Real-time monitoring of directed energy deposition additive manufacturing process using multiple sensors and machine learning

Additive manufacturing (AM) of components via Directed Energy Deposition (DED) is a complex process involving long deposits of single-walled tracks. DED has been gaining popularity due to its ability to build functionally graded parts as well as repair components. Depending on whether the feedstock material is powder-blown or wire-fed, the source of focused energy can be a laser, electron beam, plasma or electric arc. Regardless, the complex melt pool dynamics, high cooling rates, power fluctuation, changes in feedstock flow, gas flow, etc. contribute to process instability and failed parts. It has been reported that for metal AM, the defect rate can be as high as 40%. For instance, it is seen that keyhole porosity occurs due to metal vaporization from high energy whereas surface-level defects such as high roughness are caused by low energy input. The feasibility of measuring and controlling layer height has been demonstrated when researchers at the University of Tennessee, Knoxville implemented a support vector regression model to correlate voltage and current with contact-to-workpiece distance. The group also deployed a thermal camera to study the effect of interpass temperature and dwell time showing that higher temperature on previous layers results in deeper penetration of subsequent layers to cause underbuilds. Data from sensors will be combined to develop machine learning models and understand the relation with process instabilities. Under and over building are major issues in DED process and the proposed project will develop models to measure and control bead shape in real-time to reduce issues that cause process instability and part failure. A validation build will be fabricated to test accuracy and performance of machine learning models in the measurement of bead shape and interpass temperature in controlling bead shape.