Neural Network Recognition of Spherical Bodies Set Grain-Size Distribution Using Envelope of Surface

Presented in this paper is a neural network method used for recognition of spherical bodies set grain-size distribution using a data taken from envelope surface of a given set. Development was carried out in accordance with principles designed for neurocomputer-based measurement devices. These principles concern measurement devices featured by complicated nonlinear dependency between physical parameter being measured and indirect measurement values. In simple applications this dependency is substituted by scale calibration function for this measurement device. Results presented in this article were obtained using 3D-model set. Diameters of bodies, included in this set, lit the range used as standard at ore mining and processing enterprises. Described herein are the neural network algorithm, as well as a method used for informative training and test sets composition, required for effective training and verification of neural network system. Moreover, we present a comparison between neural network approach and known linear methods.