Study of the low cycle fatigue of 12Cr18Ni10Ti steel based on fractal analysis and artificial intelligence approaches

Abstract

The evolution of the structure and assessment of the age limit of steel 12Cr18Ni10Ti upon fatigue loading is considered using neural network modeling and approaches of fractal analysis of the microstructure. An algorithm for processing images of the microstructures has been developed to improve their quality. An indicator of the fractal dimension of the image is used as a quantitative indicator for assessing the evolution of the microstructure of the surface metal layer. A quantitative assessment of the structures at different stress amplitudes is carried out in a wide range of low temperatures using the fractal dimension index. Correlation of the fractal dimension index with the run of the sample material is shown. The appearance of the main crack was observed in the range of 0.7 - 0.8 from the number of cycles to failure, after which the crack growth rate increased. At a lower temperature, the main crack is formed later, but further loading results in a higher crack growth rate. Formation of the secondary phases in austenitic steel at a lower temperature occurred at earlier stages than that at a temperature of t = +20°C, which led to hardening of the material. An artificial neural network (ANN) has been developed and trained for assessing structural changes in metal proceeding from the fractal dimensionality of the microstructure images at different stages of fatigue loading. The developed neural network made it possible to estimate with a sufficiently high accuracy the number of cycles before damage of the sample and the residual life of the material. Thus, the developed ANN can be used to assess the current state of the material in a wide range of low temperatures.