| 摘要: |
| 磁芯损耗的准确测量是评估磁性元件性能及优化设计的关键。随着现代飞行器朝着轻便化、高频化及集成化的方向发展，磁性元件在飞行器系统中的功效直接影响系统的能效与稳定性，因此优化磁芯损耗成为提升飞行器整体性能和可靠性的核心任务。为了有效降低飞行器系统的功耗，探讨了温度、励磁波形和磁芯材料对磁芯损耗的影响，并基于这些因素构建了高精度的磁芯损耗预测模型。基于信号时频分析方法和随机森林重要性排序方法，提取了3个时域特征和3个频域特征作为励磁波形特征。通过多因素方差分析方法，揭示了温度、励磁波形及磁芯材料对磁芯损耗的独立效应及协同作用。为提高预测准确性，采用机器学习方法比较了集成学习、神经网络和高斯过程回归模型，最终选定优化后的高斯过程回归模型。经过贝叶斯优化迭代，该模型的预测精度显著提升，训练集和测试集的R2值均为0.99，平均绝对误差分别为2 227.46和2 200.64，与传统方法相比表现出了卓越的预测效果。 |
| 关键词: 飞行器系统 磁芯损耗预测 高斯过程回归 |
| DOI：10.20079/j.issn.1001-893x.250428003 |
|
| 基金项目: |
|
| Application of Gaussian Regression Prediction Core Loss Model in Aircraft Systems |
| YANG Chonggang,LI Ziwang,CHU Wen,JING Yimai,SHENG Guangze,ZHANG Hongwei,DAI Guangyao,YIN Bin,ZHANG Zhiyu |
| (1.AVIC First Aircraft Institute,Xi’an 710089,China;2.College of Marine Technology,Faculty of Information Science and Engineering,Ocean University of China,Qingdao 266100,China;3.Sanya Oceanographic Institution,Ocean University of China,Sanya 572000，China) |
| Abstract: |
| Accurate measurement of core loss is critical for evaluating the performance of magnetic components and optimizing their design.As modern aircrafts evolve towards lighter weight,higher frequency,and greater integration,the efficacy of magnetic components within aircraft systems directly impacts system energy efficiency and stability.Consequently,optimizing core loss has become a central task for enhancing overall aircraft performance and reliability.To effectively reduce the power consumption of aircraft systems,the influence of temperature,excitation waveform,and core material on core loss is investigated and a high-precision core loss prediction model is developed based on these factors.Using signal time-frequency analysis and random forest importance ranking methods,three time-domain features and three frequency-domain features are extracted as key characteristics of the excitation waveform.With multi-factor analysis of variance(ANOVA) method,the independent effects and synergistic interactions of temperature,excitation waveform,and core material on core loss are revealed.To improve prediction accuracy,machine learning methods are used to compare ensemble learning,neural networks,and Gaussian process regression(GPR) models.An optimized GPR model is ultimately selected.Following Bayesian optimization iterations,the model搒 prediction accuracy is significantly improved,and R-squared value of 0.99 is achieved on both the training and test sets,with mean absolute errors of 2 227.46 and 2 200.64,respectively.This optimized model demonstrates exceptional predictive performance compared with traditional methods. |
| Key words: aircraft system core loss prediction Gaussian process regression |
