Neural Network: A Potential Approach for Error Reduction in Color Values of Polycarbonate

Abstract

ABSTRACTIn current exploration, the artificial neural network (ANN) is executed to reduce the errors in color values of polycarbonate. The network consists of sigmoid hidden units and a linear output unit arranged in a feed forward backpropagation architecture. An optimal design is accomplished for 10, 12, 14, 16, 18, and 20 hidden neurons on a hidden layer with five different algorithms involving batch gradient descent, batch variable learning rate, resilient back propagation, scaled conjugate gradient, and Levenberg–Marquardt. The training data for ANN are obtained from experimental measurements. There were 22 inputs and three tristimulus color values L*, a*, and b* were used as an output layer. Statistical analysis in terms of root‐mean‐squared, an absolute fraction of variance (R2), as well as a mean square error is used to investigate the performance of ANN. The best result in terms of statistics is presented by the LM algorithm with 14 neurons in the designed ANN model. The degree of accuracy of the ANN model in reduction of errors is proven acceptable in all statistical analysis and shown in results. However, it was concluded that ANN provides a possible method in error reduction in specific color tristimulus values. © 2013 Wiley Periodicals, Inc. Adv Polym Technol 2014, 33, 21402; View this article online at wileyonlinelibrary.com. DOI 10.1002/adv.21402