Dynamic Response Analysis Of Aero-engine Based On Substructure

During the development of aero-engines,the vibration of the whole machine is a crucial issue in the dynamic analysis of aero-engines,and has always been a key factor which restrict the safety and reliability of aero-engines.However,the calculation efficiency of the model established based on the traditional finite element method is not high,and various uncertain factors in the modeling process leads to large errors in the simulation results and test results for the finite element model,resulting in the modeling and dynamic analysis of aero-engines becomes a time-consuming and labor-intensive business.Therefore,it is necessary to propose a method to improve the modeling efficiency under the premise of taking into account the modeling accuracy.For a huge multi-degree-of-freedom system like an aero-engine,the substructure method can divide the whole structure,reduce the number of degrees of freedom,and reduce the solution scale.The main research work of this paper is as follows:Aiming at the low efficiency of dynamic characteristic analysis of large and complex structural models established by traditional finite element method,a multi-level substructure modeling method was proposed.Taking the simulated rotor of an aero-engine gas generator as the research object,a multi-level substructure modeling and analysis method that can be applied to repetitive physical geometric structures is proposed based on the fixed-interface modal synthesis method.A multi-level substructure model of the rotor is established,and the effects of different substructure grading methods,the selection of residual structures and the number of external nodes on the calculation results are explored.The results show that the method greatly improves the computational efficiency of the model under the premise of ensuring the calculation accuracy,and the classification of different substructures and the selection of the residual structure will not affect the method.Under the premise of ensuring the accuracy,selecting 50%external nodes can further improve computational efficiency.For the complex rotor system of aero-engine,the high-precision dynamic modeling analysis and model reduction method of the rotor are studied.Combined with ANSYS and MATLAB,a high-precision reduced-order model of the rotor based on virtual nodes was established.On this basis,the unbalanced response of the rotor was analyzed and compared with the response results of the full model.At the same time,the influence of different polycondensation virtual nodes on the calculation results is explored.The results show that the method can not only guarantee the calculation accuracy,but also can greatly shorten the calculation time and improve the calculation efficiency.Aiming at the shortcomings of structural response prediction,such as time-consuming,complex analysis and low accuracy,a response prediction method based on deep learning is proposed.sample,to achieve the rapid response prediction of the rotor.The network is trained with the displacement of the i-th time point of the rotor 2# support center point as the input and the displacement of the i+1-th time point as the output,the network architecture and parameters are continuously adjusted,finally the remaining 30% data is used as a test set to verify the effectiveness and applicability of the network.Aiming at the big difference between the model simulation results and the actual structure experimental results and the low optimization efficiency of large-scale complex structure design,the research on fast finite element model modification of the simulated rotor of gas generator was carried out.Modal experiments are carried out on the simulated rotor of the gas generator,and the optimization objective is constructed by using the difference between the experimental frequency and the simulated frequency,so as to correct the rotor.At the same time,the model correction based on the substructure method is adopted to speed up the iteration time.The results show that the relative error between the first three-order bending frequencies of the model corrected by the finite element model correction method based on the substructure method and the frequencies obtained from the experiment is within 2%,and the time consumption is reduced by 55.24% compared with the traditional overall model correction method,greatly improving the correction efficiency.