Skip to Main content Skip to Navigation
Conference papers

A causal Gaussian mixture model decomposition for root cause identification

Abstract : Multivariate statistical process monitoring methods usually assume the Gaussianityof data. However, in practice, data are multi-modal. Therefore, it’s not always reasonable andenough to use methods that only deal with the data overall covariance matrix. As the lattermay wrap less information compared to the data distribution. Also, such prior assumption isprejudicial to the estimation of the data’ structure and the causal direction of variables. Aninteresting challenge would then be the development of relevant metrics to monitor variablesand address their causal nature in the context of the non-Gaussianity of the data. Therefore,adequate parametric tests are required to ensure an acceptable and adjustable compromisebetween false positives and false negatives. In this paper, a new statistical approach is introducedto root cause and fault path propagation analysis. The obtained results demonstrate that theproposed method performs better than the existing methods.
Document type :
Conference papers
Complete list of metadata
Contributor : Mohamed Amine Atoui <>
Submitted on : Sunday, March 21, 2021 - 1:24:35 PM
Last modification on : Sunday, March 21, 2021 - 2:45:29 PM


  • HAL Id : hal-03175805, version 1



Mohamed Amine Atoui, Vincent Cocquempot. A causal Gaussian mixture model decomposition for root cause identification. IFAC INCOM 2021, Jun 2021, Online, Hungary. ⟨hal-03175805⟩



Record views