RapidMiner framework for manufacturing data analysis on the cloud

This research proposes a manufacturing data analysis framework in the form of computing blocks. The aim is to identify the parameters/attributes that affect the production yield as well as the root cause of the manufacturing problems. The framework is designed to be flexible and exploit the cloud as a computing platform. The manufacturing data are obtained from the database of the production lines in the food industry and pre-processed. Then, the correlation analysis and the decision tree algorithm are applied. The root cause parameters of yield degradation are identified in the form of decision rules. In addition, the analysis framework is built based on the workflow concept where several computing blocks can be linked together to form a workflow graph. Each computing block can be tailored made to fit the food manufacturing data. RapidMiner, a GUI-based tool for data mining, is selected as the workflow engine. The designed statistical analysis modules are then built as plugged-ins to RapidMiner. To construct a workflow, the plant engineers can use basic data mining modules as well as our custom designed ones. Moreover, RapidAnalytics is customized to allow the computing blocks to be scheduled onto the private cloud. The designed framework is thus flexible and suitable for agile manufacturing process in Thailand's food industry.