MOBioPro - Nature-based computation in the modelling and optimization of
Partners: CCTC, CEB/IBB (U. Minho)
A number of valuable products such as recombinant proteins, antibiotics and amino-acids are exclusively produced using fermentation techniques. Additionally, for a significant number of other chemicals there is an increasing trend in the cost of the products manufactured by chemical processes, mainly due to the increasing burden of environmental compliance for those processes. Thus, there is a tremendous economic incentive to decrease the manufacturing costs of biotechnologically produced products.
However, those processes are typically very complex, involving different transport phenomena, microbial components and biochemical reactions. Furthermore, the nonlinear behavior and time-varying properties make bioreactors difficult to control with traditional techniques. For these reasons, the application of model-based control and optimization strategies to biotechnological processes, as opposed to most of the remaining engineering fields is still not widely implemented.
For that purpose there is the need to consider reliable quantitative mathematical models, capable of describing the process dynamics and the interrelation among relevant variables. Additionally, robust optimization techniques must deal with the model's complexity, the environment constraints and the inherent noise of the experimental process.
The optimization of these processes is traditionally handled by analytical and numerical methods, whose results degrade when the complexity of the problem increases. A different approach comes from the use of general purpose optimization algorithms taken from the field of Evolutionary Computation (EC). Evolutionary Algorithms (EAs) are suitable to the optimization of a number of the fermentation process parameters, namely the initial values of relevant state variables, the trajectory of controlled variables over time and even the duration of the process.
On the other hand, the modeling of fermentation processes typically makes use of white box mathematical models, based on differential equations that represent the mass balances of the relevant state variables. More complex approaches have been proposed that take into account the nonlinear and dynamic nature of the process. In this arena, Neural Networks (NNs) and Fuzzy Logic have been a focus of attention.
This project aims at studying the application of models that combine the use of Neural Networks in the modeling of the process dynamics and Evolutionary Algorithms used as optimization tools.
- Miguel Rocha, CCTC, PI
- Isabel Rocha, CEB/IBB
- Rui Mendes, CCTC
- Paulo Cortez
- Grant students (current and past): Jose P. Pinto, Paulo Maia, Rafael Carreira
Starting date: 1/7/2005
Duration: 36 months
Budget: 73000 euros
For further details, contact the PI (mrocha at di dot uminho dot pt).