Modal Analysis for Quality Control of Pre-baked Carbon Anodes

The presence of defects in baked carbon anodes can significantly affect aluminum production performance detrimentally, and cause increased emissions and cost through increased specific energy and carbon consumption. This study explores the use of principal component analysis coupled with linear discriminants to classify full-scale industrial anodes based on vibration frequency response data related to mechanical properties. Large structural changes, such as the anode’s production stage, and smaller structural changes, such as the presence of external defects, are evaluated together with an assessment of the measurement’s repeatability. The method proposed in this work tests the entire anode block, achieves superior performance compared to previously published results using modal analysis both in classification accuracy and repeatability, and is also potentially faster. The ultimate goal is to use modal analysis as an online anode quality control system.