| Abstract: | A mathematical framework was described and discussed that relates the triacylglycerol (TAG) distribution to measurements results. This model is valid for any analytical technique for which an unambiguous relation between the TAGs and the experimental data exists. The framework can be employed to estimate the TAG distribution based on any subset of these analytical techniques. Furthermore future techniques can be incorporated, eventually enabling the computation of the exact TAG distribution. In current practice the relation from measurement values to the TAG distribution is believed to be described by Coleman's theory. However some natural fats and fractions of fats are known to be non‐adherent to Coleman's theory. For both palm stearin and palm kernel oil this notion is tested and confirmed. For these fats a different approach is needed to be able to estimate the TAG distribution. Within the framework three different approaches for making an estimation of the TAG distribution based on measured data are described, discussed and compared with Coleman's results: a least squares approach, a genetic algorithm (GA) approach and an expectation maximisation (EM) approach. The results presented demonstrate that the estimated TAG distributions from the GA and EM algorithm are significantly better than Coleman's estimate and the least squares approach. Overall the EM algorithm is favoured because of its consistency. |
