Abstract:Manufacturability analysis is a critical step before manufacturing to reduce costs and risks. It is used widely in conventional manufacturing (CM) processes. However, to the best of our knowledge, there is no natural method to evaluate the manufacturability of additive manufacturing (AM) processes that have more uncertainty-derived risks and costs than CM processes. A clear definition of the manufacturability of AM processes has not been established, and there is no standard to check whether a component is manufactured successfully by an AM process, particularly for porous complex components. This study introduces the development of a new machine learning-based method to solve the problem mentioned above. It is based on the statistical measurement of experimental samples. The proposed method can be used to perform the manufacturability analysis for periodic cellular structures printed by a selective laser melting (SLM) process. A novel definition of the manufacturability of the SLM-ed periodic cellular structure was proposed. Experimental results indicate that the developed learning model (ANN model) can achieve up to 94% classification accuracy and 96% prediction accuracy, which satisfies the application requirements of the AM industry. Moreover, the developed model can be adapted for the manufacturability analysis of different AM processes.

Development of a Manufacturability Predictor for Periodic Cellular Structures in a Selective Laser Melting Process Via Experiment and ANN Modelling