Characterization of the EBM Additive Manufacturing Process Parameters using Artificial Neural Networks

Additive manufacturing also known as 3D printing is an emerging technology as it has great potential in a large variety of applications in aerospace, biomedical and automotive industry. Additive manufacturing enables the production of complicated designs, lightweight structures and novel functional features by joining materials in a layer by layer fashion to make the objects from 3D models. Nowadays there are various technologies which can implement AM using different kinds of materials, such as Powder Bed Fusion (PBF) which includes, selective laser sintering (SLS) and Electron Beam Melting (EBM). With the rapid development of AM, more complex structures with more materials can be fabricated. Specifically the Electron beam used to melt the powder in EBM is much more powerful than the laser and can result in strong mechanical properties of manufactured part with high density, offering minimal waste. However, ensuring AM part quality in EBM still remains the big challenge in the major industrial breakthrough of additive manufacturing technologies. Recently, the machine learning (ML) that is a growing field of artificial intelligence (AI) is becoming more and more important not only in Additive manufacturing but in overall manufacturing industry, mainly because of ability of ML to perform the complex task such as regression and classification. The basic purpose of Machine Learning is to make a system able to learn useful features from a data set identify the patterns and accordingly make the decisions. Currently, ML is applied in various processes of additive manufacturing like for monitoring of the AM process and making it able to take decisions on the basis of data coming from different cameras and sensors, like for defects and anomaly detection. In this thesis a framework of optimizing the influential parameters impacting on the final part quality of EBM produced parts has been implemented. First, all of such important parameters have been identified by different analysis techniques. Afterwards, the co-relation analysis of these parameters was carried out in order to find the impact of variance of these parameters on different part qualification categories. Finally, using this data a regression model was developed by comparing different machine learning algorithms to find a network that could predict the process parameter anomaly, before going towards the failure of the part.