Neuro-Fuzzy Identification of Drilling Control System Adapted to Rock Types

The research is aimed at investigating methods of forming a model for the system of adaptive control for drilling with a control object identifier. Under rapidly changing conditions of borehole drilling it is expedient to apply a strategy of the two-level adaptive control, which implies simultaneous drilling investigation and control. The subsystem of prediction is implemented on the basis of an adaptive neuro-fuzzy system. The applied neuro-fuzzy system realizes the Sugeno fuzzy inference in the form of a five-layer neural network of signal feedforward, the first layer of which contains the terms of input variables (the current signal value and its delayed values). It should be noted that the membership function type did not influence much the prediction result. While processing and analyzing the current information on the latest characteristics of drilling and while forming the adaptive control it is reasonable to apply neuro-fuzzy structures with the Gaussian membership functions.