Modified Particle Swarm Optimization-Artificial Neural Network and Gene Expression Programing for Predicting High Temperature Oxidation Behavior of Ni–Cr–W-Mo Alloys

This paper is an attempt to model the oxidation behavior of Ni-base alloys by considering the alloying elements, i.e., Cr, W, Mo, as variables. Modified particle swarm optimization-artificial neural network (MPSO-ANN) and gene expression programming (GEP) techniques were employed for modeling. Data set for construction of (MPSO-ANN) and GEP models selected from ۶۶ cyclic oxidation performed in the temperature range of ۴۰۰-۱۱۵۰ ᵒC for ۲۷ different Ni-based alloy samples at various amounts of Cr, W, and Mo. The weight percent of alloying elements selected as input variables and the changes of weight during the oxidation cycle considered as output. To analyze the performance of proposed models, various statistical indices, viz. root mean squared error (RMSE) and the correlation coefficient between two data sets (R۲) were utilized. The collected data of GEP randomly divided into ۲۱ training sets and ۶ testing sets. The results confirmed that the possibility of oxidation behavior modeling using GEP by R۲ = ۰.۹۸۱, RMSE =۰.۰۸۲۲. By consideration of oxidation resistance as criteria, Cr, Mo, and W enhanced the oxidation resistance of Ni-based alloys. The results showed that in the presence of Cr as alloying element, especially at Cr contents higher than ۲۲ wt.%, the effect of W and Mo were negligible. However, the same trend was reversed at the sample with Cr content lower than ۲۰ wt.%. In these cases, the effect of W and Mo on oxidation resistance were significantly enhanced.

کلیدواژه ها:

Artificial Neural Network ، Gene expression programming ، High-temperature alloys ، Modified Particle Swarm Optimization ، Oxidation behavior