There's a LinkedIn discussion going an at present on population balance modeling in crystallization. I posted there today as follows:
Building a population balance model / framework is quick and easy, especially if you start from a readymade template. Making the model fit your data is more time-consuming and the quality of fit (and confidence ellipsoids) may be reasonable, though the relevant 'kernels' may be too simple or too averaged to fit really well. Making accurate predictions with your model, especially with a change of scale and equipment configuration, adds a further layer of effort to be successful, as the balance among the important phenomena may shift away from your original conditions. In some industries, this level of effort may be indulged; in the fast moving pharmaceutical development project, it usually will not, depending on the phase of development and the questions that need to be answered.
Common sense tallies with our experience in this area, that users need to understand / model first the mass balance, then the energy balance and then possibly the number / population balance. In many projects, trying to start at the wrong end (the population balance) only reveals that the mass and energy balance are not understood. On the other hand, starting with mass (concentration, addition rate, solubility) and energy (temperature, cooling / evaporation rate) alongside a basic evaluation of equipment characteristics (agitation, solids suspension, heat transfer resistance) often leads to insight that solves problems without requiring a population balance approach.
In these more routine projects, on-line size data (such as FBRM) are useful as a diagnostic and to provide trend information.
Post a Comment