【作者】 吴燕玲；

【导师】 童水光； 钟崴；

【作者基本信息】 浙江大学 ， 化工过程机械， 2010， 博士

【摘要】 锅炉是国民经济中供应蒸汽和热水,实现能量的形式及载体变换的重要装备,是工业生产应用中不可或缺的。锅炉产品设计是一项复杂、繁琐、经验性强、对设计经验依赖性大、设计周期长、对经济性和可靠性要求高的大型常规工程设计。传统的锅炉产品设计主要依靠设计人员的经验,通过反复调整设计方案和重新计算来完成,优化过程不易实现,设计质量难以保证,且设计效率低、设计周期长。锅炉受热面系统设计是锅炉设计的重要内容,研究受热面系统的智能优化设计,进行优化设计方案的有向搜索,对于进一步提高锅炉产品设计自动化水平、提高设计方案优化水平、提高设计质量和效率,降低设计成本进而降低制造成本具有重要意义。本论文在“十一五”国家科技支撑计划项目(2006BAF01A46)和浙江省科技计划项目(2006C21 SA160008)的资助下,研究了锅炉受热面系统的计算机辅助设计方法,围绕受热面系统智能优化设计的三个关键技术进行分析研究：首先采用了过程系统理论中的“序贯模块迭代法”进行受热面系统设计方案性能分析；然后提出了利用遗传算法的并行计算功能,有向搜索产生多个满足设计要求的优化设计方案构成Pareto解集,实现方案的自动生成；再是运用层次分析法对Pareto解集中的设计方案进行综合评价并给出综合评价指标,为选择最终输出方案提供依据。结合以上三个技术,建立了锅炉受热面系统的智能优化设计模型,从而为锅炉受热面系统的性能分析、智能优化方案设计提供了有效方法。论文的主要研究内容如下：对锅炉受热面系统的热力性能分析进行了认知和建模。从一般意义上出发,认定锅炉属于一种进行能量形式和载体变换的过程系统,是紧凑型的换热网络。分析系统结构组成和工作原理,将整个锅炉受热面系统进行模型抽象化,认为受热面系统是由“广义部件”和“广义流体节点”组成的“广义过程单元”通过一定的顺序串接而成。选择过程系统稳态模拟理论中的“序贯模块迭代法”作为受热面系统热力性能分析的通用总体算法,实现设计方案性能自动分析。在实现锅炉受热面系统设计方案性能自动分析的基础上,提出了基于遗传算法的锅炉受热面系统优化设计方法。分别进行了单个受热面部件和受热面子系统的模型分析与搭建,建立了以热力性能分析为基础,汽水阻力性能分析及烟风阻力性能分析等多个性能分析简化为约束的遗传优化设计模型,从而实现对解空间的有向搜索并同时自动生成多个满足要求的锅炉受热面系统优化设计方案构成Pareto解集。其中,分析了设计过程的各种约束条件,如几何约束、性能约束等,以保证遗传算法得到的每个方案个体均满足设计要求。在基于遗传算法的锅炉受热面系统优化设计方案自动生成的基础上,提出了遗传算法的改进方法。采用基于实例推理与随机生成相结合的初始群体产生技术,将设计经验引入到遗传算法中,帮助快速寻找全局更优解。结合“成员分组法”与“家族竞争法”,并采用“S交叉”的自适应交叉操作,保护父代个体中的优良基因不至于在交叉操作中流失。在变异操作的环节上,采用Sigmoid函数计算自适应变异值,使得变异概率随着遗传算法的进行逐渐由大变小,增加遗传算法跳出局部最优的几率,提高群体平均适应度值。经过实例检验,这些改进方法能够促进遗传算法快速、有效地找到全局更优解,提高Pareto解集的质量。提出了基于层次分析法的综合评价技术对遗传算法自动生成的Pareto解集中各个设计方案进行综合评价。锅炉受热面系统设计涉及众多学科知识,并必须满足设计要求、制造要求、运行要求、系统维护要求等一系列确定性知识和模糊性知识。根据设计的各项要求,对评价指标进行分层归类,建立了以总投资费用和年运行费用总和最小为优化目标层,以各部件费用、燃料费用为第二层,部件的性能参数指标为第三层,Pareto解集中的个体为方案层的综合评价体系。结合专家经验按各评价指标在总优化目标中占据的重要性给予权重系数,把定性与定量结合起来,最后通过各个层次的分析导出整个方案的综合评价值,为最终选择最优方案提供依据。锅炉受热面设计属常规设计,并且带有一定的经验性。建立了锅炉受热面系统性能分析的软件模型,基于遗传算法多个方案自动生成的软件模型以及基于层次分析法的锅炉受热面系统设计方案综合评价软件模型,最终建立基于经验的锅炉受热面系统智能优化设计方法。该方法提高了设计的质量和效率,缩短了设计周期,在一定程度上降低了对设计经验的依赖,为进一步实现锅炉产品的智能优化设计指明了一个方向。更多还原

【Abstract】 Boiler is functional important and widespread utilized in industry. It is a kind of important equipment which is used to supply steam or hot water by changing energy form and energy carrier. The design of a boiler is a big project, which is complex, complicated, time consuming and relies heavily on the experience of the designers. It is a kind of routine design, but strict with economy and reliability requirment. Traditional design method was usually done by changing the parameters and calculating constantly until it reached a satisfactory solution. Heaing surface system is part of the boiler and is very important to the whole boiler. It takes a big proportion in the aspect of steel consumption. So the research of the heaing surface system intelligent optimization design is very meaningful. The optimization could improve design efficiency, quality and diminish the cost of output. In addition, it also could improve the whole design’s automatic level.The thesis was in the subsidization of "Project in the National Science & Technology pillar program during the eleven five-year plan period" (2006BAF01A46) and "Zhejiang Technology Plan Program"(200621SA 160008). It researched the boiler intelligent computer aided design, mainly focused on the three key techniques:the performance analysis of the heating surface system; a genetic algorithm based method and it’s modified model for searching the design schemes automatically and composing a Pareto solution set; analytic hierarchy process(AHP) based method for comprehensively evaluating all the schemes in Pareto. In the end, an intelligent optimization design model of the heating surface system was created. The main research contents on this subject were explained as follows:(?) Based on the cognition of the research process, analyzed and modeled the boiler heating surface system’s thermal calculation. In general sense, rechoned the boiler as a process system which was used to change energy form and energy carrier. In the microscopic point of view, analyzed the structures and the principle of the operation, abstracted the whole thermal calculation system. Reckoned the system was a net work composed of "Generalized Process Cells" in a fixed organizational structure. Took the steady-state simulation theory in the process system "Sequential modular method" as the overall calculation algorithm. Eventually got the whole performance analysis done automatically. It was improved that the method could converge automatically and relied less on the initial values.(?) On the basis of the automatical simulation for the thermal calculation system, used genetic algorithm to optimize the design procedure. The objective of the optimization was to develop and test a model of area estimating for the heating surfaces. Analyzed and modeled both the sole surface and the subsystem which is composed of several surfaces. The estimation was very complex for several performance calculations, including thermal calculation, hydraulic resistance calculation and fuel-gas & air resistance calculation, should all be done. The model of the optimization was developed based on thermal calculation while other performance calculations are simplified as constraints, which were introduced as geometric constraints, velocity constraints and temperature constraints to insure that the candidate design schemes satisfied to all the design specifications. The GA based optimization method could beamly search the whole solution space. At last, a Pareto solution set, which was composed of several optimized schemes, was produced.(?) Modified genetic algorithm models were proposed on the basis of the simple genetic algorithm. Case-based reasoning and randomly generating were combined in initial population generation process, which could help introducing design experience into genetic algorithm effectively and finding the global optimal solution efficiently. Used "Method of Members Grouping", dividing the population into two sub-groups, then the father individual and mother individual were picked from the two sub-groups respectively. Used "Sigmoid Crossover" to generate new individuals. The best two child individuals which won in the "Family Competition" were chosen as new population members. This method could avoid good quality gene of parent individuals being lost in crossover operator. In the section of mutation operation, used an adaptive mutation. The value of adaptive mutation fraction was calculated by a "Sigmoid function". In the early stage of genetic algorithm, a larger mutation fraction was given, and it became smaller gradually with the development of the algorithm, which could also improve the average fitness of the population. Examples were given to prove that these modified methods could promote the genetic algorithm to find better global optimum quickly and effectively, and eventually improve the quality of the Pareto solution set.(?) The design of the heating surface system involves much knowledge in many fields, and it has to meet a series of certain knowledge and fuzzy knowledge, such as design requirements, manufacturing requirements, operational requirements, maintenance requirements and so on. Analytic hierarchy process was used to evaluate the design schemes in Pareto solution set. With the goal of cutting the investment and operating expense, various requirements in design process were classified into two hierarchy and their weight parameters in correspond to their proportion of the cost are given under the guidance of expert experience. In this way, qualitative and quantitative were combined. Finally, the last evaluation values of the whole programs were exported by analyzing each hierarchy, giving the criterion of the best solution selection.(?) The design of boiler heating surface system is a kind of routine design, various values are calculated by theoretical formula. but the actual operation conditions can not be fully consistent with the theoretical calculation. so the boiler products design depends heavily on experience. In this thesis, intelligent optimization design of boiler schemes was proposed based on empirical, in which the design experience was expressed as all kinds of constraints. With GA’s ability of creating several design shemes simultaneously, and the AHP’s comprehensive evaluation ability, computers could create, analyze, evaluate and export projects intelligently. This method could improve the design quality, reduce the reliance on the experience, shortens the design cycle.更多还原