Nonlinear rock-physics inversion using artificial neural network optimized by imperialist competitive algorithm

Estimation of petrophysical properties from seismic attributes can be considered as rock-physics inversion problem. In general, rock-physics models are nonlinear and require nonlinear optimization algorithms to solve the inversion problem. Typically, the conventional method of inversion employs the linearized approximation of the forward model and utilizes the linear inversion methods which are usually not accurate enough and prone to be trapped in a local minimum. This paper presents a novel method of nonlinear rock-physics inversion based on artificial neural network optimized by imperialist competitive algorithm. We used Kuster and Toksöz inclusion model with spherical geometric factor as forward model to map the model parameters to the observed data. To quantify the performance of the method, we compare it with the Bayesian linearized rock-physics method. The result shows that the presented method can achieve more reliable and accurate inversion of the petrophysical parameters.