A geometrically inspired quantification approach for valve stiction using Riemannian logarithmic map. (August 2022)

• Record Type: Journal Article
• Authors: Huang, Bo; Hu, Li-Sheng
• Is Part Of: Measurement. Volume 199(2022)
• Publication Date: 2022-08
• Subjects: Valve stiction -- Riemannian manifold -- Exponential and logarithmic maps -- Quantification -- Steam turbine; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.111562
• Languages: English
• ISSNs: 0263-2241

Abstract

Abstract: Valve stiction is one of the most important failure modes of valve, which may make the control loops suffered from severe performance degradation. Therefore, valve failure detection and diagnosis have been received widely attention from researchers and industrial personals in last two decades. Up to now, flourish results on valve stiction detection and diagnosis were reported. However, to our best knowledge, how to quantify more features of the valve stiction is still remained open. Most of the reported methods on the valve stiction quantification may not be suitable to deal with slipjump of the stiction, as well as the case where the valve with stiction is used in the cascade control loops. Noticing the fact that the relationship between sampled data behaves specific shape like limit cycle with the presence of stiction, a brand new general quantification method for valve stiction based on Riemannian logarithmic map is put forward in this paper. The proposed method from the perspective of geometry expounds how to deal with the features deadband plus stickband (S-zone) and slipjump (J-zone) for single control loop, and even simple cascade loop. The experiments well demonstrated that the proposed method is feasible to accurately quantify the valve stiction using the new indicator for the deadband plus stickband, and slipjump of valve stiction. Highlights: Riemannian logarithmic map on manifold is used to quantify the valve stiction. Two methods for landmarks calculation are provided for capture routine process data. The indicators induced by logarithmic map for stiction quantification are provided. Several case studies illustrate the superiority of the proposed method.