Trend-based repair quality assessment for industrial rotating equipment

Towards artificial intelligence at scale in the chemical industry

“…Healthy unseen behaviors and previously experienced pre-failures periods can be identified, and repair activities classified on whether they were successful using targeted modeling structures. 21 Clustering algorithms in a supervised or unsupervised manner such as k-means, hierarchical clustering, Gaussian mixture models and DBSCAN have been implemented in commercial software, and are particularly useful if fault detection is to be performed during different operating states. Also in the data-driven category are ML-based classification algorithms, such as partial least squares (PLS) and partial least-squares discriminant analysis (PLS-DA), random-forests, k nearest neighbors, naïve Bayes, logistic regression, and support vector machines.…”

“…Contextualized historical data analysis allows categorization of different operating states based on their common historical multivariate data structure, thus providing additional information beyond the sole anomaly score. Healthy unseen behaviors and previously experienced pre‐failures periods can be identified, and repair activities classified on whether they were successful using targeted modeling structures 21 . Clustering algorithms in a supervised or unsupervised manner such as k‐means, hierarchical clustering, Gaussian mixture models and DBSCAN have been implemented in commercial software, and are particularly useful if fault detection is to be performed during different operating states.…”