Creation of Attribute Vectors from Spectra and Time-series Data for Prediction Model Development

Product manufacture results in high and complex amounts of data. Most prominent and information rich are spectral and process time-series data. These data are available for every process step and raw material used for product manufacture. Due to their complexity, the use is typically very limited to research and problem solving. By applying dimensionality reduction to spectral data and specific algorithms to time-series data, attribute vectors could be created. Newly created attributes, carrying process and product details can then be used for product quality prediction. Such prediction models could replace currently used time-consuming practices.