【作者】 吴世蔚；

【导师】 张云清；

【作者基本信息】 华中科技大学 ， 机械设计及理论， 2009， 硕士

【摘要】 汽车的设计问题是涉及多个学科的复杂大系统工程问题。本文引用一种可以有效解决这类问题的优化方法——指标分解(Analytical Target Cascading,ATC)。该方法可以对汽车进行层解分析,依次解决每一层的设计问题,从而实现对整辆汽车的优化。本文在指标分解方法的基础上,提出了一种指标分解扩展方法。这种方法适用于利用ADAMS/Car建模的汽车优化问题。本文通过两个工程实例分别对指标分解方法和扩展方法进行了验证。首先利用指标分解方法对一辆SUV的平顺性和操纵稳定性进行优化设计问题。SUV的整车层采用半车模型作为平顺性问题的设计模型,采用自行车模型作为操纵稳定性设计问题的设计模型。子系统层分为了四个子系统,前、后悬架设计问题采用在ADAMS/Car中建立的多体悬架模型作为设计模型,轮胎垂直刚度设计问题采用轮胎垂直振动数学模型作为设计模型,轮胎侧偏刚度设计问题采用轮胎侧偏数学模型作为设计模型。整个优化过程是在iSIGHT与MATLAB、ADAMS/Car的联合仿真优化平台上完成的。为了顺利的执行指标分解的扩展方法,本文引入了ADAMS/Car中的概念悬架模型,并利用概念悬架模型和多体模型,结合指标转换过程,完成了该方法应用于某轿车的优化设计问题。在优化过程中,本文利用筛选试验选取了灵敏度较高的设计变量,再通过拉丁方试验设计建立了二阶响应面,最后运用模拟退火和序列二次规划的组合优化方法实现了优化。更多还原

【Abstract】 Vehicle design problem is a complex large system engineering problem which usually refers to multidisciplinary analysis. This thesis presents a useful optimal method named Analytical Target Cascading (ATC) to solve this kind of problem effectively. Using the method, targets of full vehicle design problem are cascaded and design problems in each level are settled. The thesis also presents a Special Analytical Target Cascading (SATC) method based on ATC, which can be used in the design problem of the vehicle built in ADAMS/Car.The thesis proved the feasibility of ATC and SATC through two projects. The first project uses ATC to design the Handling Performance and Ride Quality for a SUV. The vehicle-level design problem contains two analysis models, a“half-car”model and a“bicycle”model. Subsystem-level analysis models for the front and rear suspensions are multibody-dynamics models. The tire models call the tire stiffness equations. The whole optimal process is executed on the platform of iSIGHT, MATLAB and ADAMS/Car.The thesis solved the problem of a passenger car by building conceptual suspension models and multibody models in order for the application of SATC. In the optimization process, the target transition process is used to transit the targets; the screening test is used to choose important design variables; second-order polynomial RSM models are built by using Latin Hypercube design and the combination of Simulated Annealing (SA) and SQP optimization methods can prevent to trap in local optimization and improve optimization speed using RSM model.更多还原