Predicting Pigment Color Degradation with Time Series Models

Abstract

The colors of pigments and dyes are affected by light exposure. Light-induced color change has an impact on various industrial and artistic applications where colored materials are frequently exposed to light throughout their life-cycle. For this reason, it is beneficial to understand the fading behaviour of pigments and simulate future degradation. In this article, we are proposing a method to forecast color change of pigments based on time series analysis. To begin with, we collect fading data from real objects with a microfadeometer, which records the color coordinates after every second of light exposure. Then, we treat this data as a time series, test for its stationarity and fit it with autoregressive integrated moving average (ARIMA) models. Finally, using a train-test split, we validate the accuracy of the ARIMA models in predicting color degradation of pigments and dyes.