【作者】 方向生；

【导师】 陈裕泉；

【作者基本信息】 浙江大学 ， 生物医学工程， 2007， 博士

【摘要】 电气化在促进生产力和人类文明飞速进步的同时,也给火灾的发生提供了更大的可能性。二十世纪九十年代以来,电气火灾在全国总火灾数中占的比例呈现上升趋势,造成的损失也逐年增加。从国外统计来看,电气火灾次数在总火灾中也占有相当的比例。电气火灾主要原因是线路过载、短路、接触不良、电弧火花、漏电、雷电或静电等引起的异常高温,导致引燃周围可燃物或引起线缆自燃。目前电气火灾的探测主要通过各种电线电缆表面温度检测和基于电磁原理检测,存在误报率高、维护困难等问题。考虑到电气火灾多由于电气故障导致线路异常高温而使得绝缘层自燃或引燃周围物质,而塑料或橡胶材质的绝缘层具有高温热分解特性,在高于额定温度工作时将释放特定的气体,因此本文提出了利用电子鼻技术进行电气火灾探测,构建的系统能够在有限时间内及时探测到电气故障隐患的存在,实现早期预警。本文在电线电缆绝缘材料及其热特性理论分析的基础上,重点分析了常用的PVC电缆热分解过程。通过对电线受热后现象的观察,并结合GC的检测结果,我们发现在电线形态发生明显变化且释放烟雾之前,已经释放出较大量的有机气体。从色谱图中可以看出,随着温度的升高,电线释放气体的浓度升高,种类也有一定的变化,但多集中在低沸点低碳链范围内,气体分析结果可作为电子鼻传感器阵列设计依据。考虑到实时探测的要求,以及电线电缆所处环境因素,本文设计了主动采样模式的电子鼻系统。在采样泵的作用下,电线释放气体大部分被收集至气室中,大大降低了环境中风等因素对气体扩散的影响,提高了气体的采样效率,使得传感器阵列能及时快速的响应。气室的设置使得传感器阵列工作在相对稳定的环境中,提高了系统的稳定性和可靠性。实验结果也表明,主动采样模式相对于被动扩散模式的探测器具有更快的响应速度,这对缩短探测时间具有重要的意义。根据电子鼻的仿生学原理,提出了传感器阵列构建原则。论文同时也提出了引入基于常温固体吸附的气体富集技术,设计了气体富集和直通双取样模式,以适应不同探测目标的特点和要求,进一步提高系统性能。碳纳米管作为一种新的气敏材料,具有灵敏度高、检测限低和常温工作等优点,也是目前研究的热点。本文研究了多壁碳纳米管的预处理和纳米金属粒子掺杂,同时我们制作了以多孔Al₂O₃为基底的金叉指电极,在其上涂敷不同处理后的碳纳米管,制备成安培型的气敏传感器。我们研究了不同处理后碳纳米管的气敏性能,理论分析和实验表明,金属粒子的掺杂改变了碳纳米管的物理结构和电特性,使得其对气体响应灵敏度和选择性都有了明显的改变。由不同碳纳米管传感器构成的阵列满足电子鼻系统对阵列交叉敏感、广谱和冗余的要求,在电子鼻系统中具有很好的应用前景。通过对电气火灾的模拟,得到不同故障条件下传感器阵列的响应模式。分别利用时域阈值算法、基于统计原理的算法和人工神经网络算法对传感器阵列的响应信号进行处理,得到电气火灾的预警模型。结果表明三种算法都可以有效的探测到电气火灾隐患,探测时间比现有文献上报导的大大缩短。由于气敏传感器的广谱性,传感器也会对干扰气体产生响应,而时域阈值算法和统计原理算法由于缺乏对气体源的辨识能力,很容易造成误报。而基于神经网络的模式识别技术不仅可以探测到传感器响应的变化,同时也能根据传感器响应特征区分是由电线释放气体还是其它干扰气体引起的响应,从而实现降低误报警率的目标,提高了系统的可靠性和稳定性。此外,利用神经网络的学习能力,系统可以方便的推广到类似的应用目标中,如变压器油状态检测等。更多还原

【Abstract】 Electrification contributes to great progress in productivity and civilization of the world.However it also causes a great number of electrical fires.Since 1990s,the proportion of electrical fires kept on increasing in China,as well as the loss of property and lives.And the same was found abroad.The predominating mechanism of electrical fire is abnormal high temperature caused by overloads,short circuit,poor connections,arcing,leakage,lightning or static electricity,which may cause ignition of the combustible substance or spontaneous combustion of wires.At present the electrical fire detection methods include wire surface temperature detection and electromagnetism principle based detection,which may cause large number of false alarms and maintenance difficulty.Considering that high temperature will cause thermal degradation of electrical insulation materials and consequently specific gas is released,we introduced electronic nose technology to application of early stage electrical fire precaution.It was able to detect the electrical failure in finite time and realize precaution.After theoretical analysis of thermal characteristics of electrical insulation material and thermal degradation process of commonly used PVC wire,we observed the PVC wire being heated and found that before the release of smoke mass organic gas could be detected by GC.The concentration increased as the heating temperature increased,while the components changed a bit.This result is basis of sensor array design.Due to the need of real-time detection,as well as the environmental factor of wire,an active sampling electronic nose was designed.Under the operation of sampling pump,the majority of released gas is collected to the air chamber. Therefore the influence of wind to gas diffusion is reduced greatly and the sampling efficiency is increased too,thus results in fast response of sensor array.The gas chamber provides the sensor array a relatively stable environment,which enhances stability and reliability of the system.The experimental results also indicated that,the active sampling mode has the quicker speed of response than that of passive diffusion mode,which lengthens the precaution time significantly.A solid adsorption based enrichment system is induced,which can also further enhance the performance of the system. Carbon nanotube as a new kind of gas sensitive material has high sensitivity, low detecting limits and can work on normal temperature,which made it the hot spot at present research.Pretreatment and the nano-metal-particle doping of multi-wall carbon nanotube（MWCNT）were studied.Different MWCNT solutions were drop-deposited onto interdigitated Au electrodes on porous Al₂O₃ substrate to make amperometric gas sensors respectively.The performance of the sensors was carefully studied.Theoretical analysis and experiments indicated that the nano-metal-particle doping changed structure and the electricity characteristic of MWCNT,thus they had obvious difference in gas sensitivity and selectivity,which were essential to form a sensor array in an electronic nose.Different response patterns of sensor array were obtained through simulant electrical failure of different breakdown conditions.Time domain threshold rule, statistical principle based algorithm and the artificial neural network were chosen as signal processing methods to work out the electrical fire precaution model.The results indicated all three algorithms could effectively detect hidden trouble of electrical wires,and the precaution time was much lager than that reported in existing literature.Time domain threshold algorithm and statistical principle based algorithm lack of the ability to distinguish between real electrical failure and nuisance,because gas sensors always have poor selectivity and will response to a lot of nuisances as cigarette smoke.And thus results in false alarms.Pattern recognition technology of artificial neural network can not only detect the response of the sensor array,but also can identify whether it is caused by gases released by wires or other disturbance gas according to the characteristics of the sensor response,thus achieves low false alarm rate and enhances reliability and stability of the system.Due to the learning ability of ANN,the system can easily extend to similar applications.更多还原