Posters
« Back
Hybrid Predictive Modelling of Geometry with Limited Data in Cold Spray Additive Manufacturing
EP35868
Poster Title: Hybrid Predictive Modelling of Geometry with Limited Data in Cold Spray Additive Manufacturing
Submitted on 28 Feb 2021
Author(s): Daiki Ikeuchi
Affiliations: The University of Sydney; Commonwealth Scientific and Industrial Research Organisation
This poster was presented at 4th International Conference on Materials and Intelligent Manufacturing
Poster Views: 125
View poster »


Poster Information
Abstract: Cold spray additive manufacturing is an emerging technology that offers unique advantages, including high production rate, unlimited product size and the ability to process oxygen-sensitive materials. However, dimensional control and accuracy in cold spray additive manufacturing are challenging, which limits its integration into commercial manufacturing systems. These problems originate from the poor understanding of the complex relationship between process parameters and the resulting fabricated geometry. This knowledge gap motivated the development of an accurate predictive model for the geometry of a cold spray track profile to overcome the problems. Recently, a machine learning approach has gained interest in developing the predictive model of such a complex additive manufacturing process due to its superior nonlinear mapping capability, as seen in other manufacturing applications. Nevertheless, such a mapping capability can be realised only with a large amount of experimental data which is often impractical to collect in additive manufacturing applications. This limited data issue has motivated the exploration of a data-efficient machine learning approach suitable for complex process modelling with limited data. Therefore, the objective of this study was to investigate a data- efficient machine learning approach to geometry prediction in cold spray additive manufacturing. The proposed approach was of hybrid modelling framework, incorporating a conventional mathematical Gaussian model into the development and learning process of a data-driven model. We compared to purely mathematical Gaussian and data-driven modelling results and showed that the proposed hybrid modelling approach provided improved predictive accuracy. The findings can contribute to the control and optimisation of the process for shorter production time and the development of build strategy for better as-fabricated surface and dimensional quality control. The approach in this study is also applicable in other deposition-based additive manufacturing technologies such as Wire and Arc Additive Manufacturing.Summary: Hybrid Predictive Modelling of Geometry with Limited Data in Cold Spray Additive Manufacturing using Machine LearningReferences: For Further Information:
(Open Access Journal Article) Data-Efficient Neural Network for Track Profile Modelling in Cold Spray Additive Manufacturing - https://doi.org/10.3390/app11041654

Our Related works:
(Open Access Journal Article) Neural Network Modelling of Track Profile in Cold Spray Additive Manufacturing - https://doi.org/10.3390/ma12172827
Report abuse »
Creative Commons