Optimization configuration of gas path sensors using a hybrid method based on tabu search artificial bee colony and improved genetic algorithm in turbofan engine

The gas turbine engine is usually working in extreme engine conditions, especially the components along the gas path are bearing the harsh environments, such as high temperature, high pressure, strong vibration and so on, which are prone to failures. The need for security and safety makes it necessary to develop an accurate and efficient monitoring and diagnostic scheme for the gas turbine components. Sensors along the gas path provide the health status information that plays an important role in the monitoring and diagnostic scheme. Owing to the harsh installation environments, the gas path sensors are so limited that it is critical to choose the optimal sensors combination to provide rich and effective status information for health monitoring and fault diagnosis. In this study, the sensitivity of gas path parameters under various gas path component faults is firstly analyzed in a turbofan engine, which verifies that diagnosis results depend sensitively on the types and orders of sensors chosen along gas path. Therefore, an optimization scheme of gas path sensors for fault diagnosis is designed and implemented using three different meta-heuristic global optimization algorithms. The genetic algorithm (GA) is firstly exploited and improved to adapt to the real problem of sensors optimization, which has been shown to be with strong convergence ability but low computation efficiency. The artificial bee colony algorithm combining tabu search (TSABC) has the characteristics of easily realization, less parameters tuning and fast searching speed, which will be seen as a more effective meta-heuristic and proved to achieve good performance of sensors optimization and fault diagnosis, however, this algorithm still exists the defect of slow convergence. Therefore, a hybrid algorithm based on TSABC and IGA (TSABC-IGA) has been then presented and compared, in which the TSABC with fast searching speed is used to obtain the initial optimal population and the IGA with strong convergence ability is exploited to choose the ultimate optimal sensors combination. The experimental studies show that the hybrid algorithm is capable of producing better or at least promising results compared to the mentioned optimization techniques for all of the fault scenarios.