Analysis and Prediction of the Physico-Mechanical Properties of Reactor Steel by Means of Artificial Intelligence and Applied Statistics

In 19551961, a fuel element with cladding made from ferrite-martensite steel for use in lead-bismuth coolant was developed at the Leipunskii Institute for Physics and Power Engineering (FEI) under the direction of V. A. Malykh. The solution of an important federal problem, viz., the development of fast nuclear submarines, also became possible after the original development of the fuel element [1]. Ferrite-martensite class steel with 12% chromium content, which is used as a structural material in fast reactors, continued to be improved in subsequent years. One direction of the new technological platform, which has now been adopted in the State Corporation Rosatom, is the development of innovative fast reactors with intense temperature and dose loads, using lead, lead-bismuth and sodium as coolant. In this connection, the development of ferrite-martensite steel, possessing enhanced strength and plasticity, has not lost its urgency. The determination of the optimal chemical composition and heat-treatment conditions is complicated by the multifactor nature and nonlinearity of the problem. Research in this field can be intensified on the basis of modern methods of multidimensional statistical analysis of experimental data and neural net modeling [2, 3]. The development and application of methods oriented toward massive calculations are largely due to the increase in the computational power of modern computers.