In this paper, we present the application of the particle-based simulation to complicated fluid flow problem with free surfaces. The particle approach is based on the SPH (Smoothed Particle Hydrodynamics) method using hyperbolic-type kernel function. The hyperbolic-type kernel function is constructed by differentiating the characteristic function (i.e., sigmoid function) based on neural network framework. The function which has the property of Dirac delta function is concomitantly applied to the smoothing kernel function in the SPH-framework. Numerical results demonstrate the workability and validity of the present approach through the dam-breaking flow problem.