Development of a new method of crack modeling and prediction algorithm

Abstract

In this report, the well known parametric method of signals and systems representation is extended to modeling and prediction of cracks on building and road surfaces. Also, a new algorithm based on complex value autoregressive neural network with split linear activation function for the determination of Complex-Value Autoregressive Moving average (CARMA) coefficients is also proposed in this report. Furthermore, mathematical derivation and detail analysis of the proposed CARMA based Complex-Value Neural Network (CVNN) algorithm is also discussed in this work.