Automatic Route-finding with Non-monotonic-trend Factor for Large-scale Plant Maintenance Task

When a component is replaced in the maintenance of a large plant, for example several dozen meters size, the replacement-operation's cost is highly affected by its transportation route in the plant. When a component is transported by an overhead-crane, the number of corners in the route is desired to be minimized because it represents the number of required lanes, which affects the cost. Given that the corner count is minimized, maximizing average collision margin, which is represented by the average distance between the component and the obstacles in the plant, is required to reduce the workload of adjusting component's postures to avoid collision. The route length comes as the third factor. Dijkstra method, which is a well-known route-finding method, only accepts factor which increases or decreases monotonically during route-finding. While factors with monotonic trend such as corner count and route length can be optimized by Dijkstra method, we could not optimize average collision margin which is non-monotonic-trend factor in our previous work. In this paper, we propose an improvement on route-finding method so that it can optimize multiple factors for plant maintenance including average collision margin, a factor with non-monotonic trend. The proposed method optimizes average collision margin using Dijkstra method by introducing cascade evaluation of multiple factors with vector cost function and limiting the update condition for average collision margin during route-finding. The proposed method is evaluated with an actual plant data. The evaluation result shows that the average collision margin of a route found by the developed method was improved by maximum 17% with the same corner counts compared to the previous method using a weighed sum of the factors.