Prediction and Optimization of Heavy Clay Products Quality

Abstract

Th e eff ects of chemical composition, fi ring temperature (800-1100 °C), and several shape formats of laboratory brick samples on the fi nal product quality were investigated. Prediction of the fi nal laboratory products parameters was evaluated by second order polynomial regression models (SOPs) and artifi cial neural networks (ANNs), and aft erwards compared to experimental results. SOPs showed high r 2 values (0.897-0.913 for compressive strength models, 0.942-0.962 for water absorption, 0.928 for fi ring shrinkage, 0.988-0.991 for water loss during fi ring, and 0.941 for volume mass of cubes models). An ANN model, coupled with sensitivity analysis, was obtained with high prediction accuracy: 0.866-0.939 for compressive strength models, 0.954-0.974 for water absorption, 0.882 for fi ring shrinkage, 0.982-0.988 for water loss during fi ring, and 0.920 for volume mass of cubes models. Th e optimal samples' chemical composition and fi ring temperature were chosen depending on a fi nal usage of the raw material in heavy clay brick industry.