10 Février – Thesis defense - Pinku Yadav

10 h30 Amphi HSE - IUT Bordeaux (Gradignan)

Drift detection in the L-PBF process using in-situ monitoring and data analytic techniques.

Metal AM provides a high degree of freedom in terms of design flexibility and functionalities. It is recommended to enhance the commercial machines with quality assurance instruments to achieve the full potential of metal AM.  Nowadays, machine suppliers focus on improving quality assurance and providing more insights into the process space with in-situ machine controls. It led to improved part quality, reliability, and repeatability per consumer’s confidence-building measures. However, the full potential of the quality control instrumentation cannot be released if the captured in-situ data is not treated in a meaningful way and correlated with different process steps. It is vital to correlate the in-situ signal to process characteristics such as anomalies to monitor and increase the overall quality of the final part.
This thesis presented a methodology to detect and identify the process anomalies in the Laser-Powder Bed Fusion process using the commercial in-situ instrumentation. First, it is essential to find the best-suited parameters to study the correlation between the process space, including the process parameters (laser power, scanning speed, feedstock material, etc.) and process outcome (melt pool signals). Initially, an investigation based on the microstructural and mechanical aspects of AlSi7Mg0.6 alloy is carried out, and the best-fit scanning strategy is identified for further study.
Different types of in-situ instrumentation such as infrared cameras and photodiode-based commercial systems are studied to understand the laser-powder interaction better. Camera-based co-axial EOSTATE Optical Tomography supplied by EOS GmbH is exploited, and a case study based on data analytics techniques is proposed to identify the potential drift layers.  Similarly, machine learning-based methodology is proposed to extract critical features at a global and local scale for photodiode-based Melt Pool Monitoring in-situ module supplied by SLM Solutions GmbH. As we know, the laser-powder interaction is dependent not only on process parameters but also on the powder bed spread. However, powder bed spread quality can be influenced by feedstock characteristics such as humidity, granularity, .etc. Therefore, it is vital to correlate the metal pool signal to powder spread health to understand better. To do so, the critical characteristics of the powder bed spread are identified using Layer Control System and computer vision algorithms. The proposed methodology successfully identifies powder bed spread anomalies and their influence on the melt pool signal. Also, a case study is presented to investigate the efficiency and robustness of the proposed methodology.
Additionally, to provide the experimental data for the constitutive-based modeling approach (Work Package 1), the thermomechanical response of the L-PBF processed Inconel 718 alloy is investigated. The influence of deformation conditions, i.e., temperatures and strain rates, are studied using hot torsion tests. It is noticed that the deformation conditions strongly influence the microstructural refinement influenced by dynamic recrystallisation and recovery. Also, the material's initial state, such as production state (in our case, the L-PBF process), plays a critical role.  The L-PBF processed Inconel 718 samples showed a deformation activation energy of 353 KJ/mol-1. The samples deformed at 1000  ̊C showed the highest elongation %, which can be attributed to the dynamic behavior of the material.

Event localization