| 引用本文: | 程祥辉,胡志星,张亚辉,胡小锋,刘跃雄.雾化性能预测驱动的航空发动机燃油喷嘴选配[J].哈尔滨工业大学学报,2024,56(12):105.DOI:10.11918/202309032 |
| CHENG Xianghui,HU Zhixing,ZHANG Yahui,HU Xiaofeng,LIU Yuexiong.Selective assembly for aero-engine fuel nozzle driven by atomization performance prediction[J].Journal of Harbin Institute of Technology,2024,56(12):105.DOI:10.11918/202309032 |
|
| |
|
|
| 本文已被：浏览 64次 下载 34次 |
码上扫一扫！ |
|
|
| 雾化性能预测驱动的航空发动机燃油喷嘴选配 |
|
程祥辉1,胡志星2,张亚辉3,胡小锋1,4,刘跃雄2
|
|
(1.上海交通大学 机械与动力工程学院,上海 200240; 2.中国航发南方工业有限公司,湖南 株洲 412002; 3.上海交通大学 海洋装备研究院,上海 200240; 4.上海市网络化制造与企业信息化重点实验室(上海交通大学),上海 200240)
|
|
| 摘要: |
| 为减少航空发动机燃油喷嘴装配中的反复拆卸重装,提高一次装配成功率,提出了一种基于雾化性能预判的关键零部件选配方法。首先,基于喷嘴历史装配数据构建喷嘴几何精度-雾化性能实例库；然后,考虑到样本空间大小和喷嘴几何精度波动较大、一致性差的影响,采用自适应综合过采样方法对样本空间进行扩充,同时利用改进的K-means聚类算法对连续属性离散化处理；最后,通过关联规则挖掘算法建立几何精度与雾化性能之间的关联关系,并利用规则适应度评价方法量化每条规则的准确性,基于这些关联规则集构建喷嘴雾化性能预判模型,用于指导喷嘴装配。研究结果表明,利用某双油路离心喷嘴的旋流器和副喷口的装配数据进行验证,与决策树、支持向量机和人工神经网络等方法进行比较,本方法提出的喷嘴雾化性能预判模型的预测效果最好,预测精度高达98.33%。可以对不同零件组合后的喷嘴雾化性能进行有效预判,进而减少无效装配,提高喷嘴的装配效率。 |
| 关键词: 燃油喷嘴 选配 关联规则挖掘 样本扩充 连续属性离散化 |
| DOI：10.11918/202309032 |
| 分类号:TH164 |
| 文献标识码:A |
| 基金项目:国防基础科研项目(JCKY2021110B048) |
|
| Selective assembly for aero-engine fuel nozzle driven by atomization performance prediction |
|
CHENG Xianghui1,HU Zhixing2,ZHANG Yahui3,HU Xiaofeng1,4,LIU Yuexiong2
|
|
(1.School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China; 2. Aecc South Industry Company Limited, Zhuzhou 412002, Hunan, China; 3. Institute of Marine Equipment, Shanghai Jiao Tong University, Shanghai 200240, China; 4. Shanghai Key Laboratory of Advanced Manufacturing Environment (Shanghai Jiao Tong University), Shanghai 200240, China）
|
| Abstract: |
| In order to reduce the repeated disassembly and reassembly in aero-engine fuel nozzle assembly and improve the success rate of one assembly, a key part selective assembly method based on atomization performance prediction was proposed. First, based on the historical assembly data of nozzle, the nozzle geometric precision-atomization performance case library was constructed. Next, considering the impact of large fluctuations in sample space size and nozzle geometric accuracy, as well as poor consistency, the sample space was expanded by adaptive comprehensive oversampling method, and simultaneously the continuous attribute was discretized by improved K-means clustering algorithm. Finally, the correlation between geometric accuracy and atomization performance was established by association rule mining algorithm, and the accuracy of each rule was quantified by rule fitness evaluation method. Based on these association rule sets, the nozzle atomization performance prediction model was constructed to guide nozzle assembly. The research results show that the nozzle atomization performance prediction model proposed in this paper has the best prediction effect, with a prediction accuracy of up to 98.33%, compared with methods such as decision tree, support vector machine, and artificial neural network, which can effectively predict the atomization performance of different parts combination, thus reducing invalid assembly and improving the assembly efficiency of nozzle. |
| Key words: fuel nozzle selective assembly association rules mining sample expansion continuous property discretization |
|
|
|
|
