Prediction of plywood bonding quality using an artificial neural network

Abstract

Abstract The bonding quality test is one of the most important of all tests performed on plywood, because it determines the suitability of boards for use in the type of exposure they are intended for. Because this test involves aging pretreatment, results are not available in <24–97 h after manufacture, depending on the type of board, and therefore any error in the manufacturing process is not detected until 1–4 days later. To solve this time problem, an artificial neural network was developed as a predictive method to determine the suitability of board bonding through other properties that can be determined in less testing time: thickness, moisture content, density, bending strength, and modulus of elasticity. The network designed WAS a feedforward multilayer perceptron trained by supervised learning after normalization of the input data, and allowed the bonding test result to be predicted with 93% accuracy.