Sums of amplitudes analysis – A new non-parametric classification method for time series deviation evaluation in pharmaceutical processes

Abstract

The evaluation of various types of data and the understanding of their properties is the backbone of scientific work. Time series data are abundant in the investigation of manufacturing processes. Since they result from measurements of parameters over time, time series have distinctive statistical characteristics. Focused on bulk solid dosing, the authors illustrate the potential for misinterpretations if these characteristics are ignored in data evaluation. The sums of amplitudes analysis (SAA) is developed, based on discrete Fourier transformations as a novel non-parametric statistical method for description of time series. Subsequently, the authors use the SAA to investigate bulk solid dosing data. In this process, artifications of data of external catch scales, commonly used for the evaluation of bulk solid dosing, are found, and discussed.