Research On 3D Printing Path Planning Algorithm For Complex Thin-walled Models

3D printing has been widely used in many fields.Path planning is an important part in 3D printing.the path planning method optimization can not only improve the molding quality,but also improve the printing efficiency.However,the traditional path planning method is not ideal when printing complex thin-walled structures.Based on the intelligence of reinforcement learning,we propose a path planning method for complex thin-walled structures based on Q-learning.Firstly,we transform the path planning task in reinforcement learning into a full-traversal problem,the optimization goal is to improve the printing efficiency and molding quality,that is,to minimize the total number of lift and turn of the print head,we design the corresponding printing optimization mechanism to punish the action of lift and turn.Finally,we establish the simulation environment of a slice,use the Qlearning algorithm with the printing optimization mechanism to find the best path.Experiments are carried out on different thin-walled models,and test the algorithm performance in terms of the complexity of the model.The results show that the performance of our algorithm is superior to that of the traditional method when printing complex thin-walled structures.On this basis,we propose a path strategy optimization algorithm combining Double DQN and Dueling DQN.First,establish a prediction network and a similar target network.we improve the stability of the algorithm by delaying the update of the target network,use the experience replay pool to accelerate the convergence.We solve the problem of over-estimation of Q value by finding out the action with the maximum value in the main network and then calculating the Q value of the action in the target network.In addition,add two additional subnetworks to the original network structure to evaluate the value stream and the action advantage stream respectively.Separating values can significantly improve the learning efficiency of the algorithm when the agents take different actions but their corresponding values are equal.Expirements on several different thin-walled models and in terms of the number of model layers show that the proposed strategy optimization method can further improve the efficiency and accuracy of the algorithm.