Room For Improvement

At MRL, we see opportunities for innovation in every problem we tackle. Process optimization, particularly qualification-aware process optimization, is the ultimate challenge for the additive manufacturing community and has been a focus at MRL since entering the field. The traditional process optimization workflow involves additively manufacturing numerous coupon samples over a wide range of values for a few important parameters – typically laser power, scan speed, and hatch spacing. Each of these samples must then be characterized to determine the parameter combination that results in the ideal properties. This is a time- and cost-intensive process that helped inspire the development of MRL’s technology for high-throughput Microstructure Characterization and Mechanical Testing, to improve the efficiency of this method.

These high-throughput tools remain critical for fast and reliable qualification, yet we recognized further opportunity for improvement in the process optimization workflow. Far more parameters than power, scan speed, and hatch spacing affect the microstructure and properties of an additively manufactured part. Our ICME software, iCAAM, has been developed as the perfect marriage between modeling and data science for, not only fast prediction of microstructure and properties, but also multi-objective optimization in the large design space of laser powder bed fusion. Model-based process optimization with iCAAM can identify the best parameter sets before printing, which saves time and material cost.

Changing Verification

Parallel to these modeling efforts was our research and development of Sensors to monitor the health of our machines and learn more about machine behavior and material response. By using Machine Learning and Data Analytics tools to increase both the speed of modeling predictions and the efficiency of sensor data collection and analysis, we are now working toward live process optimization. In this case, a fully closed loop between machine, sensors, and iCAAM will allow for in-situ detection and assessment of anomalies in sensor data, live prediction of the impact on the structure and properties of the part, and machine learning-informed decision making on how to move forward. The next step may involve changing the process parameters of the machine according to iCAAM predictions, pausing the build to wait for operator input, or cancelling the build to save time material if a critical defect has been detected. This is the future that we see in process optimization, and we are working to bring it closer every day.