"Granger-Causal Failure Vectors: Extracted Thermal-Mechanical Signals from the AI4I 2020 Dataset"

"This study tackles the difficulty of Root Cause Analysis (RCA) in industrial settings by introducing a synchronized causal matrix based on the AI4I 2020 Milling Machine dataset. Global industrial telemetry often suffers from statistical dilution due to steady-state operations, but this study uses an event-windowed filtering strategy to find 235 high-entropy failure instances. We made a strong set of features for causal inference by adding 30-step time lags to mechanical torque and process temperature. The study shows a significant 10.53% drop in the Residual Sum of Squares (RSS) by using a Java-based Granger Causality engine. These results show that mechanical overstrain is the main cause of thermal instability. This gives a reliable standard for predictive maintenance and automated fault diagnosis in CNC operations."