【作者】 赵欣；

【导师】 黄金杰；

【作者基本信息】 哈尔滨理工大学 ， 计算机软件与理论， 2025， 博士

【摘要】 3D打印技术作为一项深刻影响现代工业体系和生产方式的先进制造技术,有效推动了产品设计、制造与应用模式的创新与变革。与传统制造方式相比,3D打印可实现从数字模型到物理产品的一体化制造,有效缩短了产品开发周期,降低制造过程中的复杂性与成本。这些优势使其在个性化定制、小批量生产以及快速原型制作中不可替代。自20世纪90年代以来,3D打印技术已逐步拓展至航空航天、医疗器械、建筑工程、消费电子等行业。随着工业4.0和智能制造的推进,3D打印被视为未来制造业发展的关键技术之一。尽管3D打印技术展现出了诸多优势,但在实际生产中仍然面临一系列关键挑战,包括几何数据的处理效率低下、分层规划的精度不足、结构参数优化复杂以及热效应控制困难等问题。这些问题不仅影响了3D打印的成型质量和生产效率,也制约了其技术优势的进一步发挥及产业规模化应用。为解决这些问题,本文围绕3D打印工艺规划中的关键环节进行系统性优化研究。结合计算机科学中的计算机图形学、数据结构与算法、数值分析建模与优化和进化算法等理论基础,提出3D打印工艺规划的优化方法。具体研究内容如下:(1)针对三维几何模型中存在大量相似结构,导致数据冗余高的问题,提出了一种基于相似特征聚类的几何文件压缩编码方法。首先,设计了相似特征聚类方法,通过识别模型中相似或相同的面片并建立其变换关系,从而实现对相似特征的聚类表示;同时,设计了基于哈希编码的网格拓扑数据结构,减少了相似面片数据的存储。实验结果表明,与压缩前相比,文件体积可减少70%以上。(2)针对分层规划中打印精度和效率较难平衡的问题,提出了一种基于特征识别的混合自适应分层方法。通过在离散三角网格上识别模型的高曲率区域和孔洞特征,设计适应于不同特征区域的自适应分层算法,从而针对不同区域灵活调整分层规划。在关键特征区域,采用基于体积最优的自适应分层算法,以提高打印精度;在非关键特征区域,采用基于局部误差平衡的自适应分层算法,以确保打印质量与打印效率的平衡。实验结果表明,该方法在不显著增加分层层数的前提下,显著提升了分层打印的精度,减少了分层时的阶梯效应。(3)针对3D打印过程中结构工艺参数复杂和多目标优化困难等问题,提出了一种基于沉积过程分析的结构工艺参数优化方法。首先,设计了沉积过程分析算法,以分析3D打印过程中制件的拉伸变形、打印时间和材料消耗;接着,使用基于响应曲面法(Response Surface Methodology,RSM)的改进预测模型,建立结构工艺参数与响应目标之间的多项式回归关系,以减少沉积过程分析次数并引导优化过程;最后,采用基于参考向量进化算法(Reference Vector Guided Evolutionary Algorithm,RVEA)的改进搜索策略对多目标优化问题进行求解,并引入角度惩罚距离来平衡解的收敛性和多样性。实验结果表明,沉积过程分析结果与实际打印测试结果之间,在拉伸变形、打印时间和材料消耗方面的误差分别约为7.92%、2.74%和1.01%。此外,沉积过程分析结果与预测模型结果之间,在拉伸变形、打印时间和材料消耗方面的误差分别约为6.16%、6.93%和5.48%。(4)针对3D打印过程中热效应参数复杂和多目标优化困难等问题,提出了一种基于温度场分析的热效应工艺参数优化方法。首先,设计了温度场分析算法,结合生死单元技术,以求解3D打印过程中制件的翘曲变形、von Mises应变和弹性应变;接着,基于高斯过程回归(Gaussian Process Regression,GPR)的改进预测模型和自适应拉丁超立方体采样(Adaptive Latin Hypercube Sampling,ALHS)算法,构建热效应参数与响应目标之间的回归模型,以减少温度场分析次数并提高预测精度;最后,采用基于萤火虫算法(Firefly Algorithm,FA)的改进搜索策略对多目标优化问题进行求解,结合随机飞行策略以实现热效应工艺参数高效搜索。实验结果表明,温度场分析结果与预测模型结果之间,在翘曲变形误差、von Mises应变和弹性应变方面的平均误差分别约为0.92%、0.22%和0.39%。综上,本文针对3D打印工艺规划中的关键问题,提出了一系列优化算法和解决方案。通过在数据处理、分层规划、结构参数和热效应参数等方面的优化方法研究,显著提升了3D打印的效率和制件质量,为3D打印技术的发展提供了进一步的技术支撑和理论依据。更多还原

【Abstract】 As an advanced manufacturing technology exerting profound influence on modern industrial systems and production paradigms,3D printing effectively drives innovation and transformation in product design,fabrication,and application models.Compared with traditional manufacturing,additive manufacturing enables fully integrated production from digital model to physical artifact,substantially shortening development cycles and reducing both process complexity and cost.These advantages render it indispensable for personalized customization,small-batch production,and rapid prototyping.Since the 1990s,3D printing has gradually penetrated industries such as aerospace,medical devices,construction engineering,and consumer electronics.With the advent of Industry 4.0 and the rise of smart manufacturing,additive manufacturing is widely regarded as one of the key technologies for the future of the manufacturing sector.Despite its many strengths,3D printing still faces a series of critical challenges in practice—among them,low efficiency in handling geometric data,insufficient precision in slicing and layer planning,complexity in structural parameter optimization,and difficulty controlling thermal effects.These challenges not only compromise build quality and production efficiency but also constrain the full realization of the technology’s advantages and its scalable industrial deployment.To address these issues,this paper conducts a systematic optimization study of the key stages in 3D printing process planning.Drawing on foundational theories from computer science,including computer graphics,data structures and algorithms,numerical modeling and optimization,and evolutionary algorithms,we propose several novel methods for improving additive manufacturing planning.The specific contributions are as follows:(1)Geometric File Compression via Repetitive-Feature Clustering.To tackle the high data redundancy caused by extensive repeated structures(e.g.,symmetric parts,lattice infill,thread textures)in 3D models,we introduce a compression-coding scheme based on repetitive-feature clustering.First,we establish a repetitive-encoding framework that identifies similar or identical facets,clusters them according to their geometric transformations,and represents each cluster by a single reference facet plus its transformation parameters.In parallel,we design a hash-based topological data structure for the mesh to store connectivity information compactly.Experimental results demonstrate that this approach preserves model integrity while reducing file size by over70%compared to the original uncompressed geometry.(2)Hybrid Adaptive Slicing Based on Feature Recognition.Balancing precision and efficiency in layer planning remains difficult.We propose a hybrid adaptive-slicing method that first detects high-curvature regions and hollows on the discrete triangular mesh,then applies region-specific adaptive algorithms.In critical feature areas,a volume-optimal adaptive slicing algorithm is used to maximize accuracy;in noncritical zones,a local-error-balancing adaptive slicing ensures a compromise between print quality and speed.Experiments show that our method significantly improves slicing precision and reduces the staircase effect without markedly increasing the total number of layers.(3)Structural Process-Parameter Optimization via Deposition-Process Analysis.The structural parameters of the deposition process involve complex multi-objective trade-offs.We develop a deposition-analysis algorithm to quantify part deformation(e.g.,tensile warping),build time,and material consumption during printing.Based on this analysis,we construct an improved predictive model using Response Surface Methodology(RSM)to approximate the relationship between process parameters and response targets,thereby reducing the number of expensive deposition analyses and guiding the optimization.Finally,we solve the resulting multi-objective problem with an enhanced Reference-Vector Guided Evolutionary Algorithm(RVEA),incorporating an angular-penalty distance metric to balance convergence and diversity.On average,the errors between deposition-analysis predictions and physical test results for tensile deformation,build time,and material usage are approximately 7.92%,2.74%,and 1.01%,respectively.Moreover,the predictive-model results differ from the deposition-analysis outcomes by only about 6.16%,6.93%,and 5.48%,respectively.(4)Thermal-Effect Process-Parameter Optimization via Temperature-Field Analysis.Thermal effects introduce further complexity and multi-objective challenges.We first design a temperature-field analysis algorithm,augmented with element activation(“birth–death”)techniques,to calculate part warpage,von Mises strain,and elastic strain during printing.We then build an enhanced Gaussian Process Regression(GPR)predictive model,coupled with Adaptive Latin Hypercube Sampling(ALHS),to model the mapping from thermal parameters to response targets.This surrogate reduces the frequency of costly temperature-field analyses while improving prediction accuracy.Finally,an improved Firefly Algorithm(FA),augmented with a random-flight strategy,efficiently searches the multi-objective space.Experimental validation shows that the average errors between temperature-field analysis and surrogate predictions for warpage,von Mises strain,and elastic strain are approximately 0.92%,0.22%,and 0.39%,respectively.In summary,this work presents a suite of optimization algorithms and solutions targeting the critical issues encountered in 3D printing process planning.By advancing methods in geometric data handling,adaptive slicing,structural-parameter optimization,and thermal-parameter optimization,we substantially enhance both the efficiency and quality of additive manufacturing,providing further technical support and theoretical grounding for its continued development.更多还原