@article{scholars12486,
             doi = {10.1016/j.chemolab.2020.104143},
            year = {2020},
            note = {cited By 14},
          volume = {206},
           title = {Detection and severity identification of control valve stiction in industrial loops using integrated partially retrained CNN-PCA frameworks},
       publisher = {Elsevier B.V.},
         journal = {Chemometrics and Intelligent Laboratory Systems},
            issn = {01697439},
          author = {Henry, Y. Y. S. and Aldrich, C. and Zabiri, H.},
        keywords = {Article; automation; classification algorithm; control valve stiction; controlled study; convolutional neural network; data analysis; dimensionality reduction; feature detection; feature extraction; friction; image analysis; industry; mechanical stress; nonlinear system; online monitoring; prediction; principal component analysis; priority journal; simulation; statistical model; time series analysis},
             url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85090286201&doi=10.1016\%2fj.chemolab.2020.104143&partnerID=40&md5=08d3a980e80d858fa43339d0c70abae8},
        abstract = {The wear and tear of control valves is a common problem encountered on process plants, owing to continuous movements of the valves. Aging of control valves leads to operational problems, such as valve stiction. Detection and severity identification of valve stiction remain an extensive research area even today, since the behaviour of these values tend to be nonlinear and is not necessarily easy to detect. In this paper, an integrated framework based on the use of convolutional neural networks (CNN) and principal component analysis (PCA) is proposed for both the detection of stiction, as well as identification of the severity of stiction. More specifically, a CNN is used to extract features from time series, while PCA acts as a dimensionality reduction tool to visualize the extracted features. Both the T2 and Q-statistics of the PCA model are applied for automated and predictive monitoring of control valve stiction. The CNN is then constructed to identify the severity of stiction as {\^a}??weak{\^a}?? or {\^a}??strong{\^a}??. The detection results on industrial benchmark loops show the ability of the proposed method to retain the generalization property and balance of false-positive and false-negative detections of the latest methods published in literature, while having the key advantage of being readily extendible to the identification of the severity of stiction. Results based on simulated data also show the promising capability of the proposed method to be used in online predictive monitoring for process plants which may be beneficial to alert the instrumentation/maintenance teams on the current and future health of their valves. {\^A}{\copyright} 2020 Elsevier B.V.}
}