Predicting the ultraviolet radiation protection by polyester-cotton blended woven fabrics using nonlinear regression and artificial neural network models.

INTRODUCTION: Ultraviolet protection factor (UPF) of fabric is mainly influenced by fabric cover which is dependent on the primary fabric construction parameters like yarn count and thread density. UPF can be modeled by using these primary fabric parameters as inputs by the help of nonlinear regression as well as artificial neural network (ANN).
OBJECTIVES: The objective of this study is to develop prediction models for fabric UPF using nonlinear regression and ANN techniques and compare their relative efficacy.
METHODS: Thirty-six woven fabric samples were produced by varying weft related parameters like proportion of polyester, weft count and pick density. Nonlinear regression and ANN models were developed from same experimental data sets. Prediction accuracy of both the models was evaluated and compared. Trend analysis was performed to check the generalization ability of the ANN model.
RESULTS: UPF was well estimated from the three primary fabric parameters by both the nonlinear regression and ANN models. However, ANN model demonstrated better prediction accuracy than the nonlinear regression model.
CONCLUSIONS: Fabric UPF can be predicted with high degree of accuracy using ANN and nonlinear regression models. These models can be used to estimate the UPF of woven fabrics without spectrophotometer based test.