Design of Grinding Operations on the Basis of Modeling with Application of Artificial Neural Network

Abstract

The increase in production efficiency is largely determined by the design stage of products and technological processes using modern computer-aided design tools. The absence of universal mathematical models describing grinding necessitates the search for new approaches and methods for describing and formalizing this process. The complexity of the solution of this problem is due primarily to the multifactority of the process and the probabilistic nature of the phenomena taking place. Creation of a universal mathematical model of the grinding process, which allows to determine the parameters of the tool and the processing mode, is possible on the basis of application of neural network programming. The paper proposes to use an artificial neural network (ANN) in order to solve the problem under uncertainty, the main difficulty of work with which is connected with the construction and training of the network. To model the grinding process, an ANN was constructed, the hyperbolic tangent being used as the activation function. The conducted studies showed the possibility of building a network on 6 neurons in a hidden layer. Evaluation of the performance of a neural network, the structure of which consists of three layers: an Input one - 7 neuron-receptors; Hidden - 6 neurons; Output - 6 neurons, showed that the error of the trained ANN lies in the range from 0.1% to 2%. The construction of the model includes the following steps: loading into the system of training data arrays; calculation of the parameters of the future model; analysis of initial data and output of information about the quality of training arrays; creation of structure and training of neural network. The developed model is used as the basis for the automated design of the regime-tool equipment for flat grinding operations, consisting of several modules: Modeling, Model Development, Model Archive, Analyzer, Reference, User Interface. The use of specialized CAD of grinding operations with the possibilities of automating the choice of abrasive tools and process parameters, created on the basis of a mathematical model using ANN, contributes to improving the quality and reducing design time.