【作者】 郭亚娟；

【导师】 孟光；

【作者基本信息】 上海交通大学 ， 机械设计及理论， 2010， 博士

【摘要】 配管是空调器的重要部件,也是空调器中最薄弱、最容易损坏的结构件之一,一直以来配管都是空调器结构开发过程中的一个难点。仿真设计是空调设计方法的发展趋势,对配管系统进行仿真设计,可以减少不必要的物理样机的制作,降低设计成本,缩短设计周期,加快新产品推向市场的速度,从而增强企业产品的竞争能力。论文受日本Daikin空调“空调配管系统的动力特性试验分析与有限元模拟”和“空调配管系统的减振设计研究”两期项目资助。提出在配管表面局部环绕粘弹性阻尼层吸收配管的振动;对配管系统仿真研究的关键技术——配管系统的建模和阻尼层位置及形状优化进行了研究,并通过了大量的试验和有限元对比验证了模型的正确性。论文是开发空调配管系统虚拟设计、性能评估以及结构优化平台的重要组成部分,该研究从根本上改变了传统设计思路,使得现行的设计方案在设计进行的同时就能对空调压缩机匹配的振动性能进行仿真评估。以环绕粘弹性阻尼层的空调配管为研究对象,开展了粘弹性复合结构动力特性计算方法和基于代理模型的复合结构动力特性优化等研究工作。论文从一类沥青型粘弹性材料力学特性研究入手,开展了材料力学特性的试验研究,粘弹性本构方程参数优化以及模型参数验证等工作;进而在已知材料力学参数的基础上,对周向环绕阻尼层的U型管和多种型号的配管系统进行了动力特性试验研究和有限元建模工作;最后依据试验设计方案,对配管系统建立了三种近似模型,并以结构质量最小和结构阻尼比最大为优化目标,利用遗传算法开展了基于近似模型的多目标结构优化研究。论文首先利用机械动态热分析仪对试验中所用的沥青阻尼片进行材料力学性能分析,测试结果表明,这种沥青阻尼片具有粘弹性材料的特点,材料的弹性模量,损耗模量和损耗因子随着频率的增加呈递增趋势,但增长速度逐渐下降;当温度增加时,弹性模量,损耗模量和损耗因子都迅速下降,且损耗因子β不存在明显的峰值。基于材料的测试数据,提出采用GA-BFGS混合算法对多目标多参数的粘弹性模型进行参数优化。该算法对初始参数取值无限制,收敛速度快,具有全局收敛性,适合多参数多峰值优化问题。文章对常用的STD模型、GHM模型和ADF模型进行参数优化,综合分析认为,ADF模型更适合作为此类沥青阻尼片的本构关系模型。为验证优化所得到的ADF模型参数的有效性,采用粘贴自由层阻尼片的悬臂梁进行振动测试和有限元建模研究。悬臂梁试验与有限元计算表明,采用ADF模型的计算结果与试验结果吻合较好,ADF模型可以较精确的描述该类阻尼材料的本构关系。由于粘弹性材料力学特性随频率的变化而变化,动力特性分析时难以考虑结构的频变特性,为此,提出采用频率迭代-模态应变能-振型叠加相结合的方法计算粘弹性结构固有频率、阻尼比和动力响应,即采用迭代法计算复合结构的固有频率;进而基于模态应变能法,求得模态阻尼比;最后在模态特性分析和阻尼特性分析的基础上,采用模态叠加方法计算结构的位移响应。通过对3种不同型号的U型管以及12种不同型号尺寸的空调配管系统的试验和有限元分析,研究了各种空调配管的固有特性,同时验证了所提出的有限元计算方法的适用性和可靠性。针对结构动力特性计算过程复杂耗时的特点,建立了空调配管系统的二次多项式响应面、高次多项式响应面、Kriging最优内插法和BP神经网络等四种近似模型,编制了相应的通用程序,并首次采用正交试验设计和均匀试验设计相结合的试验设计模式训练近似模型。训练结果表明后三种近似模型都有较高的近似精度,泛化能力也很强,其中,BP神经网络模型和Kriging模型的近似结果稍好于高次多项式响应面模型的近似结果,可以在优化计算中代替复杂耗时的有限元计算。在配管系统的结构优化设计中,结构阻尼比是评价结构减振性能的一个重要指标,在其它性能参数都相同的情况下,阻尼比越大则说明结构的减振性能越好。同时配管周向环绕阻尼层后大大增加了结构的重量,也提高了设计成本。以环绕粘弹性阻尼层的空调配管系统质量最小,结构阻尼比最大为优化目标,采用改进的多目标遗传算法NSGA-Ⅱ对高次多项式响应面、Kriging最优内插法和BP神经网络近似模型进行优化分析。从结构阻尼比和结构质量之间的Pareto曲线中可以看出,随着阻尼层用量增加,结构的阻尼比不断增大,当宽度和大于58mm时,阻尼比增加速度非常小,综合考虑阻尼比和阻尼层用量,阻尼层质量应选取在140 g~160 g之间,此时结构质量增加为13.4%~29.5%,结构一阶阻尼比在7.5%~10.8%。依据最优方案的试验验证,环绕阻尼层后,配管系统的振动位移衰减可达19.8dB,加速度衰减达26dB。本文的研究工作,是空调配管仿真设计的一个重要组成部分,为配管系统的仿真建立了精确的有限元模型和简单可行的有限元计算方案,同时也为空调配管的减振提供了一种新的方法和切实可行的优化计算方案,并对类似复合结构的动力特性试验、有限元建模以及复杂结构的多目标优化分析均具有一定的参考价值。更多还原

【Abstract】 Pipe system is an important part of air conditioner, and it is also the insubstantial and fragile unit for a long time, the improvement of pipe system is a difficult design process of air conditioner. Visiul design is becoming the trend of the design and developmeng of air conditioner. Simulation design for pipe system can replace the examination of model macnine, decrease the cost and shorten the design time. Meantime, the productive tempo will be cuttingdown greatly, which will strengthen the competitive power for company.This dissertation was supported by Daikin Co. A new vibration attenuation method was suggested in this paper, with partly circling damping layer on the pipe to absorb the vibration, and strengthen the reliability. Finite element (FE) modeling and damping optimization of pipe system were researched in this paper. Examination and FE verfivation were carried out in gross for pipe system with different type and size. It is the key technique for the simulation of air conditioner system. This research will change the traditional design concept, and the design process and performance evaluation can carry out in the same time.The main research work involves calculation method of dynamic property for viscoelastic compound structure, and optimization design of dynamic property parameters based on approximate model. Begin with the experimental study of viscoelastic material; the parameters’optimization and verification are carried out according to the test result. Vibration test and finite element modeling on U-shaped pipe and pipe system with this kind of viscoelastic material are performed carefully. At last, three kinds of approximate model about the viscoelastic damping position and shape parameters vs. dynamic response parameters are trained according to experiment design result. A multi-objective optimization model is established, which defines mass and modal damping as optimal objective. And the improved Non-dominated Sorting Genetic Algorithm (NSGA-Ⅱ) is employed as optimal method.The mechanical properties experiment was carried out in Dynamic Mechanical Thermal Analyzer (DMTA). The frequency sweeping result shows both storage modulus and loss modulus increase with the activating frequency, and ascends at a higher rate. And the loss factor has not an obvious peak in 1-200 Hz range. An approach based on GA-BFGS hybrid algorithm is introduced to determine the parameters in a constitutive relation for viscoelastic materials according to the experimentally obtained material mechanical properties. This hybrid algorithm can reduce the iteration times greatly and the results are more exact than single GA algorithm, meantime, the initial values of the design variables are free without limits, which is very important for multi-variable optimization. STD model, GHM model and ADF model were analized and compared from the complexity and computer speed in the sedond chapter. And the ADF model was chosen as the optimum constitutive model for this kind of viscoelastic asphalt material.In order to validate the parameter estimation results, dynamic response analysis of free layer damping (FLD) cantilever beam was performed using experimental and numerical simulation approach. The numerical simulation results have good agreements with experimental results which verified the effectiveness of this method. In view of the virtues of completely compatible with finite element method (FEM), the ADF model is suitable to analyze complicated structure.In view of the frequency dependence of viscoelastic damping material, a suggest method of frequency iterative--modal strain energy--modal superposition was used to calculate the frequency, damping ratio and displacement response. Firstly, considering the frequency dependency property of viscoelastic material, FE iterative and modal strain energy method is adopted to solve the modal frequencies and loss factors. And modal mode and corresponding load are extracted from each modal. Moreover, the displacement response of node at each modal frequency was calculated directly. The displacement response in some points of viscoelastic compound structure can be solved by mode superposition. The comparison of numerical simulations with experiment results indicates that the proposed method is effective for analyzing the viscoelastic material with frequency dependency property. This method can be preformed easily using FE software and a simple program. Therefore it has the potential capability to deal with actual structures with a more complex geometry or boundary conditions.Finite element analysis are often time consuming and costly, especially in structure optimization. Therefore, there is necessary to use approximate model, derived from small numbers of computer runs, to replace additional FE tests during optimization. In this work, an optimization framework is proposed to combine experimental design, approximate model and genetic algorithm to form a computationally efficient global optimization method. Firstly, FE calculation is preformed according to experimental design to create a data set of optimization parameters and corresponding model responses. This data set is then used to train the metamodel to receive model optimization parameters as input and an approximation to the response of the numerical as output. An approximation to the solution is then obtained with the metamodel used in place of the expensive numerical model in optimization algorithm. The approximate model used in this paper includes response surface method, Kriging and neural network. At last, dynamic response calculation is carried out to demonstrate the quality of the approximation optimization models.Damping is an important valuation index of vibration attenuation property. With the same parameters values, the bigger the damping value is, the better the vibration attenuation property is. But, damping layer on the pipe system will add the mass greatly, which increase the design cost. Therefore, the minimun mass and the maximum damping were defined as the optimal objective. And NSGA-Ⅱalgorithm is employed as optimal method. From the Pareto curves between mass and damping, we can see the damping of the pipe structure is growing with the increase of width of the damping layer. But when the width is more than 58mm, the rise rate of the damping is very slow. So, we can choose the mass range is 140 g~160 g as the optimal design range, and the mass ratio range is 13.4%~29.5%, the damping ratio is 7.5%~10.8%. The test result shows with optimal damping layer, the vibration displacement reduced 19.8dB, accelerate reduced 26dB.It is an important part of virtual design of air conditioner’s pipe system. A precise FE model and a simple, feasible calculation method were established in this paper for pipe system. Meantime, new vibration attenuation way was suggested and the corresponding optimal algorithm was given. The results of this thesis may have some guideline value to similar dynamic character analysis of viscoleastic material, and to test, modeling and optimization dynamic response of compound structure.更多还原