【作者】 赵望达；

【导师】 徐志胜； 吴敏；

【作者基本信息】 中南大学 ， 消防工程， 2006， 博士

【摘要】 火灾，特别是建筑火灾对人类生命财产的损害是有目共睹的。为了研究建筑火灾发生、发展以及产生破坏的机理，有必要建立一套建筑火灾模拟试验系统对各种建筑构件在明火的作用下力学性能变化进行模拟测试。而测试的模拟升温过程必须严格符合ISO834国际标准。本论文围绕“建筑火灾模拟试验智能测控系统开发和研究”开展了深入的理论和试验研究工作。主要研究内容包括：(1)开发设计了一套包括立式炉、卧式炉、液化气供气系统、点火装置和建筑构件加压装置的火灾模拟试验系统。其创新性特点一是采用液化气加热方式，最大限度地模拟了实际建筑火灾明火升温的特性；二是研制的智能测控系统，满足了建筑火灾模拟试验对多种类建筑构件测试的机动性、综合性和可靠性要求。(2)提出了一个适合火灾模拟试验系统炉温估计的基于主元分析和径向基函数神经网络的软测量模型，解决了火灾模拟试验中炉温在线估计的难题。对于总共采集的119组18路数据，将80组用于主元分析和RBF神经网络初试化训练，确定较为理想的中心距离、权值和隐层数，经MATLAB程序对余下的39组数据用训练好的软测量模型进行测试，构建了在线的炉温估计模型。给出该软测量模型与加权平均方法的估计结果对比，表明了论文设计的软测量模型在炉温估计精度和稳定性方面比常规的加权平均方法有着明显优越性。(3)提出了一种新颖的基于模糊专家PID控制的复合智能控制器结构。深入研究了PID控制器3个参数K_P、K_I、K_D与控制器特性动态和稳态特性的关系，提出了智能在线整定K_P、K_I、K_D参数的方法。进行了专家模糊PID控制器同常规PID控制器控制效果对比试验，从给出的5个时段的对比曲线可以看出，设计出的智能专家PID控制比常规PID控制在静态误差和稳定性等方面有着明显的优势。给出了炉温升温实测结果同标准值的对比曲线。证明炉温控制效果完全符合国际标准ISO834要求，控制精度和稳态误差等性能指标优于设计要求。(4)提出了一种融合热电偶、热电阻和DS18B20数字温度传感器等不同等级温度参数的虚拟仪器数据采集系统结构。研制开发了兼具串行通信和增强型EPP并行通信模式带多台单片机和工控机硬件结构的54路温度采集处理系统。(5)开发了应用Visual Basic 6.0、单片机汇编语言、FRANKLIN C语言和MATLAB语言等多开发平台工具的独特的软件系统。整个软件系统包括炉温数据采集软件、通信接口软件、软测量软件、神经网络的MATLAB训练软件、智能模糊专家PID控制器软件以及试验流程管理软件等子系统。(6)提出了一种预埋温度传感器进行温度实测并同温度场计算方法进行对比的新颖的试验研究方法。从理论和试验角度分析了建筑构件在火灾模拟试验炉中的温度场分布，并同测试温度数据进行比较，给出了各测点的对比曲线结果。证实了所开发的建筑火灾模拟试验系统进行建筑构件火灾性能测试的有效性、实用性和先进性。更多还原

【Abstract】 The damage to the human lives and properties of fire especially building fire is obvious to all. In order to study the happening, developing as well as the destruction mechanism of building fire, it is necessary to establish a set of building fire simulated experiment system, which can simulated test the mechanics performance change of each kind of building component under the action of flame. It must be according to the international standard ISO834 for the system to add fire load to building component in the heating process. Particular theoretic and experiment research tasks are carried on the study and development of a building fire simulated experimental intelligent measurement and control system in this dissertation. The primary research work include:(1) A buildings fire simulated experimental system including a vertical furnace, a horizontal furnace, a set of liquefied petroleum gas system and a device to add pressure to experiment building component is designed. There are two innovation characteristics: on the one hand, using liquefied petroleum gas to furthest simulate the flame heating process of building fire; and on the other hand, the measurement and control system can satisfy the flexible, comprehensive and reliable test speciality of building fire simulation test, which requires to the multi-types building component.(2) A soft measurement modeling based on principal component analysis and RBF neural network is introduced, which adapts the furnace temperature estimate of fire simulated experiment system and can solve this puzzle in fire simulated experimental. For the collected 119 groups 18 loops data, 80 groups are used in PCA and RBF neural network preliminary discipline to determinate appropriate center distance, weight and hidden layer. The other 39 groups data are used to test the disciplined soft measurement model to complete an on-line furnace temperature estimate model. The comparison result of this model with the weight average method illuminates that the model has obvious advantages in the precision and stability of furnace temperature estimate.(3) A new multi intelligent controller based on Fuzzy expert PID control is presented. The relation between 3 PID controller parameters (K_p, K_I, K_D) and the dynamic or stability characteristic of PID controller is thoroughly studied, and a method to intelligent on-line adjust K_p, K_I, K_D is developed in the dissertation. After the comparsion experiment between Fuzzy expert PID and general PID, 5 curves are shown in different time of the heating process, which indicate that the Fuzzy expert PID has obvious advantages in control precision,stability and so on comparing with general PID At last, 5 comparison curves of furnace temperature measured in real time to the standard value are given, and these curves indicate that the control result for fumace temperature is completely according with the demand of the international standard ISO834, and what is more, the performance index such as control precision, stability error and so on are much more excellent than the design request.(4) A data collection stytem with virtual instrument constructure is introduced, which include different type temperature sensors such as thermocouple, thermal resistance and DS18B20 digital thermometer. 54 channel temperature collecting and processing system is developed, and a particular measurement and control hardware constructure is designed, which contain a RS-485 serial interface for MCUs communicating and a EPP parallel interface between MCU and IPC.(5) The software system is developed by a multi-platform tool including Visual Basic 6.0, MCU assemble language, FRANKLIN C and MATLAB. The system is consist of a furnace temperature data collecting software, a communication software, a soft measurement software, a neural network training software based on MATLAB, a Fuzzy expert PID controller software and an experiment process management software.(6) A novel experiment method is put forward by comparing embedding temperature sensors into test object to measure temperature with temperature field calculationan. The temperature field distribution of building component in the fire simulated experiment furnace is analyzed both from theoretic and experiment aspects. As a result, curves of all measuring data comparing with analysis and calculation data are given. The result proves that the building fire simulated experiment system developed by the author have availability, practicability and advancement in building component anti-fire performance test.更多还原