Multiple Ramp-variable Retort Temperature Control for Optimal Thermal Processing

Abstract

Variable retort temperature (VRT) thermal processing has been recognized as an innovative method to improve food product quality and save process times. The key to designing a VRT thermal process is to choose a reasonable (optimal) VRT temperature profile for a given food product and package being thermally processed. The selection of optimal retort temperature profiles with a multistage ramp function involving multiple variables is complex and difficult to handle by conventional optimization methods. Newer concepts such as artificial neural networks (ANN) and genetic algorithms (GA) have the potential to deal with such complex situations. The objective of this study was to develop ANN-GA based procedures for the selection of the optimal retort temperature profile under multistage ramp-variable (MRV) retort temperature control for optimizing thermal processing. ANN concepts were used for developing dynamic prediction models for process time, average quality retention and surface cook value, in which the input variables were ramp time (30-70min) and step temperatures (104-134 °C) in four consecutive stages of retort processing. GA-ANN based optimization procedure was then developed using a commercial GA and ANN software, and used for searching the best combination of retort ramp/soak sequence in each of the four stages to give a continuous variable temperature profile that achieves the optimization objective under imposed constraints conditions. The statistical results of modelling performance for all ANN models were: average correlation coefficient r2 = 0.95 and relative error Er = 2.05%. The optimization results and processing efficiency using GA were affected by main GA configuration parameters including initial population number, mutation rate and crossover rate, and the optimal configuration parameters of GA for this study were determined by trials. The optimal retort temperature profiles meeting different optimization objectives for different can sizes were obtained using the GA-ANN optimization method. Compared with the constant retort temperature, the MRV process could improve both process time (up to 43%) and surface quality (up to 24%) significantly. The results suggested that the hybrid artificial intelligence techniques of neural networks and genetic algorithms can be efficiently used for modelling and optimization of the MRV retort temperature control for thermal processing.