Double Objective Optimation Assembly Method Of Coaxiality And Initial Unbalance For Aero-Engine Rotors

Aeroengine is one of the key research fields of “Made in China 2025”.The core engin is assembled by multi-stage rotors.Coaxiality and initial unbalance out of specification cause engine vibration,leading to blade rubbing,significantly affecting engine performance.Therefore,it is urgently needed to develop the optimal assembly method for multi-stage rotors to improve the assembly quality of the aeroengine.“Single inclined surface”and “saddle surface” are two surface error models of the contact surface for aero-engine rotors.In this paper,a double objective optimization assembly method of coaxiality and initial unbalance for aeroengine rotors is proposed to optimize the coaxiality and the initial unbalance of multi-stage rotors based on the two surface error models.The propagation principle of geometric eccentricity and mass eccentricity are researched and the vector projection model of coaxiality and initial unbalance under the “single inclined plane” model are established.On this basis,the error sources are analyzed according to the assembly process and the prediction methods of coaxiality and initial unbalance under the “saddle plane” model are established.Based on the prediction methods,a double objective optimization model is established to improve the assembly quality of the multi-stage rotors by optimizing the assembly angles of the rotors.The main research contents include:Firstly,the vector projection models of coaxiality and initial unbalance for multi-stage rotors assembly are established to solve the problem of coaxiality and initial unbalance out of specification under the "single inclined plane" model.The propagation processes of geometric eccentricity and mass eccentricity in assembly are analyzed.And the distribution functions of geometric eccentricity and mass eccentricity of rotors are obtained.The prediction models of coaxiality and unbalance for multi-stage rotors assembly are established.The process of errors interaction is studied,and the principle of vector cancellation of unbalance for multi-stage rotors assembly is expounded.Secondly,the coaxiality and initial unbalance prediction methods based on the neural network are proposed to solve the problem that the vector projection models do not apply due to the change of the contact form of the rotor connection surface under “saddle surface model”.Based on the structural characteristics and the stack-assembly process of multi-stage rotors for aeroengine,the error sources of coaxiality and unbalance are identified.And the neural networks for coaxiality and unbalance prediction are established.According to the relationship between the cost function of cross validation and the network performance,the network hyperparameters are optimized.The prediction for coaxiality and initial unbalance of final assembly is realized.Thirdly,a double objective optimization assembly method for aeroengine rotors is proposed to achieve the synchronous optimization of coaxiality and initial unbalance.A double objective optimization model of coaxiality and initial unbalance is established based on the theory of multi-objective optimization.Considering the actual assembly constraints of the bolt holes and the limit of the coaxiality and the initial unbalance,the model was solved by the genetic algorithm.The non-inferior assembly angles group are obtained to minimize the coaxiality and the initial unbalance synchronously.Lastly,the vector projection models and prediction methods based on the neural network for multi-stage rotors of coaxiality and initial unbalance are verified,respectively.Besides,the effectiveness of the double objective optimization assembly method is verified.Experiments show that the prediction error of the coaxiality and the initial unbalance under the "single inclined plane" model based on the vector projection models are 5.1 ?m,196 g·mm,respectively;the maximum prediction error of the coaxiality and the initial unbalance under the "saddle surface plane" model based on the neural network are 6.9 ?m,695 g·mm,respectively,meeting the assembly requirements.The doulbe objective optimization assembly method can give qualified assembly angles according to the actual requirements.Therefore,the double objective optimization assembly method of coaxiality and initial unbalance can not only improve the accuracy of rotor assembly,but also optimize the quality of rotor assembly.The double objective assembly method can provide technical guidance for error allocation in the process of rotor design and assembly phase optimization in the actual assembly process.