Mechanistic models enable profound knowledge management
Mechanistic models base on the laws of mathematics, physics and physical chemistry and inhere all elementary information of the system dynamics. The model parameters have an actual physical meaning, which provides fundamental benefits for the scientific interpretation and understanding. The model equations thereby present a clear distinct language, optimal for sharing growing process knowledge with a team.
Typically, a first draft process is scratched in the early development phase. This process is then refined throughout the development life cycle. Using a model based approach as implemented in DSPX, it is easy to manage and share this growing process knowledge within a team. Same as the actual process, also your model evolves in time. Starting with a draft model in early phase, the model evolves in time, by adding more and more available information to it. The improvement is thereby clearly documented in terms of the model equations and parameters, both saved in the form of a DSPX project file. So instead of sharing complex process specifications, lab- and pilot-scale columns within your team, you simply hand over your DSPX project file. The DSPX datastructure is centered around components: columns, resins, molecules and even methods are treated as individual items. Such items can be stored easily in a project-independent database. When starting a new project, you can then simply refer to previously used components in the database.
Integrated design of experiments (DoE) is a common approach to reduce the experimental load, while increasing the amount of information. The key idea is to plan an experimental series, such that the most information is gained from a minimum number of experiments.
However, a real experiment takes up to two days and costs on average EUR 20,000. Downstream process simulation with DSPX is a high-potential innovation and a gamechanger. It allows to replace these lab experiments by computer simulations, that achieve the same results within seconds and at the price of computing power only.
In comparison to the traditional one-factor-at-a-time approach, the level of process knowledge increases quickly already with few experiments. After several hundreds of simulated experiments, the maximum level is reached and no more additional information is generated by adding more experiments.
At current time, the required level for regulatory acceptance is however reached with a few hundred experiments – a lot faster and cheaper than in the fully experimental approach using one-factor-at-a-time.