(706d) Optimal Controller Design Based on Efficient Ant Colony Optimization Algorithm. Case Study: Chemical Process Control

In this work, an advanced optimal control solution method inspired by biological systems is proposed. Biological systems have shown success of solving difficult problems encountered in nature and hence, serve as motivation for the formulation of optimization and advanced chemical process control approaches. For example, the behavior of the natural groups such as ants, bees and swarms demonstrates that self-organization and cooperation by following simple rules of interaction can result in a wide range of optimal patterns [1]. Inspired by antâ??s foraging behavior, this work proposes a biomimetic optimal control strategy for chemical process control specifically for non-linear fermentation processes. The proposed strategy employs the computationally efficient optimal control solver entitled efficient ant colony optimal control (EACOC) algorithm. The EACOC algorithm is a metaheuristic optimization technique inspired by the antsâ?? foraging behavior which utilizes probabilistic and stochastic concepts for solving large-scale optimization problems. A brief description of this algorithm is presented below.

In basic ant colony optimization (ACO) algorithm, artificial ants are stochastic candidate solution construction procedures that exploit a pheromone model and possibly available heuristic information of the mathematical model. The artificial pheromone trails are used as means of communication among the artificial ants. Pheromone decay allows the artificial ants to forget the past history and focus on new promising search directions. The pheromone values are updated according to the information learned in the previous iterations and hence, the algorithmic procedure leads to accurate and potentially global optimal solution. EACO algorithm improves the performance of the conventional ACO algorithm for combinatorial, continuous and mixed-variable optimization problems by introducing Hamersley sequence sampling (HSS) for probabilistic elements of ACO. The initial solution archive diversity for continuous and mixed-variable optimization problems plays an important role in the performance of ACO algorithm. The uniformity property of the HSS technique is exploited to avoid clustering of the initial solution archive in a small region of the potential solution space. Moreover, ACO algorithm is a probabilistic method where several randomized probability functions are involved in the algorithm procedure. The distribution of these random numbers affects the performance of the ACO algorithm. At this step, the multidimensional uniformity property of HSS is also introduced to improve the computational efficiency of the ACO algorithm. The capabilities of the proposed methodology are illustrated using benchmark problems and real world case studies of computer aided molecular design problems. Specifically, these problems include the optimal design of solvent for extraction of acetic acid from waste process stream using liquid-liquid extraction [3] and adsorbent selection for naturally occurring radioactive materials of natural gas fracking produced water [4].

The EACOC algorithm is developed based on the efficient ant colony algorithm and orthogonal collocation methods. The proposed algorithm is implemented for optimal control of dynamic fermentation processes where the objective function is to track the product concentration set-point. The challenges in this fermentation process are the steady-state multiplicity and the oscillations in the concentration profile due to nonlinearities present in the process model [2, 5, 6, 7]. This approach successfully overcomes these challenges. The implementation results using EACO show potential improvement in the solution quality and reduction in computational time when compared to a gradient-based solver. Thus, the proposed algorithm can be used as an alternative to the gradient-based optimal control algorithms for optimization of large-scale dynamic problems.

References:

1. Hristu-Varsakelis D. and Shao C. â??A bio-inspired pursuit strategy for optimal control with partially constraints final stateâ?. Automatica 2007; 43:1265-1273.
2. Sridhar L. â??Elimination of oscillations in Fermentation Processâ?. AIChE Journal 2011;57(9):2397-2405.
3. Gebreslassie BH, and Diwekar UM. Efficient ant colony optimization (EACO) for computer aided molecular design: Case study solvent selection problem. Computers and Chemical Engineering 2015: 78: 1-9.
4. Benavides PT, Gebreslassie BH, and Diwekar UM. Optimal design of adsorbents for NORM removal from produced water in natural gas fracking. Part 2: CAMD for adsorption of radium and barium. Chemical Engineering Science 2015: 37: 977-985.
5. Lima F. V., Li S., Mirlekar G. V., Sridhar L. N. and Ruiz-Mercado G. J., â??Modeling and advanced control for sustainable process systemsâ?. Sustainability in the Analysis, Synthesis and Design of Chemical Engineering Processes, G. Ruiz-Mercado and H. Cabezas (eds.), Elsevier, 2016.
6. Mirlekar G. V., Gebreslassie B. H., Diwekar U. M. and Lima F. V., "Design and implementation of a biomimetic control strategy for chemical processes based on efficient ant colony optimization." Presented at AIChE Annual Meeting, Salt Lake City, Utah, November 2015.
7. Li S., Mirlekar G. V., Ruiz-Mercado G. J., and Lima F. V., â??Development of Chemical Process Design and Control for Sustainabilityâ?. Submitted for publication, 2016.