Process Optimization and Proactive Quality Control to Increase Investment Casting Throughput

Abstract

Blue Origin is an aerospace company with ambitious throughput goals in response to increased commercial space exploration. Pressure to increase throughput is especially apparent within its BE-4 engine business, as the engines support Blue Origin and its customers. Blue Castings is one of the primary in-house manufacturing plants that supports BE-4 production; the plant manufactures rocket engine components through a process called investment casting. Investment casting, by nature, is a complex process involving long rework times, high incidence of defects, and significant process variability. These characteristics contribute to the discrepancies between Blue Origin’s target BE-4 production rate, the production rate feasible at Blue Castings, and its actual delivery rate. This thesis explores how defect management and prevention techniques can improve throughput at Blue Castings and reduce the number of Blue Origin’s schedule delays attributable to Blue Castings. The research began with a baseline investigation and analysis of Blue Castings’ actual and best-case throughput rates compared to its goal. Two gaps were identified: 1) a gap between actual and feasible throughput, and 2) a gap between feasible and target throughput. The analyses highlight the need for better process and quality management to close both gaps. Through a mixed-method approach, the researcher explored and piloted process and data improvements to understand their impact on throughput. This included qualitative and quantitative data collection through on-site interviews, random sampling of defect data, and queries from the manufacturing execution system. With this data, the researcher investigated how machine learning can predict rework severity and support defect prevention. A case study on a selected part number demonstrated the potential to improve throughput by reducing unnecessary rework. By aligning stock-on surface criteria to downstream machining requirements, average rework loops were reduced from thrice the industry benchmark to below the benchmark. This increased capacity at the rework work center and improved the overall delivery of this part. The research also demonstrated how a cross-functional collaboration to formalize producibility lessons reduces the creation of defects, promotes systematic knowledge-sharing, and accelerates improvements similar to the stock-on surface case study. In parallel, this research evaluated how Blue Castings could improve defect documentation and tracking without causing significant additional effort for operators. The researcher’s findings highlight the limitations of handwritten weld maps and inconsistent data capture practices on effectively preventing defects. Digitization of defect tracking is recommended to enable consistent defect data collection and improved root cause and trend analyses. As data quality improves, applying classification ML models for predictive analytics can scale throughput. This work provides recommendations for Blue Castings to implement mechanisms that reduce rework, improve producibility, and increase throughput to align with Blue Origin’s goals.