Prediction of the thickness of a thin paint film by applying a modified partial-least-squares-1 method to data obtained in terahertz reflectometry

摘要

We applied a multivariate analysis method to time-domain (TD) data obtained in terahertz (THz) reflectometry for predicting the thickness of a single-layered paint film deposited on a metal substrate. For prediction purposes, we built a calibration model from TD-THz waveforms obtained from films of different thicknesses but the same kind. Because each TD-THz wave is approximate by the superposition of two echo pulses (one is reflected from the air-film boundary and the other from the film-substrate boundary), a difference in thickness gives a relative peak shift in time in the two echo pulses. Then, we predicted unknown thicknesses of the paint films by using the calibration model. Although any multivariate analysis method can be used, we proposed employing a modified partial-least-squares-1 (PLS1) method because it gives a superior calibration model in principle. The prediction procedure worked well for a moderately thin film (typically, several to several tens of micrometers) rather than a thicker one.