Particle swarm optimization as alternative tool to sensory evaluation to produce high-quality low-sodium fish sauce via electrodialysis

Electrodialysis (ED) has been proposed as a means to produce low-sodium fish sauce to serve the need of health-conscious consumers. Artificial neural networks in combination with genetic algorithm and experimental sensory results was previously used to optimize the ED process to produce high-quality low-sodium fish sauce. However, a large number of training data and exhaustive computational resource were required in such a case. In this study, particle swarm optimization (PSO) is proposed to optimize the ED process. Changes in the total nitrogen, total amino nitrogen and total aroma compounds concentrations were generated from our previously developed phenomenological model and used for the optimization; no experimental sensory data were needed. PSO indicated that fish sauce should contain 14.4% (w/w) salt if all the quality indicators were to be optimized; this result agreed well with the result of an independent sensory test.