Research On Key Technologies Of Geometry Computing In Fused Deposition Modeling

3D printing is a hot research field at present. Fused deposition modeling as one of the most widely used techniques for 3D printing, along with its low costs, convinient maintenance, has gained widespread concern and attention, however, there are still some shortages in fused deposition modeling. That the manufacturing precision is not so accurate is the most obvious problem. Further, the model can not be printed without inside or outside support. To address these problems, five algorithms was proposed, which aim at improving the printing precision, saving the printing consumable materials and printing time, and making the manufacture product closer to the original physical structure. The main contributions of this dissertation are as follows:(1) A novel algorithm for calculating the optimal oriented placement of the model was proposed. By analyzing the support structure material, the model surface printing supplies, the printing time, the model surface accuracy, the salient features and the processing costs of the model, the algorithm constructs the relation function between these six indexes and the oriented angle of the model. Then an optimized objective function of the model was introduced and obtained using the improved Powell method. Experimental results demonstrate that the proposed algorithm can acquire the optimal placement angle ofthe model, improve the printing accuracy, save the printing consumable materials and printing time, and reduce the processing time.(2) An algorithm generating a stable and low costs support structure was proposed. Defined four fuse print constraints by analyzing the influencing factors of fuse deposition forming. The support structure was divided into three categories according to the constraints, then calculate the minimum fuse support areas according to the four constraints. Finally, the external support was generated by using the minimum spanning trees to connect these three types of support structures. Compared with the traditional algorithms, the proposed algorithm takes into consideration of the most basic elements of forming "fuse" and ensures the intact fuse forming, resulting in smaller supporting areas and better supporting effects. The experimental results show that the generated support structure can stably support the model and the materials-consuming, time-consuming are less than the traditional algorithms.(3) Proposed an algorithm divided the models that beyond printing space into small parts, which can be put into the print space and printed without external support. Firstly, partition the model surfece using the region growing method and analyze the normal vector direction in order to determine the candidate dividing directions, select the candidate dividing direction planes to segment the model, if the sub-model after division is not pyramidal. Then use the same method to segment the sub-model until all of the sub-model is pyramidal. Adopting beam search method to construct the dividing trees, and using the evaluation function to appraise the dividing nodes, optimal partition was created. The experimental results show that the algorithm can divide the model into small parts that fit into the printing space and can be printed without any supports.(4) According to the demand of printing hollow objects, proposed an algorithm divided the model into parts which can be printed without internal support. Firstly, we utilized region growing scheme to derive a set of candidate surfaces, whoseinner surfaces can be printed without support.. We then employed Monte Carlo tree search (MCTS) over the candidate surfaces to obtain the optimal set cover. All possible candidate subsets of exact cover from the optimal set cover were finally obtained and the bounded tree was used to search the optimal exact cover that is the optimal segmentation. The experimental results show that the algorithm can be used in various kinds of models, can print hollow models without internal supports, and save the printing materials and printing time.(5) A low cost compact packing algorithm for printing multi models in one printer. The placement of the model is represented by vectors. The solution of the packing problem can be represented by a set of rotation vector and displacement vector. A fast methods for calculating the volume of the external support of the models in one printer and for calculating the size of the bounding box of the packing models are put forward as the solutions to the value functions. The problem is solved by local learning particle swarm algorithm with dynamic neighborhoods. Finally the optimal packing plan was obtained. The experimental results show the proposed algorithm can pack the model tightly in the printing space, effectively improve the print space utilization.