) are used to extract the mentioned modal parameters. The extracted parameters are used as the classification process feature of the neural network. Multi-layer Perceptron (MLP) is used as the mapping model of the network. The technique adopted for the system is the Levenberg-Marquadt (LM) and Scaled Conjugate Gradient (SCG) Backpropagation technique.