Abstract
A work planning system for decommissioning the Fukushima Daiichi Nuclear Power Station has been developed. This system, designed to formulate robust plans without experience using typical work-centred decommissioning methods and processes, is a significant step towards improving work plans, including continuously dismantling equipment and related resources, such as remotely operated robots and human workers.
The processing speed for optimising the work schedule is proportional to the square of the number of activities. In an inexperienced project, such as the Fukushima Daiichi decommissioning project, the person-hour calculation formula for the dismantling quantities has yet to be determined. Hence, we need to examine how long hours it takes to complete each activity with a mock-up facility and physical simulation corresponding to several scenarios. Moreover, we must evaluate optimal task durations by combining potential activities after acquiring possible person-hours for a specific scenario. It has been proven that optimising a hundred activities per month for typical dismantling work took several hours. For this reason, optimisation processing applied to many activities simultaneously has the problem of being unrealistic.
We have introduced a novel model that effectively expresses the process's flow, addressing the abovementioned issue. We calculated the work time using a discrete event simulation (DES) calculation by adjusting the parameters related to processing throughputs. In DES, multiple robot operations for dismantling equipment by various operators can be modelled with a state flow diagram. This diagram includes waiting time for events, delay time, and storage capacity limitations for repeated activities. Based on this state flow model, DES can calculate the total durations of complex combinations of activities for multiple scenarios. Since the system state is changed only when an event occurs to reduce calculation time, it is expected that simulation calculation can be performed at high speed rather than a continuous model.
