Additive manufacturing (AM) offers an efficient way of building complicated geometry in a shorter and cost-effective manner. Though process parameters could be optimised to decrease dimensional deviation of fused deposition modelling (FDM) manufactured components, getting acceptable dimensional accuracy from FDM remains a challenge compared to other AM processes. In this study, a geometric benchmark part was designed based on the NIST (National Institute of Standards and Technology) which comprises different geometrical features, and it was fabricated by varying process parameters on an FDM machine. Dimensions of these features on a fabricated benchmark component were determined using the coordinate measuring machine, and the machine learning (ML) approach was employed to predict the dimensions and deviation. ML models built (linear regression, LASSO, Ridge, random forest, XGBoost) based on the measured dimensions were able to predict the geometrical dimensions (length, width, height, diameter, thin walls/ slots, angle) and dimensional deviation without fabricating the components.

增材制造 (AM) 提供了一种以更短且具有成本效益的方式构建复杂几何形状的有效方法。尽管可以优化工艺参数以减少熔融沉积建模 (FDM) 制造组件的尺寸偏差，但与其他 AM 工艺相比，从 FDM 获得可接受的尺寸精度仍然是一个挑战。在这项研究中，基于 NIST（美国国家标准与技术研究院）设计了一个几何基准部件，该部件包含不同的几何特征，并通过 FDM 机器上的不同工艺参数制造。使用坐标测量机确定制造的基准部件上这些特征的尺寸，并采用机器学习 (ML) 方法来预测尺寸和偏差。建立的机器学习模型（线性回归，