Turbofan engines are widely used in modern aircraft. Low-pressure turbines are the heaviest components of turbofan engines, and reduction of their weights is very effective in improving the specific fuel consumption and overall efficiency of these engines. One of the methods of decreasing the engine weight is to decrease the number of blades which is accompanied by an increase of the blade loading. For this purpose, high-lift airfoils can be used. As the occurrence of flow separation is very probable in high-lift blades, the recognition of the location and size of the separation bubble is important to assess the energy loss of flow. In this research, T106D-EIZ high-lift cascade is simulated by two-dimensional Unsteady Reynolds-Averaged Navier-Stokes (URANS) equations with Shear Stress Transport (SST) turbulence model and γ-Re_θ transition model in two Reynolds numbers 200,000 and 60,000 at a constant isentropic exit Mach number of 0.4, which represent a typical flow condition in low-pressure turbine. The results show that when Reynolds number is high, the separation bubble remains small on the suction side and the separated shear layer returns to the blade surface, and the energy loss of flow decreases. On the other hand, at a low Reynolds number, the separation bubble grows and energy loss increases. Separation bubble is not directly detectable in an evaluation of pressure distribution. However, proper orthogonal decomposition of the pressure field provides the capability to identify flow structures including vortex stretching, the onset of flow separation, and flow reattachment. When the separation bubble is long, large vortical structures are formed on the suction surface. Release of these large vortices can increase the profile loss by more than 50 percent.

The load type imposed on the structures is one of the important issues of the modal identification Experimental methods. Generally the loads applied to a structure for dynamic testing are divided into two categories: artificial stimulation and ambient loads. Applying artificial loads to large structures such as bridges and tall buildings is difficult, costly and in some cases impossible. For this reason, modal identification of such structures is generally done by ambient vibration tests. However this experimental methods, also include problems such as large noise amplitude relative to the measured responses that this causes errors in the results and in some cases leads to unrealistic modes. As a solution, modal information can be calculated from several different methods and compared with each other to ensure the accuracy of the results. In this paper, a new scheme for natural frequencies extraction of structures from their ambient vibration is presented. For this purpose, the combination of two mathematical techniques of random decrement (RD) and proper orthogonal decomposition (POD) methods were used. The reason for using these two methods, is their ability to reduce the noise effects. In other words, combining of these two methods can lead to a very powerful tool for extracting structural frequencies from its ambient vibration under high amplitude noise conditions. The proposed algorithm consists of three steps: In the first step, after measuring the acceleration response of the structure at the appropriate points, the effects of random vibration are eliminated from the response by RD method and only dynamic properties of the structure remain in the acceleration records. Secondly, the acceleration records are separated into several structural modes using the proper orthogonal decomposition technique and finally, at the last step, the proceeded responses are transformed by the fast Fourier transform into the frequency domain to extract the natural frequencies of the structure. The strength of the proposed method is its robustness to the use of very high amplitude noise data, which is one of the challenges in the ambient vibration experiments. The accuracy of the proposed algorithm was evaluated by numerical modeling and experimental study. To investigate the efficiency of the new method, the numerical and experimental results were compared with the frequencies obtained from commonly modal identification methods such as extended frequency domain decomposition (EFDD) and stochastic subspace identification (SSI). A very good agreement was observed between the results of methods. Furthermore, Studying the effect of noise on the new algorithm results shows that increasing the ratio of noise to acceleration amplitude up to 250, did not affect the results precision and the main frequencies of the structure can be obtained with good accuracy. In this study, the effect of the number of sensors used in the ambient vibration test also was investigated on the accuracy of the new algorithm results. It was concluded that the minimum number of sensors (even one number) and repetition of the experiment can be used to extract structural frequencies from its ambient vibration with high accuracy. The results of this study showed that the new method can be used as a suitable tool to determine the natural frequencies of structures from its ambient vibration under severe noise conditions and to control the results obtained from other methods.