Investigation of the Possibility of Reconstructing Strain Diagrams Using the Neural Network Approach

摘要

The results of neural network computational-experimental estimations of the strength properties of steels based on the diagrams of surface local elastoplastic deformation are given. The presented neural network approach to conversion of the indentation diagram into the diagram of uniaxial tension makes it possible to unify the specimen-free technique of determining the main mechanical properties of the materials of equipment. Neural network models for constructing point and interval estimates for different levels of elastoplastic deformation are described.