A decision and control technique based on fuzzy control, neural networks and genetic algorithms for optimization of a fruit-storage process

A skilled grower can deal well with such complex systems with his own intuition and experience. In this study, a decision and control technique mimicking a skilled grower´s decision process was proposed and then applied to the optimization of a fruit-storage process. It includes a fuzzy controller and two decision systems I and II consisting of neural networks and genetic algorithms to imitate the expert´s manners. A skilled grower´s decision process can be simplified into three steps: the construction of a mental model describing the input-output relation of an unknown system learning (“Learning”); the selection of the best strategy through simulation of the identified mental model (“Selection”); and the action of the best strategy in a highly skilled manner (“Action”). In the decision system I, fruit responses as affected by the relative humidity are first identified using neural networks (“learning”) and then the optimal setpoints of the relative humidity are searched for through simulation of the identified model using genetic algorithms (“selection”). Furthermore, a fuzzy controller tuned optimally adjusts the relative humidity (“action”). The optimal setpoints of the relative humidity resulted in better qualities of fruits during storage. The fuzzy controller gave a better control performance of the relative humidity. It was found that a decision and control technique is useful for realizing the optimization of production systems