基于强化学习的航空零件复合材料缠绕机床运动优化方法

Motion Optimization Method of Six-axis Fiber Placement Machine for Aerospace Components Based on Reinforcement Learning

大型结构件自动铺缠是航空工业中重要的自动化工艺,这些结构件中存在曲率较大的加工区域,加工过程中产生的剧烈转角变化降低了实际加工速度,增加了加工时间。文中研究了一种针对六轴铺缠机床的优化方法,目的在于保证较大负载转轴运动平稳的前提下,提高加工效率。在刀具路径已经确定的情况下,依据刀具路径和机床运动性能搭建用于强化学习的简单仿真环境,使用PPO策略决策在不同状态下分配转台轴的运动量,进而规划出相应的机床各轴加工轨迹。选用铺缠路径对两种不同分配冗余轴的策略进行比较,验证效果表明,强化学习规划路径能够在保证实际进给速度的前提下有效降低关节的速度波动。

The fiber placement of large structural components is widely used in the aerospace industry. Since the structural components have large volume and complex surfaces, a six-axis machine tool with redundant degrees of freedom(DOFs) is selected for the fiber placement process. This paper establishes an optimization method for the motion of redundant DOFs to improve the efficiency of the machine under the premise of smooth movements of shaft with heavy loads. With a determined tool path, a reinforcement learning model and a simulation environment is built. The PPO strategy is used to determine the movements of redundant rotation axes. Compared with generic algorithm(GA), the proposed method can reduce speed fluctuating and process time.