Method of forming a database for learning expert system to evaluate stability of the welder

Abstract

The technique of building a database of training examples of training artificial neural network underlying the expert system. It is assumed that the quality of the welded joint depends on the stability of the welding process. Stability manual arc and mechanized welding depends on the skill of the welder. To create a database using the field of quality of the welded joint, built by the method proposed by Professor. G. Lyudmirsky. A method for constructing the area as including welding samples according to the scheme allows for smooth transitions from high-quality weld portion to be defective, that allows to define the quality of the welded joint. As a result, the quality of the resulting area is used to synthesize a database of training examples, which are pairs of input-output artificial neural network. A technique for training artificial neural network, which allows a fast mode to train, without losing the quality of education. It is shown that the proposed method of forming a database for learning expert system to reduce the time and number of experiments without loss of data reliability and the quality of learning expert system.