Pig chilling process analysis based on constrained EM algorithm

A chilling process simulation approach is presented to analyze a few pig chilling experimental records from a meatpacking house. Two classes of chilling process vectors (CPVs): temperature slope based cooling process vector and loss based losing process vector were defined. They were utilized to demonstrate the transfer of heat and mass from carcass. As both wind and no-wind working conditions were included in each experimental record, a constrained expectation maximization algorithm was proposed to simulate different working conditions simultaneously, i.e.: E step to construct the Wind CPV (WCPV) and M step to construct the No_Wind CPV (NWCPV); in both E step and M step, partial order constraints were considered, and a cloud model based on evolutionary algorithm CBEA was utilized as the optimization algorithm. Based upon 37 experimental records, the CPVs were constructed and applied to simulate the chilling process of pig carcasses, the mean error was 0.93℃ for cooling process, 1.51‰ for loss process. Meatpacking companies can apply this approach to analyze the practical pig chilling process data.