Gas turbine fault identification has been used worldwide in many aero and land engines. Model based techniques have improved isolation of faults in components and stages’ fault trend monitoring. In this paper a powerful nonlinear fault identification system is developed in order to predict the location and trend of faults in two major components: compressor and turbine. For this purpose Siemens V94.2 gas turbine engine is modeled one dimensionally. The compressor is simulated using stage stacking technique, while a stage by stage blade cooling model has been used in simulation of the turbine. New fault model has been used for turbine, in which a degradation distribution has been considered for turbine stages’ performance. In order to validate the identification system with a real case, a combined fault model (a combination of existing faults models) for compressor is used. Also the first stage of the turbine is degraded alone while keeping the other stages healthy. The target was to identify the faulty stages not faulty components. The imposed faults are one of the most common faults in a gas turbine engine and the problem is one of the most difficult cases. Results show that the fault diagnostic system could isolate faults between compressor and turbine. It also predicts the location of faulty stages of each component. The most interesting result is that the fault is predicted only in the first stage (faulty stage) of the turbine while other stages are identified as healthy. Also combined fault of compressor is well identified. However, the magnitude of degradation could not be well predicted but, using more detailed models as well as better data from gas turbine exhaust temperature, will enhance diagnostic results.
Gas turbine engine dynamic model based on variable-memory LSTM architecture
