A Novel Method for Monitoring Single Variable Systems for Fault Detection, Diagnostics and Prognostics
Volume 6, Number 5, September 2010 - Paper 7 - pp. 477-486
J. WESLEY HINES, JAMIE COBLE, and B. KEITH BAILEYDepartment of Nuclear Engineering, The University of Tennessee, Knoxville, U.S.A.
(Received on August 29, 2009, revised on March 26, 2010)
This paper introduces empirical modeling techniques for process and equipment monitoring, fault detection and diagnostics, and prognostics. The paper first provides a brief background and an overview of the theoretical foundations and presents a new method for applying these methods to systems which only have one useful measured variable. Instead of using a traditional auto-associative model to estimate the fault free parameter values, nominal operating features are inferred from the operating conditions of the system. This newly proposed system is called Stressor-based Univariate Monitoring Method (SUMM). A case study is presented for the application of this method to an aircraft generator that includes normal feature prediction over different operating conditions, actual feature measurement and residual generation, and fault detection and identification. Application of the proposed SUMM system to the simulated aircraft generator data includes fault detection and identification. The results presented here highlight application of the method to data including dynamically changing loads. A methodology for developing a corresponding prognostic model is given.
Click here to download the paper.
Please note : You will need Adobe Acrobat viewer to view the full articles.