Predicting hydrophobicity of silica sol-gel coated dyed cotton fabric by artificial neural network and regression

Abstract

Artificial neural network (ANN) and multiple linear regression (MLR) have been used to predict the hydrophobicity of silica sol-gel coated dyed cotton fabric using different nanoparticle concentrations, dye concentrations, dye types and cross linker types as predictors. A total of 32 samples have been dyed with reactive and direct dyes using two dye concentrations at HT dyeing machine. To develop nano roughness on dyed fabric, with an aim to create super hydrophobic dyed cotton, different concentrations of silica nanoparticles with a combination of silane hydrophobes (alkyltrialkoxysilanes), and silane cross-linkers, i.e. tetraethoxysilane (TEOS) and teramethoxysilane (TMOS) are applied by sol-gel technique using dip-dry-cure process. The hydrophobicity is measured by AATCC spray rating technique. The coefficient of determination (R²) indicates that there is a strong correlation between the measured and the predicted values with a trivial mean absolute error; ANN is found to be more powerful predicting method than MLR. The most influencing variables revealed through correlation coefficient and P-values of regression model are silica nanoparticle and dye concentration. Empirical and statistical models have been proposed to predict dyed cotton fabric hydrophobicity without any prior trials, which reduces cost and time.