Using a set of GM(1, 1) models to predict values of diagnostic symptoms. (February 2015)

• Record Type: Journal Article
• Authors: Tabaszewski, Maciej; Cempel, Czeslaw
• Is Part Of: Mechanical systems and signal processing. Volume 52/53(2015)
• Page Start: 416
• Page End: 425
• Publication Date: 2015-02
• Subjects: Condition monitoring -- Time series -- Model selection -- Grey system theory; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2014.08.013
• Languages: English
• ISSNs: 0888-3270

Abstract

Abstract: The main purpose of this study is to develop a methodology of predicting values of vibration symptoms of fan mills in a combined heat and power (CHP) plant. The study was based on grey system theory and GM(1, 1) prognostic models with different window sizes for estimating model parameters. Such models have a number of features that are desirable from the point of view of data characteristics collected by the diagnostic system. When using moving window, GM(1, 1) models tend to be adaptive. However, selecting an inappropriate window size can result in excessive forecast errors. The present study proposes three possible methods that can be used in automated diagnostic systems to counteract the excessive increase in the forecast error. A comparative analysis of their performance was conducted using data from fan mills in order to select the method which minimises the forecast error. Highlights: We model the time series of vibration symptoms using a set of rolling models GM(1, 1). We compare proposed methods for reducing forecast error of vibration symptoms. We proposed a method for dynamic selection window length for fan mills vibration symptoms. Finally, we received a relatively (for industrial environments) small forecast error.