| Abstract: | Traditionally, improvement of the constrained alternating least squares (ALS) solution has been executed by the addition of a priori information in the initial estimates and or constraints. However, there are cases where this information simply does not exist or is impossible to acquire under the process conditions. Therefore, new strategies are required to produce starting estimates close to the actual solution without the need of a priori information. Quantitative iterative target transformation factor analysis (QITTFA) is developed as a solution to this problem. The QITTFA approach combines the strengths of both iterative target transformation factor analysis (ITTFA) and simple‐to‐use interactive self‐modelling mixture analysis (SIMPLISMA) to (1) produce a solution space spanned by the independent factors such that the variance contribution due to noise is reduced, (2) to iteratively refine the solutions space prior to ALS and (3) to select the most pure variables from the refined solution space using the purity criterion. It has been observed that the QITTFA approach markedly improves the conventional SIMPLISMA and second derivative SIMPLISMA performance in the presence and absence of selectivity. In addition, components of differing spectral characteristics (narrow or broad spectral features) can be resolved, without a priori knowledge of the shapes of the pure components. This has been demonstrated with a simulated high performance liquid chromatography‐diode array detection (HPLC‐DAD) dataset, a laboratory‐based UV–Vis calibration dataset and a gaseous near infrared (NIR) dataset from an industrial process. Copyright © 2007 John Wiley & Sons, Ltd. |
