Continuous advances of sensor technology and real-time computational capabilities allow developing industrial quality monitoring tools based on sensor signals acquired during the process itself. This yields notable benefits with respect to traditional quality control performed on the output of the process (i.e., the manufactured part). However, many discrete manufacturing processes violate the most common distributional assumptions used in Statistical Process Control (SPC). A particularly challenging violation consists of the existence of multiple in-control states (a.k.a. operating modes), which produces a stream of data from different distributions that follow one another over time. The processes that exhibit such a behaviour are referred to as a “multimode processes”. The paper discusses the use of SPC approaches for sensor signal monitoring in the presence of multiple operating modes, where nonparametric and data-adaptive learning methods are combined together. The study focuses on the use of the K-chart and the kernel density estimation (KDE) methodologies. Real industrial examples are discussed to highlight the need for nonparametric methods in industry and to demonstrate the performances of the K-chart in the presence of multimode processes.
Quality Monitoring of Multimode Processes via Signal Data
GRASSO, MARCO LUIGI;COLOSIMO, BIANCA MARIA;
2015-01-01
Abstract
Continuous advances of sensor technology and real-time computational capabilities allow developing industrial quality monitoring tools based on sensor signals acquired during the process itself. This yields notable benefits with respect to traditional quality control performed on the output of the process (i.e., the manufactured part). However, many discrete manufacturing processes violate the most common distributional assumptions used in Statistical Process Control (SPC). A particularly challenging violation consists of the existence of multiple in-control states (a.k.a. operating modes), which produces a stream of data from different distributions that follow one another over time. The processes that exhibit such a behaviour are referred to as a “multimode processes”. The paper discusses the use of SPC approaches for sensor signal monitoring in the presence of multiple operating modes, where nonparametric and data-adaptive learning methods are combined together. The study focuses on the use of the K-chart and the kernel density estimation (KDE) methodologies. Real industrial examples are discussed to highlight the need for nonparametric methods in industry and to demonstrate the performances of the K-chart in the presence of multimode processes.File | Dimensione | Formato | |
---|---|---|---|
Quality Monitoring of Multimode Processes via Signal Data.pdf
Accesso riservato
Descrizione: Paper definitivo
:
Publisher’s version
Dimensione
948.15 kB
Formato
Adobe PDF
|
948.15 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.