Kouadri, AbdelmalekChaib, AhmedNamoune, Abdallah2024-07-122024-07-122009Kauadri, A., Chaib, A. ve Namoune A. (2009). Fault detection in a complex system: a new statistical-based approach. Maltepe Üniversitesi, İnsan ve Toplum Bilimleri Fakültesi. s. 61.9.78605E+12https://www.maltepe.edu.tr/Content/Media/CkEditor/03012019014112056-AbstractBookICMS2009Istanbul.pdf#page=76https://hdl.handle.net/20.500.12415/2534Fault detection in stochastic dynamical systems is usually done by the generation of residuals directly reflecting the magnitude of the faults. This faults’ indicator is used to evaluate deviations created from normal operating conditions and measurements of the system. This test is almost always very difficult to implement in the multi-faults case. In this paper, we propose a new detection index based on descriptive statistics. In general, statistical data can be described as a list of subjects and their associated data. We have chosen the statistical method, namely a measure of statistical variability, which shows how the data differs. In physical systems, variability may result only from random measurement errors: instrument measurements are often not perfectly precise. One way is to assume that the quantity being measured is constant and that the variation between measurements is caused by observational errors. The coefficient of variation (CV) is a good measurement of the dispersion degree of a given data randomness. It is defined as the ratio of the standard deviation to the mean. Therefore, to assess the detection in the multi-faults case, it is preferrable to use the CV as a fault detection index. To estimate the average CV’s for each signal and its confidential intervals (CI), we have carried out a number of numerical simulation experiments on a Three Tank System DTS-200. The CV and CI are calculated from twenty independent runs, in the same operating conditions, where a single run consists of 10240 samples for each signal.enCC0 1.0 Universalinfo:eu-repo/semantics/openAccessConference Object6261