Intelligent prediction of rock bolt debonding employing the fractal theory and relevance vector machine (FT-RVM) with piezoceramic transducers

Abstract

A new intelligent prediction model incorporated fractal theory and relevance vector machine (FT-RVM) was proposed to detect the debonding status of the rock bolt by using the piezoceramic transducer-induced stress waves. In the FT-RVM model, the original signals under different debonding status are firstly decomposed by the wavelet packet decomposition, the box dimension of decomposed signal is extracted by fractal theory (FT). The fractal box dimension of decomposed signals and root mean square value of the original signal are used as the as the inputs to the FT-RVM model, and the different debonding status of the glass fiber reinforced polymer rock bolt is the output. After the training, the prediction model is used to estimate the debonding status of the rock bolt. In the FT-RVM model, the kernel function utilized is the Gaussian radial basis function, and its optimal value is obtained by using particle swarm optimization (PSO). The experimental results show that the average relative error of the FT-RVM prediction model is 3.04%, and the accuracy and reliability of the model are high, which demonstrates the intelligent identification of GFRP rock bolt debonding status. The proposed intelligent prediction model based on FT-RVM could be used to quickly evaluate rock bolt debonding status.