基于智能算法与模糊控制的室内环境反向设计方法研究

发布时间：2018-05-25 03:39

本文选题：遗传算法 + 人工神经网络　；参考：《天津大学》2015年博士论文

【摘要】：室内环境反向设计理论是当前研究的热点问题。反向设计能够根据设计目标的特定要求得到相应的设计变量值,为室内环境设计过程提供理论指导,具有广阔的商业前景。在全面分析国内外反向设计理论和发展趋势的基础上,本文以计算流体力学(CFD)为基础,将遗传算法、人工神经网络、模糊控制等技术相结合,提出了一种高效的室内环境反向设计方法,应用于Blay模型与客机座舱模型的反向设计问题。对Blay模型,选取入口速度与温度为设计变量,实现了监测点处的速度与温度的反向设计;对座舱模型,选取入口速度、入口温度、入口角度、入口及出口位置为设计变量,实现了控制区域的热舒适、吹风感、空气龄、局部速度及垂向温差等的多目标优化设计。主要成果如下:首先,研究了基于遗传算法的反向设计方法。使用单目标遗传算法对Blay模型进行了反向设计。使用多目标遗传算法对飞机座舱环境进行了反向设计,利用非支配关系解决了多目标设计过程中的个体排序问题。利用区间遗传算法获得了座舱模型的区间解,并使用超体积评价解的区间特性,得到了8个区间解,并确定了最优区间解。研究发现,仅使用遗传算法进行反向设计时,计算量很大,难以满足实际设计要求。将人工神经网络引入反向设计中,利用人工神经网络预测新个体的目标值,减少了反向设计的计算量。为保证设计精度,同时使用CFD和神经网络计算新个体的目标值。提出了对数归一化方法和优选初始种群方法提高反向设计效率。进一步提出了自更新神经网络,实现了训练样本量的自适应,并使Blay模型与客机座舱模型反向设计的计算量分别降低了49.4%和60.8%。将模糊控制技术应用在反向设计中。利用模糊控制与神经网络获得个体的计算概率,结合模糊控制与遗传算法获得个体的进化区间。在二维及三维座舱模型的反向设计中,上述计算概率和进化区间控制使计算量再分别下降26.6%和24.6%。总之,本文提出的反向设计方法提高了室内环境反向设计效率,客机座舱反向设计的计算量减少了71.3%。
[Abstract]:The theory of reverse design of indoor environment is a hot issue at present. The reverse design can get the corresponding design variable value according to the specific requirements of the design objective, which provides theoretical guidance for the interior environment design process, and has a broad commercial prospect. Based on the comprehensive analysis of the theory and development trend of reverse design at home and abroad, this paper combines genetic algorithm, artificial neural network, fuzzy control and so on, based on computational fluid dynamics (CFD). This paper presents an efficient reverse design method for indoor environment, which is applied to the reverse design of Blay model and passenger cabin model. For the Blay model, the inlet velocity and temperature are selected as the design variables to realize the reverse design of the velocity and temperature at the monitoring point, and for the cabin model, the inlet velocity, inlet temperature, inlet angle, entrance and exit position are selected as design variables. The multi-objective optimization design of thermal comfort, blowing sense, air age, local velocity and vertical temperature difference in the control area is realized. The main results are as follows: firstly, the reverse design method based on genetic algorithm is studied. A single objective genetic algorithm is used to reverse design the Blay model. Multi-objective genetic algorithm (MGA) is used to reverse design the cockpit environment, and the non-dominated relation is used to solve the individual scheduling problem in the multi-objective design process. The interval solution of cockpit model is obtained by using interval genetic algorithm. By using the interval characteristic of hypervolume evaluation solution, eight interval solutions are obtained, and the optimal interval solutions are determined. It is found that when genetic algorithm is only used for reverse design, it is difficult to meet the actual design requirements because of the large amount of calculation. The artificial neural network is introduced into the reverse design, and the target value of the new individual is predicted by using the artificial neural network, which reduces the calculation amount of the reverse design. In order to ensure the design accuracy, CFD and neural network are used to calculate the target value of the new individual. Logarithmic normalization method and optimal initial population method are proposed to improve reverse design efficiency. Furthermore, a self-updating neural network is proposed, which realizes the self-adaptation of the training sample size, and reduces the computation of the inverse design of the Blay model and the cockpit model of the passenger plane by 49.4% and 60.880%, respectively. The fuzzy control technique is applied to reverse design. Fuzzy control and neural network are used to obtain individual computing probability, and fuzzy control and genetic algorithm are combined to obtain individual evolution interval. In the reverse design of two-dimensional and three-dimensional cockpit models, the computational probability and evolutionary interval control decrease the computational load by 26.6% and 24.6g, respectively. In a word, the reverse design method proposed in this paper improves the efficiency of the reverse design of the indoor environment, and reduces the calculation amount of the cockpit reverse design of the passenger plane by 71.33.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：V223

【相似文献】