An independent component analysis and mutual information based non-Gaussian pattern matching method for fault detection and diagnosis of complex cryogenic air separation process

The conventional principal component analysis (PCA) based pattern matching methods have been applied to dynamic process monitoring. However, they do not take into account the non-Gaussian features in industrial processes and are also more focused on fault detection instead of fault diagnosis. In this paper, an independent component analysis and mutual information based non-Gaussian pattern matching approach is developed for fault detection and diagnosis of complex chemical processes. The presented approach is applied to a simulated cryogenic air separation process and the application study demonstrates that the developed non-Gaussian pattern matching method can effectively monitor the complex air separation process with strong capability of fault detection and diagnosis.