Pipe Routing Design for Aero-Engine Using Monte Carlo Tree Search Method

Abstract

Pipe routing design (PRD) is a complex process that involves a large search space and requires experienced professionals. Despite advancements in aero-engine design, PRD remains a stage that is not completely automated. In this article, we present a new approach to PRD by formulating it as a Markov decision process and proposing a pipe routing design method based on Monte Carlo Tree Search (PRD-MCTS). Firstly, this paper uses an intelligent algorithm to look for enough paths for each pair of joints. Secondly, PRD-MCTS regards each path as a candidate choice, and then PRD-MCTS randomly chooses a path and calculates the probability of collision-free routing until the set time. The method selects the path with the highest probability and updates the environment for the next selection. A simplified environment from the aero engine verifies the correctness of the methods.