ACTIVITY MONITORING SYSTEM USING NEURAL NETWORK IN PRODUCTION SHOP FLOOR

Recently, high-mix low-volume production system is introduced in many factories, because manufacturing companies have to meet various needs for consumers. In high-mix low-volume production system, there are many setup operations & maintenance operations. Therefore, production overhead costs like set-up operation are increasing. Activity Based Costing (ABC) and Activity Based Management (ABM) which can accurately consider those costs are attracting a lot of attentions as the new costing method. But these have some problems. High running cost is one of problem in ABC and ABM, because it is difficult to measure an activity consumption. In this study, we develop "Activity Monitoring System" that is able to use at low cost.. Concretely, we collect monitoring data at the line production by ray sensors, ammeters and passage sensors. And the monitoring data is used to distinguish activities of workers. Additionally, the proposed algorithm including neural network estimates the activity if the activity is not able to distinguish. The effectiveness of the proposed monitoring system is evaluated. As a result, we found quantity of monitoring data on neural network required for estimation with the high precision.