Tampere University of Technology

TUTCRIS Research Portal

Probabilistic Modelling of Defects in Additive Manufacturing: A Case Study in Powder Bed Fusion Technology

Research output: Other conference contributionPaper, poster or abstractScientific


Original languageEnglish
Number of pages6
Publication statusSubmitted - 31 Dec 2018
EventCIRP Conference on Manufacturing Systems -
Duration: 1 Jan 2000 → …


ConferenceCIRP Conference on Manufacturing Systems
Period1/01/00 → …


Implementation of additive manufacturing into product manufacturing suffers from the challenge of part defects prediction. Due to interdependencies of design variables and manufacturing parameters in achieving suitable part quality, modelling methods are necessary to provide simulation capabilities for part quality analysis at early stages of product development. A systematic methodology is proposed to extract cause-effect relationships among variables and to transform this causal model into a Bayesian network. The Bayesian network is then used to predict the effect of specific design and manufacturing parameters on part defects and to estimate the needed input parameters backwards, based on acceptable output values.


  • Additive manufacturing, Powder bed fusion, Probabilistic model, Bayesian network, Dimensional Analysis Conceptual Modeling