The focus of this paper is process improvement in a fish-sheet production line in a snack factory in Thailand. Preliminary measurement suggests that moisture of fish sheets after the roasting process significantly deviate from the provided specification, reporting a process capability index (C pk ) of post-roasting moisture to be only 0.04. Prior to the start of this project, the manufacturer relied on machine operators to subjectively adjust production parameters during production to reduce

more »

... sture variation in fish sheets. Despite experience of operators, data suggest that the method is ineffective in reducing moisture variation. Moreover, it is found that significant variation originates from frozen minced fish which is a key ingredient of fish sheets, resulting in high moisture variation in mixed ingredient. Two intervention procedures were implemented to improve the post-roasting moisture C pk index. First, the mixing procedure was modified to allow flexible adjustment of water in mixing ingredients based on objective measurements. Second, post-mixing processes are operated under constant parameter values. Optimal values are solved using response surface method based on 3x3 factorial design of experiments of air-drying temperature and roasting conveyor belt speed. The C pk index is predicted to improve to 0.59 after the water-adjusting intervention and production parameters are set up so that the air-drying temperature is set to 61 º C and the roasting conveyor belt speed is set to 64 Hz.  Index Terms-design of experiment, food manufacturing, mass-production manufacturing, moisture variation control, response surface method Thananat Jitapunkul received B.S. and M.Eng. degrees in electrical engineering and computer science from Massachusetts Institute of Technology, Cambridge, MA in 2010. He is pursuing a master's degree in Engineering Business Management at Regional Centre for Manufacturing System Engineering, Faculty of Engineering, Chulalongkorn University, Bangkok Thailand. His current research interest is optimization of manufacturing process using statistical and machine learning techniques.