A novel interval‐halving framework for automated identification of process trends

Abstract

Qualitative process trend representation is an useful approach to model the temporal evolution of sensor data and has been applied in areas such

A framework for on-line trend extraction and fault diagnosis

“…Recently, Dash et al (2004) proposed an interval-halving-based algorithm for automatic trend extraction. The primitive-based language (Janusz and Venkatasubramanian, 1991) is used for the representation of the qualitative trends.…”

“…) are the most natural representation of features that have been widely used for fault diagnosis (FD). In this work, an algorithm for the on-line extraction of qualitative trends is developed based on an existing technique for off-line trend extraction, namely, interval-halving (Dash et al, 2004). A framework for on-line fault diagnosis using the qualitative trends is also developed.…”

“…The primitives B, D, E and G are nonlinear primitives. One may note that the interval-halving algorithm uses a second-order polynomial to directly extract nonlinear primitives (Dash et al, 2004). It is possible to propose more specific primitives such as sharp increase or decrease, etc.…”

“…A detailed review of QTA is not intended here. For a detailed discussion on various contributions to the research on trend-analysis and/or similarity-based fault diagnosis, an interested reader is referred to the literature (Vingron and Argos, 1989;Konstantinov and Yoshida, 1992;Stephanopoulos, 1994a, 1994b;Faloutsos et al, 1994;Agrawal et al, 1995;Wong et al, 1998;Perng et al, 2000;Okabe and Masuyama, 2001;Wong et al, 2001;Colomer et al, 2002;Johannesmeyer et al, 2002;Sun et al, 2003;Sundarraman and Srinivasan, 2003;Dash et al, 2004Dash et al, , 2003Venkatasubramanian et al, 2003b;Lu et al, 2003;Hetland, 2004;Charbonnier et al, 2005;Maurya et al, 2007a). Keogh and coworkers (Fu et al, 2008;Mueen et al, 2009) have made seminal contributions in the area of similarity-based pattern-query in time-series databases and its extensions to shape analysis and motif discovery.…”

Exceptional Events Management Applied to Roller Compaction of Pharmaceutical Powders

“…There have been a number of approaches proposed to detect, diagnose, and control faults based on dynamic trend analysis and signed directed graph (SDG) based causal models [12][13][14][15][16][17][18][19][20][21][22][23]. Of these, we find the methodologies proposed by Rengaswamy et al [21], Maurya et al [16][17][18][19], and Dash et al [13] to be more appropriate for our needs and hence we adapt them suitably to address our problem. Thus, this study demonstrates the application of SDG and qualitative trend analysis (QTA) to detect and diagnose exceptional events related to the operation of roller compaction (a common step in most tablet manufacturing processes), as well as suggest mitigation strategies to alleviate and prevent the progression of their effects.…”