Generation of Turbine Maps Using a Fusion of Validated Operational Data and Streamline Curvature Method

Abstract

This paper shows how operational data in combination with a calibrated stream line curvature method (SCM) and fleet statistics can be used for turbine map generation. The operational data storage system of industrial gas turbines can be used, in case the customer agreed to provide the data, for fleet statistics, degradation behavior investigation or component map generation. The generation of updated component maps using operational data is mainly necessary for older gas turbines types since the available numerical gas turbine models often do not represent the current state of knowledge. The process for component map generation based on operational data incorporates several steps explained in detail in this paper. The first step is a full thermodynamic evaluation, including the calculation of relevant parameters like isentropic efficiencies of compressor and turbine. Furthermore, the compressor mass flow and the turbine inlet temperature are determined. This step is accompanied with the calculation of systematic and random uncertainties for all required performance parameters. The thermodynamic evaluation is coupled with a data validation system. This system incorporates signal checks, a statistics based anomaly detection, a Kalman filter based single fault isolation und a fuzzy logic based multiple fault isolation. After the evaluation and validation of the data, aging effects are eliminated. In the next step, data sets from different sites are consolidated and shifted to meet the fleet average. The first step in calibrating the SCM is the rasterizing of the operational data. The Jacobian matrix of the SCM for the loss factors to be calibrated is generated automatically at the raster points. Afterwards, the calibration is done taking into account measurement uncertainties as well as different systematic uncertainties for the loss factors in order to achieve the minimum variance result. The updated SCM, calibrated to actual engine data, is then used for component map generation.