Semi-supervised high-uncertainty deep canonical variate analysis for fault diagnosis in blast furnace ironmaking

Yuelin Yang, Chunjie Yang, Xiongzhuo Zhu, Hanwen Zhang, Haifeng Zhang, Zhiqi Su, Siwei Lou

Abstract: Highlights•A semi-supervised fault diagnosis method called SHDCVA is proposed to tackle non-linearity, dynamics, noise, and limited labeled data issue in BFIP.•A high-uncertainty deep canonical variate representation is developed to capture complex nonlinear dynamic characteristics in noisy environments.•A robust semi-supervised framework leverages limited labeled and abundant unlabeled data for better fault diagnosis model accuracy and robustness.•Real-world validation confirms that SHDCVA outperforms existing fault diagnosis methods, demonstrating its efficacy and reliability.