A novel method for intelligent reasoning of machining step sequences based on deep reinforcement learning

Abstract

High-quality and efficient process planning methods are crucial for ensuring product manufacturing quality. However, traditional methods have several drawbacks, namely, they are time-consuming, highly dependent on expert experience, and involve considerable repetitive workloads. To overcome these limitations and enhance the efficiency and intelligence of process planning for complex structured parts, this study proposes a machining step sequence reasoning method based on deep reinforcement learning. First, historical process data are preprocessed to convert the knowledge stored in the process files into structured and vectorized data. Second, the process routes and feature step sets serve as inputs, and a proximal policy optimization algorithm is employed to train the historical process instances. The sequencing patterns discovered during training are then integrated with advanced sorting strategies to efficiently generate the machining step sequences. To evaluate the effectiveness of the proposed method, 50 complex structured parts were tested, with 25 representative parts selected for detailed comparative analysis. The training performance of the proposed algorithm was evaluated against those of the advantage actor-critic and soft actor-critic algorithms. In addition, the reasoning results of various state-of-the-art algorithms were analyzed using these test cases. Experimental results demonstrate that the proposed method is effective and competitive for process planning of complex structural parts. Therefore, this study provides practical guidance for enhancing the efficiency and intelligent automation of process planning of complex structural parts.