Data-Driven Divide-and-Conquer for Estimating Build Times of 3D Objects.

Accurate estimation of build times of 3D objects is of great importance in the different phases of an additive manufacturing process. The existing physics-based models can very precisely tackle this task but at the cost of spending a considerable amount of time. An alternative solution is to use a data-driven machine learning method for build time estimation (BTE). However, estimating build times of a dataset of objects with diverse and heterogeneous characteristics is a challenging task for a single learning algorithm. The aim of this study is therefore to investigate the value of the divide-and-conquer strategy in partitioning the dataset into subsets of homogeneous objects to facilitate the BTE task for the examined learning models. The comprehensive experiments performed in this study prove that this strategy is indeed capable of providing accurate BTEs and can outperform the performance of a single learning method trained with all objects in the dataset.