A new approach using probabilistic outputs of support vector machine for identification of mean shift in multivariate processes Destek vektör makinesi ile elde edilen olasılık çıktılarına dayalı yeni bir istatistiksel süreç izleme yöntemi
Journal of the Faculty of Engineering and Architecture of Gazi University, cilt.39, sa.2, ss.1099-1112, 2024 (SCI-Expanded, Scopus, TRDizin)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 39 Sayı: 2
- Basım Tarihi: 2024
- Doi Numarası: 10.17341/gazimmfd.1192354
- Dergi Adı: Journal of the Faculty of Engineering and Architecture of Gazi University
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Art Source, Compendex, TR DİZİN (ULAKBİM)
- Sayfa Sayıları: ss.1099-1112
- Anahtar Kelimeler: detection of shift, identification of shift, Multivariate statistical process control, support vector machines
- Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
- Gazi Üniversitesi Adresli: Evet
Özet
In multivariate statistical process control, it is as important to identify the variable(s) that cause shift as it is to detect the significant shift in the distribution parameter vector. The important problem with most of the model-based methods developed to detect the shift in the mean parameter of the joint distribution function of multivariate processes is that they require assumptions such as the normality of process distribution and the independence of observations, and therefore do not have a flexible use. In addition to detecting the shift in the mean parameter vector, the current methods for determining the set of variable(s) causing this shift also have an important disadvantage in terms of computational burden, as well as limitations from those assumptions. This study proposes a new data-based method that allows to detect both the shift and the source of the shift using support vector machines and does not have assumption limitations. Simulation study for various process conditions has shown that the proposed method has better performance than traditional methods and can be used flexibly in different process structures.