Data Fusion Modeling of Lumber Moisture Content Sensors Using Chebyshev Functional Link Networks

Abstract

Lumber moisture content sensors operating in harsh environment are easily influenced by ambient factor parameters. Data fusion technique is proposed to combine data from several sources into a single unified description. A novel single layer functional link network (FLN) using Chebyshev polynomials is used for this purpose to compensate for the nonlinear response characteristics and complex nonlinear dependency of the environmental parameters on the sensor characteristics. FLN eliminates the hidden layers of conventional neural networks by expanding the input pattern into a high order dimensional space. Compared to the multilayer perceptron (MLP), Chebyshev FLN has the similar performance and less computational complexity.