Neural Network Photothermal Depth Profiling of a Heat Source Distribution Application to Water Migration in Starch Sheets

Abstract

The determination of the water migration profile in starch is used as an experimental model system for the depth profiling of heat source distributions. A depth dependent water concentration, which has diffused into the sample during a timed exposure to a humid atm osphere, acts as a specific position dependent heat source. The direct and inverse relation between the water content and the frequency spectrum of a photothermal signal detected by a pyroelectric tranducer in contact with the sample is investigated. The inverse problem is treated by a neural network and a table lookup method. The profile reconstruction is illustrated for simulated and experimental cases.