基于神经网络集成的家用轿车全生命周期成本估算与性能指标预测  

Cost estimation and performance index prediction for the whole life cycle of family car based on neural network integration

作　　者：陈晓川[1] 袁杰[1] 吴迪[2] 杜红彬[3] 

机构地区：[1]东华大学机械工程学院,上海201620 [2]大连理工大学计算机学院,辽宁大连116024 [3]华东理工大学自动化系,上海200237

出　　处：《机械设计》2010年第1期5-9,25,共6页Journal of Machine Design

基　　金：国家自然科学基金资助项目(60704013)

摘　　要：面向成本的设计(Design For Cost,DFC)是从设计的角度降低全生命周期成本(Life Cycle Cost,LCC)的设计方法。从DFC的角度,通过分析得到家用轿车的设计特征主要有外形尺寸、发动机功率、排量等参数,采用基于特征的神经网络集成方法,通过实例计算表明在概念设计阶段就可以估算其LCC,为降低其LCC奠定了重要基础。在计算BP神经网络权值时分别采用了Levenberg-Marquardt,LM法和遗传算法(Genetic algorithm,GA),对两种方法的计算结果进行了神经网络集成,集成后的结果更好。最后采用类似方法,对家用轿车的部分性能指标(100 km耗油量和车身质量)进行了预测。The design for cost （ DFC ） is a designing method for lowering the whole life cycle cost （LCC） from a design point of view. From the angle of design and through analysis, the designing characteristics that mainly contain parameters of outline dimensions, power of engine and delivery capacity etc. of family ear were obtained. By adopting the characteristics based neural network integration method it has been indicated by means of a living example that its life cycle cost （LCC） could then be estimated during the phase of conceptual design, and thus laid an important foundation for lowering its LCC. The LM （Levenberg- Marquardt） method and genetic algorithm （GA） have been adopted respectively while computing the weights of BP neural network. The neural network integration was carried out on the calculation results of these two kinds of algorithms, and found that the result after integration is so much the better. Finally, by adopting the similar method a prediction on partial performance indexes （oil consumption/100 kilometers and car-body mass） of family car was carried out.

关 键 词：家用轿车 面向成本的设计 全生命周期成本 神经网络集成 遗传算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]