Abstract
Multimodal optimization problems are commonly found in engineering problems, and their solution can be very challenging for metaheuristic approaches. In this work, the use of a recently proposed multimodal metaheuristic method was analyzed - the Multimodal Flower Pollination Algorithm - in two fluid phase equilibrium problems: (i) the calculation of double azeotropes and (ii) parameter estimation in a thermodynamic model. Two different formulations were also considered in the double azeotropy problem. In the azeotrope calculation, a statistical analysis was conducted in order to verify if the algorithm performance is affected by the the problem formulation. The computational results indicate that the methodology provides robust results and that the objective function employed affects the computational performance.
Highlights
Multimodal optimization problems can be a challenging test for stochastic optimization algorithms (Platt, 2016), considering the task of locating all minimum/maximum points of the problem at hand
We will present a brief description of the Multimodal Flower Pollination Algorithm (MFPA), proposed by Gálvez et al (2017)
The results indicated a statistical difference between the two formulations employed in the double azeotropy problem, favoring the strategy proposed by Bonilla-Petriciolet et al (2009)
Summary
Multimodal optimization problems can be a challenging test for stochastic optimization algorithms (Platt, 2016), considering the task of locating all minimum/maximum points of the problem at hand This kind of problem has been studied with crowding, sharing, niching and speciation techniques, among others (Thomsen, 2004; Parrot and Li, 2006; Cuevas and Reyna-Orta, 2014). Standing out among them — in terms of simplicity of implementation — are the Multimodal Cuckoo Search (MCS) (Cuevas and Reyna-Orta, 2014) and the Multimodal Flower Pollination Algorithm (MFPA) (Gálvez, Cuevas, and Avalos, 2017) These algorithms were tested in typical benchmark functions, but lack tests and validation in real engineering problems
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have