基于BO-MLP神经网络的涡轮增压器管路疲劳寿命预测
PDF下载 (157)谢重阳,汪 涛,赵帅帅.基于BO-MLP神经网络的涡轮增压器管路疲劳寿命预测[J].宁波大学学报(理工版),2025,38(5):72-79.DOI:10.20098/j.cnki.1001-5132.2024.1126
XIE Chongyang,WANG Tao,ZHAO Shuaishuai.Fatigue life prediction of liquid-filled pipelines of turbocharger based on BO-MLP neural network[J].Journal of Ningbo University(Natural Science & Engineering Edition),2025,38(5):72-79.DOI:10.20098/j.cnki.1001-5132.2024.1126
| Title: | Fatigue life prediction of liquid-filled pipelines of turbocharger based on BO-MLP neural network |
| 作者: | 谢重阳, 汪 涛, 赵帅帅 |
| Author(s): | XIE Chongyang, WANG Tao, ZHAO Shuaishuai |
| 关键词: | 涡轮增压器充液管路; 贝叶斯优化算法; 疲劳寿命预测; MLP神经网络 |
| Keywords: | liquid-filled pipelines of turbocharger; Bayesian optimization algorithm; fatigue life prediction; MLP neural network |
| 分类号: | TH134 |
| DOI: | 10.20098/j.cnki.1001-5132.2024.1126 |
| 文献标识码: | A |
| 摘要: | 为解决汽车涡轮增压器充液管路疲劳寿命难以预测的问题，提出一种基于贝叶斯优化算法优化多层感知机神经网络（BO-MLP）的预测方法。首先利用充液管路高保真仿真模型分析筛选出几个最具关联性的设计变量作为特征输入，其次搭建BO-MLP神经网络模型架构，再采用最优拉丁超立方采样方法获取样本数据，并引入K-fold交叉验证，最后将交叉验证所得平均相对误差作为目标函数对模型进行迭代训练。结果表明：管路系统中波纹管结构参数对疲劳寿命影响显著；BO-MLP神经网络模型在小样本数据上预测效果符合预期，结果误差均在10%以内，训练集的平均误差为5.46%，测试集的平均误差为4.86%，满足工程应用需求。 |
| Abstract: | To address the problem of fatigue life prediction of liquid-filled pipelines of automobile turbocharger, a method based on Bayesian optimization algorithm to optimize multilayer perceptron neural network (BO-MLP) is proposed. Firstly, several most relevant design variables are extracted as feature inputs by using the high-fidelity simulation model of the liquid-filled pipelines. Secondly, the BO-MLP neural network model architecture is constructed, with which the optimal Latin hypercubic sampling method is employed to obtain the sample data, and the K-fold cross-validation is applied. Finally, the average relative error obtained from cross-validation is used as the objective function to train the model and the predicted results of the model are compared with the real results. The results show that: the structural parameters of the bellows in the piping system have a large influence on the fatigue life; the prediction effect of the BO-MLP neural network model on the small sample data meets the expectation, and the errors of the results are within 10%, with an average error of 5.46% on the training set and 4.86% on the test set, which meet the engineering requirements |
| 参考文献 /References: | [1] RAFIEE R, SHARIFI P. Stochastic failure analysis of composite pipes subjected to random excitation[J]. Construction and building materials, 2019, 224:950-961. [2] ZHOU Y, MU D S, HAN Y B, et al. Application of the nonlinear fatigue damage cumulative on the prediction for rail head checks initiation and wear growth[M]// Mechanical Fatigue of Metals. Chamber: Springer International Publishing, 2019:239-244. [3] ZHANG F L, YUAN Z H, ZHANG F Z, et al. The analysis and estimation of vibration fatigue for pipe fitting in aviation hydraulic system[J]. Engineering failure analysis, 2019, 105:837-855. [4] LI H F, WEI J L, LI S H, et al. Fatigue life prediction of high-speed train bearings based on the generalized linear cumulative damage theory[J]. Fatigue & fracture of engineering materials & structures, 2023, 46(6):2112- 2120. [5] 顾秀聪, 王轲, 李奕霖. 基于振动时域信号并行雨流处理的振动疲劳寿命预估[J]. 测试技术学报, 2022, 36(3): 194-198. [6] 董江, 于浩, 文敏, 等. 基于实测数据的随机振动疲劳寿命预测方法[J]. 航天器环境工程, 2022, 39(3):279- 286. [7] 宓乐峰, 孙丹, 王旭, 等. 基于正交试验的波纹管结构参数优化及疲劳寿命分析[J]. 压力容器, 2023, 40(3): 38-46. [8] SU Y, LI J, YU B, et al. Fast and accurate prediction of failure pressure of oil and gas defective pipelines using the deep learning model[J]. Reliability engineering & system safety, 2021, 216:108016. [9] ZHANG H, TIAN Z G. Failure analysis of corroded high-strength pipeline subject to hydrogen damage based on FEM and GA-BP neural network[J]. International journal of hydrogen energy, 2022, 47(7):4741-4758. [10] XIE Y G, GAO C G, WANG P Z, et al. Research on vibration fatigue damage locations of offshore oil and gas pipelines based on the GA: improved BP neural network[J]. Shock and vibration, 2023, 2023(1):2530651. [11] JING G X, LI S B, XIAO S, et al. Research on fatigue reliability assessment of engine cylinder head based on neural network[J]. International journal of fatigue, 2023, 175:107800. [12] ZHANG X X, PAN Y J, XIONG Y, et al. Deep learning-based vibration stress and fatigue-life prediction of a battery-pack system[J]. Applied energy, 2024, 357: 122481. [13] 于桂杰, 赵崇, 迟建伟, 等. 基于人工神经网络的连续油管疲劳寿命预测[J]. 中国石油大学学报(自然科学版), 2018, 42(3):131-136. [14] 朱鸿杰, 吕志刚, 邸若海, 等. 改进MD-MTD的神经网络锂电池寿命预测仿真[J]. 西安工业大学学报, 2022, 42(6):620-626. [15] 李国亮, 张佳强, 韦亮, 等. 基于BP神经网络的不同内压下连续油管疲劳寿命预测[J]. 机械设计与制造工程, 2023, 52(4):97-101. [16] 许志鹏, 刘婵, 冯红翠. 基于有限元和RBF神经网络的液压支架前连杆疲劳寿命预测[J]. 机械设计, 2024, 41(1):110-116. [17] 汪涛, 谢重阳, 白永洁. 汽车涡轮增压器充液管路振动疲劳分析与优化[J]. 液压与气动, 2024, 48(8):67-76. [18] 蒋昂波, 王维维. ReLU激活函数优化研究[J]. 传感器与微系统, 2018, 37(2):50-52. [19] HALAMKA J, BARTOŠÁK M, ŠPANIEL M. Using hybrid physics-informed neural networks to predict lifetime under multiaxial fatigue loading[J]. Engineering fracture mechanics, 2023, 289:109351. |
| 备注/Memo: | 收稿日期：2024−11−27 宁波大学学报（理工版）网址：http://journallg.nbu.edu.cn/ 基金项目：浙江省自然科学基金（LQ22E050004）；宁波市“科技创新2025”重大专项（2022Z015）；2024年度宁波市“科创甬江2035” 关键技术突破计划项目（2024Z159） 第一作者：谢重阳，博士/讲师，主要研究方向为机械系统动力学。E-mail: xiechongyang@nbu.edu.cn 宁波大学学报（理工版）网址：http://journallg.nbu.edu.cn/ |
