In model-based diagnostics, a simulation model is used to simulate the same operating conditions as the system to be diagnosed to detect and identify anomalies. For this type of analysis, the diagnostic results may be affected by multiple sources of uncertainty. The most common uncertainty to consider is measurement noise. Other sources of uncertainties may originate from the simulation model, instrumentation setup and numerical issues, such as tolerances. While these are often overlooked, they may affect the result to various extent.
In this paper, a multi-point model-based gas path analysis method is proposed and evaluated in the presence of both measurement noise and model uncertainties. The multi-point algorithm addresses the issue of the diagnostic system being underdetermined, having more health parameters than measurements available for diagnostics. It obtains a unique solution through an optimization, where the deviation in health parameter estimation for the operating conditions going into the analysis is minimised. Model uncertainties are introduced in the system by intentionally skewing the characteristics of the rotating components. The objective function is then reconfigured with a, for the gas turbine diagnostic field, novel method taking model uncertainties of the component maps into account. Through this it is possible to reduce the effect of model uncertainties on the diagnostic result. The study shows that through this approach, the uncertainties in diagnostic results are reduced by
$3.7{\rm{\% }}$ for the evaluated operating conditions.