【作者】 崔远慧；

【导师】 唐祯安；

【作者基本信息】 大连理工大学 ， 电路与系统， 2013， 博士

【摘要】 以分布式车载多传感器数据采集网络系统和危险化学品泄漏气体传感器标定系统的研发为基础,系统研究了危险化学品运输槽罐车安全状态分析模型、传感器信号的预处理方法、传感器信号间的相关性以及多传感器数据融合的安全状态评估方法。首先,研制了用于采集槽罐车运行状态和槽罐中危险化学品安全状态参数的数据采集网络系统。该网络系统由车载数据采集终端和远程中央控制中心组成。其中,数据采集终端采集来自安装于槽罐车不同部位的多只传感器的检测信号,其中包括检测槽罐车运行参数的速度、加速度和倾角传感器信号；检测槽罐中危险化学品压力、液位和泄漏气体的传感器信号；检测槽罐车周围环境温、湿度的传感器信号；以及检测非法打开槽罐车进出料口舱门的门开关传感器信号。远程中央控制中心接收并处理由数据采集终端通过无线方式发送而来的车载传感器信号以及通过GPS发送而来的槽罐车位置和速度信号,并综合评估危险化学品运输槽罐车的安全状态及在必要时给出预、报警信息。在远程中央控制中心,根据各传感器特征值的特点,给出了适用于不同传感器的数据预处理方法,并通过线性相关性分析,获得了不同传感器信号与同一特定安全状态之间的线性相关系数,进而提出了利用多传感器信号相关性判断运输危险化学品槽罐车安全状态的方法。其次,为了掌握危险化学品泄漏气体的性质以及使监测危险化学品是否泄漏的气体传感器的数据准确可靠,设计并组建了一个温湿度可调的、高精度、小批量、动态闭环自动标定系统。该系统由四级杆质谱仪、质量流量计、温湿度控制单元和计算机程控系统等部分组成。其中,温湿度控制单元可根据要求控制测试腔的环境温度为0～100℃、湿度为0～100%RH；计算机程控单元将四级杆质谱仪采集到的数据经最小二乘法处理后控制质量流量计的输出,保证配气精度的同时实现配气过程的闭环控制。采用该系统,对目前工业领域使用量最大的12种危化品气体进行了不同浓度配制试验,利用Agilent6890气相色谱仪对4批共132个气体样本进行了浓度检测,系统的最小相对误差为0.01%,最大相对误差为10.61%。讨论了气体种类、环境温湿度等条件对该系统配气精度的影响。第三,为了准确的评估槽罐车运输危化品的安全状态,从集合论的角度,研究了不同集合域的安全状态评估算法。具体讨论了康托集下利用主传感器信号评估安全状态的阈值法、变化率阈值法和改进的C4.5决策树算法。其中,阈值法算法简单,但误报率较高；变化率阈值法因增加了变化趋势的判断,误报率显著降低；改进的C4.5决策树算法因具有较好的处理缺省数据和带有噪声数据的能力,误报率更低。为避免因主传感器故障而造成的误报漏报,又建立了基于系统方法论的多传感器数据融合分析模型,研究了基于信号相关函数的模糊综合评价算法和可拓工程算法。其中,模糊综合评价算法中设计了用于度量每个传感器信号与其它各个传感器信号之间相互支撑程度的误差函数和每个传感器信号与特定的危险化学品运输槽罐车安全状态之间隶属关系的模糊综合函数。利用该算法,对一组含有泄漏和正常安全行驶两个状态的数据进行了判断分析,与阈值法相比,危险状态判断准确率为100%。但该算法对各个传感器信号的稳定性要求较高,且对不同安全状态之间各传感器信号的差异性要求显著,所以又研究了可拓工程函数算法。提出了使用层次分析法计算不同传感器信号之间相互支撑程度,使用可拓函数计算每个传感器信号与特定安全状态之间隶属关系的新观点。利用该算法,进行了一级泄漏、二级泄漏和正常安全行驶三个安全状态的判断分析,危险状态判断准确率为100%,并正确给出了当前安全状态的趋势值,为险情预报提供了基础。最后,为了验证算法的鲁棒性和可靠性,根据实际槽罐车的罐箱体状态,搭建了一个小型罐箱实验系统开展了大量车载实验研究,并参加了863课题合作单位开展的千公里级的陆路和海路考核。结果表明所研究的算法可以实用于危险化学品运输槽罐车的安全状态评估。更多还原

【Abstract】 The paper focuses on how to evaluate the tank vehicle’s safety state for hazardous chemicals transportation. A distributed and multi-sensor monitoring system is designed and a gas sensor calibration system is established, an analytical model for the tank vehicle’s safety status is proposed, and muiti-sensor information algrithems are studied to evaluate the safety status.First of all, a data acquisition network system is designed to collect the running status of the vehicle and the safety parameters of the hazardous chemicals. The network system includes a data acquisition terminal and a remote central control unit. The signals from eight kinds of sensors distributed in different positions are collected in the data acquisition terminal. These sensors can be divided into four sections. The first section is an acceleration sensor and an angle sensor for detecting the tank vehicles’running status. The second section is a gas sensor, a pressure sensor and a liquid level sensor for monitorting the safety status of the hazardous chemiclas in the tank. The third section is a temperature sensor and a humidity sensor for obtaining the ambient environment. The last section is a door switch sensor for an illegal operation. Remote central control unit is used to wirelessly receive data from the data acquisition terminals and evaluat the safety status of the vehicles comprehensively. If a danger occurred, the managers can receive a forecasting or an alarm message. According to the characteristics of the sensors, suitable pre-processing methods are approved for different sensors. And the correlation coefficient between different sensors and a safety status was gained by signals correlations. Furthmore, a new method based on the multi-sensor signal correlation was put forward to detect the vehicle’s safety status.Secondly, the gas sensor is the core element to assess the state of the leakage. In order to guarantee the accuracy of the gas sensor, a gas sensor calibration system is designed and established. This system can configure gas concentration in a closed loop automaticly and continuously with a high precision. And this system can calibrate fourty gas sensors one time. This system consists of a quadrupole mass spectrometer, multiple mass flow controllers, a tempreture and humidity regulator unit and a computer. It can regulate the tempreture range from0to100℃, and the humidity from0to100%RH. Meanwhile, it can control the outputs of the mass flow controllers by dealing with the data from the quadrupole mass spectrometer by the least squares. In order to validate the accuracy of the system,132samples of12kinds of hazardous chemicals gas with different concentrations are sampled. And all these samples were tested by the Agilent6890gas chromatograph.The minimum relative error of the system is0.01%, the maximum relative error was10.61%. According to the results, the relationship between the system accuracy and the ambient environmental temperature humidity conditions is studied.Thirdly, from the respective of set theory, several evaluation algorithms are studied. A threshold method, a ratio threshold method and a revised C4.5decision tree algorithm are studied.The threshold method is simple with higher false rate. While, the ratio threshold method due to the increment trend judgment decreased the false rate rapidly.And the lowest false rate appeared on the revised C4.5decision tree algorithm with a high ability to deal with missing data and noise data. In order to eliminate the false rate, a novel distributed multi-sensor analysis model is put forward based on the system theory, and a fuzzy comprehensive evaluation algorithm and an extension engineering algorithm are studied. The fuzzy evaluation algorithm consists of a error function and a fuzzy evaluation function. The error function expresses the relation between a sensor and other sensors, and the fuzzy evaluation function displays the relation between a sensor and a safety status. A data set containing leakeage status and normal status is analyzed. Compared with the threshold method, the risk judgment accuracy rate of this algorithm is100%. Owing to the signals limited in strict boundary conditions, the fuzzy comprehensive algorithm only can predict two safety statuses. Extension engineering function algorithm is studied. It consists of an analytic hierarchy process and an extension function. The same data set used in the fuzzy function algorithm is analyzed, risk judgment accuracy rate is also100%. Meanwhile, a tendency of the safety status is given precisely in real time.Finally, in order to validate the robustness and reliability of the algorithms, a small-scale tin box experiment system is carried out. And a large number of vehicle experimental researches are simulated and tested. Meanwhile, these algorithms are adopted in an exercises launched by the863project team.Research results show that the algorithms can be used to evaluate the safety status in hazardous chemicals transportation by tank vehicles.更多还原