摘要
现有振动式给料柔性上料系统的工作效果多数依靠出厂前人工手动调试,手动振动参数调试存在耗时长、费人力问题,且调试结果有较强针对性,导致系统柔性能力不足。提出一种基于改进Q-learning算法的柔性上料系统,结合柔性上料系统自身特性对传统Q-learning算法的奖励函数和ε-贪婪策略进行改进,令柔性上料系统自学习寻找一组有较优料件振动效果的振动参数。由实验可得,所提算法能减少人工在柔性上料系统调试上的参与度,证明了该算法在柔性上料系统中的可行性和有效性,且与传统Q-learning算法相比较,改进Q-learning算法更符合柔性上料系统的实际应用。
Most of the existing vibrating feeding flexible feeding systems rely on manual debugging before delivery.Manual vibration parameter debugging is time-consuming and labor-intensive,and the debugging results are highly targeted,resulting in insufficient flexibility of the system.A flexible feeding system based on improved Q-learning algorithm was proposed.Combined with the characteristics of flexible feeding system,the reward function of traditional Q-learning algorithm and ε-the greedy strategy was improved so that the flexible feeding system can learn from itself to find a group of vibration parameters with better vibration effect.The experiment shows that the algorithm can reduce the participation of human in the debugging of the flexible feeding system,which proves the feasibility and effectiveness of the algorithm in the flexible feeding system.Compared with the traditional Q-learning algorithm,the improved Q-learning algorithm is more suitable for the practical application of the flexible feeding system.
作者
丁慧琴
曹雏清
徐昌军
李龙
DING Huiqin;CAO Chuqing;XU Changjun;LI Long(School of Computer and Information,Anhui Polytechnic University,Wuhu 241000,China;Wuhu Robot Technology Research Institute,Harbin Institute of Technology,Wuhu 241000,China)
出处
《现代制造工程》
CSCD
北大核心
2023年第4期87-92,129,共7页
Modern Manufacturing Engineering
基金
安徽省教育厅科学研究重点项目(KJ2020A0364)。
