摘要
To address the challenges of high real-time requirements in industrial fault diagnosis, this paper proposes a fast root cause diagnosis method based on Kernel Ridge Regression Granger Causality (KRRGC). Traditional Granger causality methods rely on linear regression, which fails to capture nonlinear variable relationships and performs poorly in some nonlinear scenarios. The innovative KRRGC replaced conventional linear Granger tests, in which kernel ridge regression maps raw data to a high-dimensional space via a Gaussian kernel to effectively capture nonlinear causal relationships, while ridge regression with L2 regularization avoids overfitting. To ensure valid significance testing, a multi-causality algorithm was designed to split datasets into independent subsets, and raw data was segmented into a combined matrix for repeated causal inference, reducing false causality and enhancing result consistency. The proposed method was validated by numerical simulations and the Tennessee Eastman process, which can accurately identify fault root cause and has a runtime only 50% of that of comparative models, proving to be an efficient and accurate solution for complex industrial scenarios. © 2026 IEEE.