【作者】 刘亮；

【导师】 孙伟； 孙清超；

【作者基本信息】 大连理工大学 ， 机械工程（专业学位）， 2021， 硕士

【摘要】 航空发动机转子作为发动机的关键部件之一,其装配后的精度程度对于实际工作状态有着重要的影响。在装配过程中,由于零件的制造误差的不确定性以及受力后零件发生变形,转子的装配精度很难得到保证,另外还会出现“曲轴型”、“弯弓型”转子装配的现象,导致航空发动机转子装配的一次性成功率低,需要通过人工试错法、修配法多次装配。本文提出了一种基于机器学习的航发转子装配精度预测与优化技术,融合人工智能算法和优化算法,实现预测航空发动机转子的装配精度和提高装配效率的目的。针对航空发动机转子零件制造误差的随机性、装配过程中变形的不可控性,提出利用有限元和人工智能算法实现高效准确的做出装配精度的预测;针对装配成功率低这一问题,提出一种基于新的目标函数的优化算法,可以避免装配过程中出现“曲轴型”转子的现象。本文的主要研究如下:(1)首先介绍了航发转子仿真技术的研究,包括形貌提取与重构技术,仿真的前处理、后处理,并对仿真结果进行了分析,最后采用了自动化的方法获得多组数据用来构建机器学习数据集。(2)构建了两种航发转子精度预测模型:基于单体等效模型的刚性堆叠模型和基于机器学习的弹性预测模型。首先使用刚性模型进行了预测,并与获得的仿真结果进行了对比分析,然后训练了多组机器学习模型进行预测,并对不同的机器学习模型进行了对比。(3)介绍了航发转子的安装相位优化算法研究,通过对多种遗传算法模板进行实验,选出了一种最优的算法模板;并提出了多种用来消除“曲轴型”转子的目标函数,从中选出一种效果最好的;另外对算法做了缓存优化,使其计算时间极快仅需10s左右,最后对结果进行可视化方便观察效果。(4)介绍了智能化装配软件的设计、功能及其简单的使用方法,另外对软件中的一些小细节进行了阐述,展示了该软件具有简易良好的操作性和人性化的用户操作体验。更多还原

【Abstract】 As one of the key parts of the engine,the aero-engine rotor has an important impact on the actual working state.During the assembly process,due to the uncertainty of the manufacturing errors of the parts and the deformation of the parts after the force is applied,the assembly accuracy of the rotor is difficult to be guaranteed.In addition,the phenomenon of "crankshaft" and "bend-shaped" rotor assembly may also occur.As a result,the one-time success rate of the aero engine rotor assembly is low,and it needs to be assembled repeatedly through manual trial-and-error method and repair method.In this paper,an aircraft engine rotor assembly accuracy prediction and optimization technology based on machine learning is proposed,which combines artificial intelligence algorithm and optimization algorithm to predict the assembly accuracy of aero-engine rotor and improve the assembly efficiency.Aiming at the randomness of manufacturing error of aero-engine rotor parts and the uncontrollability of deformation in assembly process,the finite element and artificial intelligence algorithm are used to achieve efficient and accurate prediction of assembly accuracy.Aiming at the problem of low assembly success rate,an optimization algorithm based on new objective function is proposed,which can avoid the phenomenon of “crankshaft” rotor in the assembly process.The main research of this paper is as follows :(1)Firstly,the research of aviation rotor simulation technology is introduced,including shape extraction and reconstruction technology,simulation pre-processing and post-processing,and the simulation results are analyzed.Finally,an automated method is used to obtain multiple sets of data to build the machine learning data set.(2)Two aircraft rotor accuracy prediction models are constructed: a rigid stack model based on a single equivalent model and an elastic prediction model based on machine learning.First,the rigid model was used to make predictions,and compared with the simulation results,then multiple sets of machine learning models were trained to make predictions,and different machine learning models were compared.(3)Introduced the research on the installation phase optimization algorithm of the aeroengine rotor.Through experiments on a variety of genetic algorithm templates,an optimal algorithm template was selected;and a variety of methods used to eliminate the "crankshaft" rotor were proposed.The objective function is to select the best effect;in addition,the algorithm is cached and optimized,so that the calculation time is extremely fast,only about 10 s,and finally the results are visualized to facilitate observation of the effect.(4)Introduced the design,function and simple use method of intelligent assembly software.In addition,it explained some small details in the software,showing that the software has simple,good operability and user-friendly user experience.更多还原