Comparison of artificial neural network and multilinear regression analysis models in estimation of pulp flow speed from low coherence Doppler flowmetry measurement data

Low Coherence Doppler Flowmetry (LCDF) measurement produces a signal, which frequency domain characteristics are in connection to the speed of the flow. In this study performances of Artificial Neural Network (ANN) and Multilinear Regression (MLR) methods in prediction of pulp flow speed from the LCDF measurement data were compared. In the study the pulp flow speed was estimated distinctly from consecutive frequency bands of the LCDF data with both methods. The smallest estimation error in flow speed with the ANN method was 20% and with the MLR method 30%, depending on the selected frequency band. The results indicate the relationship between characteristics of the LCDF measurement and pulp flow speed includes remarkable number of nonlinear components. The result is in line with theoretical calculations about the Doppler shifts occurrence in the LCDF data.© (2007) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.