摘要
为实现汽车碰撞试验假人响应曲线数据集的类别辨识,研究了面向智能优化算法的问题转换与构造方法。针对假人曲线数据的特征处理与分类过程,提出了一种基于社会蜘蛛粒子群优化算法(SSI-PSO)的碰撞试验多变量时序数据特征选择与分类方法;利用汽车碰撞试验采集的假人曲线数据,测试和验证了该方法。结果表明:本文方法可获得面向假人曲线数据分类的最佳特征组合方式与较小规模的神经网络结构;该方法的假人曲线分类模型性能提升17.5%、分类精度达到96.5%。因而,实现了对碰撞试验假人响应曲线标注信息的有效分类。
This paper investigated an optimization problem transformation and construction method for heuristic optimization algorithm to realize the category identification of dummy curve dataset from automobile crash test.A method of feature selection and classification was proposed for multi-variable time-series data in crash test based on a social spider inspired particle swarm optimization(SSI-PSO)for the feature processing and for the classification process of dummy curve data.The proposed method was tested and validated by using the dummy curve data collected from automobile crash test.The result shows that the optimal feature combination and the small-scale neural network for dummy curve classification are obtained by the proposed method.The performance of dummy curve classification model improves by 17.5%and classification accuracy reaches 96.5%based on the proposed method.Therefore,the labeling information of dummy response curve from crash test is classified effectively.
作者
李晗
刘钊
朱平
LI Han;LIU Zhao;ZHU Ping(School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China;School of Design,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《汽车安全与节能学报》
CAS
CSCD
北大核心
2022年第2期259-268,共10页
Journal of Automotive Safety and Energy
基金
国家自然科学基金(U1864211,11772191)
上海市自然科学基金(21ZR1431500)。
关键词
汽车碰撞
安全数据集
多变量时序数据
社会蜘蛛粒子群优化(SSI-PSO)算法
特征工程
监督学习
启发式优化算法
automobile crash
safety data
multi-variable time series data
social spider inspired particle swarm optimization(SSI-PSO)algorithm
feature engineering
supervised learning
heuristic optimization algorithm
