Hybrid Algorithms for the Design of Sensor Networks of Nonlinear Systems

SANCHEZ MABEL; HERNÁNDEZ JOSÉ; CARNERO MERCEDES

San Francisco

Congreso; 2016 AIChE Annual Meeting; 2016

American Institute of Chemical Engineers

In this work, the metaheuristic proposed by Carnero et al. (2013) is extended to the design of SNs of processes whose operation is described using the set of nonlinearalgebraic equations associated to the mass, component and energy balances. The approach uses an Estimation of Distribution Algorithm (Hauschild and Pelikan, 2011), i.e. a global search engine, in combination with a local search heuristic. This is Strategic Oscillations within the framework of Tabu Search. Application results of the hybrid technique have demonstrated a rewarding performance for solving the same problem for linear systems of incremental size. It achieves a high reproducibility of the good solutions using a controlled computational time. The incorporation of the process knowledge contained in the balance equations is extremely useful to guide the search. In this work, a technique is presented to form the initial population of the Estimation of Distribution Algorithm in such a way that at least the observability constraints of the key variables are satisfied. The performance of the proposed design methodology is evaluated and compared with the corresponding ones to other existing strategies for small and medium size benchmarks. Regarding a large scale case study, best known solutions are provided, and their reproducibility is analyzed.