WEBINAR : Efficient Inspection and realtime quality monitoring for ultra-precision polishing of Inertial Confinement Fusion Shells

On December 5, 2022, researchers at the National Ignition Facility at Lawrence Livermore National Laboratory achieved a landmark breakthrough in Inertial Confinement Fusion (ICF), producing an energy output that exceeded the laser input for the first time. The success of an ICF experiment hinges on multi-stage lapping and polishing of fuel capsule shells to nanometer-scale finish and devoid of major surface defects. This talk presents our work with LLNL on assuring surface quality of the fuel capsule shells. Rare surface defects, such as deep pits, can severely degrade ICF performance. Exhaustive inspection across multiple finishing stages to detect these defects is both cost- and time-prohibitive. Conventional scalar surface quality quantifiers fail to capture the manifestation of rare surface pits. We investigated novel inspection strategies that substantially reduce measurement burden while retaining confidence in defect-risk estimation. Here, we impose a multivariate probabilistic bound on pit distribution estimation error to determine the minimal number of surface scans needed to guarantee a specified confidence level. This enables reliable assessment of deep pit risk using approximately 5 – 6 scans (a 5- to 10-fold reduction), thereby substantially reducing the inspection time per shell at each finishing stage. These challenges also motivate the need to move beyond reactive, post-process inspection toward proactive, process monitoring methods that can detect and mitigate the process anomalies that lead to these defects. In this context, we leverage generative machine learning methods conditioned on polishing process parameters combined with shell tracking to identify deviations from expected motion patterns that may lead to surface defects. We also introduce a deep learning model that can track the evolution of pit populations across the polishing stages, capturing rare-event manifestations that scalar surface quantifiers miss. Predictive insights from these models inform possible triaging of at-risk parts and more informed process planning decisions. Taken together, these contributions illustrate how integrating inspection efficiency, process understanding, and predictive decision support can advance manufacturing quality control in settings where rare anomalies and high-consequence requirements demand more than conventional approaches.

PRESENTERS:

Satish Bukkapatnam, PhD
Regents Professor, Sugar and Mike Barnes Department Head Chair, Industrial & Systems Engineering
Texas A&M University

Shashank Galla
Graduate Research Assistant, Texas A&M Engineering Experiment Station (TEES)
Texas A&M University

Presented by SME Technical Activities

Visit https://advancedmanufacturing.org/webinars for more webinars and an interactive experience with visuals.