Assessments of performance deterioration and its prediction are of major importance to the gas turbines operation and maintenance. The task called Gas Path Analysis (GPA) deals mainly with diagnostics of faults in the engine flow path (the efficiency and capacity changes of the modules), malfunctions in the engine control loops, changes in power off-takes and air-escapes. This approach is based on comparison of the measured thermodynamic data (temperatures, pressures, shaft speeds etc.) at present with the accepted reference data with a subsequent application of different mathematical methods meant for explanation of the data differences. These methods make use of gas turbines mathematical models, often linear in the form of so called fault matrices, representing the effect of different faults on measured parameters. The use of standard non-linear steady state performance models for investigation of in-service problems is essentially restricted because of the lack and inaccuracy of measured information. Application of such models requires reliable and sufficient information about measured performance parameters of the considered engine to carry out a good identification of the engine internal state. In practice there is usually no such information. Even the heavily instrumented development engines data list is often insufficient for a proper identification. In mathematical terms the gas turbine performance diagnostics is a badly-posed problem, with a high degree of uncertainty. So the low and insufficiently accurate instrumentation of production engines has resulted in a need for creating special diagnostic tools exhibiting a special approach to the task. In contrast to the other linear methods dealing with the GPA problem (known to the author), the proposed approach is based on the probability analysis of fault matrices and not on the solution of systems of linear algebraic equations. That makes it principally different from the other GPA methods using fault matrics. The new method has shown high level of working capacity. The application of the proposed diagnostic approach shown in this paper embraces main problems that may occur in operation. It is the routine estimate of performance deterioration shown by an example of analysis of the long-run test bed data, then it is the sifting of sensor faults and control loop settings changes and at last the diagnostic results themselves. The data used is the real test bed data of two-shaft engines with a very low by-pass ratio.

This content is only available via PDF.
You do not currently have access to this content.