【作者】 张宇明；

【导师】 林建辉；

【作者基本信息】 西南交通大学 ， 测试计量技术及仪器， 2002， 硕士

【摘要】 我国的摆式列车研究已经进入工程化阶段，作为此项研究的一部分，检测系统的硬件配置及软件算法是其中的一个重要环节。本文的主要内容是研究检测系统中传感器的选型、故障检测与容错及相关的信号处理算法，为控制系统提供准确可靠的未平衡离心加速度信号和滤波延时补偿信号。 本文根据摆式列车线路试验的大量数据，分析了试验方案中传感器的性能，提出了传感器的选型和改进的方向。研究了传感器的故障检测和容错算法，论述了三种针对单个传感器故障检测的方差算法，以及针对硬件冗余的并联和2／3(G)(三取二)表决系统的容错算法，并详细论述了上述算法的推导及实现过程。设计了多种对加速度信号进行滤波的经典数字低通滤波器，通过对滤波器的性能分析和滤波结果比较，总结了滤波器的优化方向；讨论了基于梯度搜索的两种自适应滤波算法。分别设计了滤波法、小波去噪法和积分法来提取倾摆脉冲(滤波延时补偿信号)，论述了这三种算法提取倾摆脉冲的原理，提出了一种鲁棒性好、适应性强的“高进低出”法来确定三种提取倾摆脉冲算法中的门限值：最后用这三种算法对一次线路试验的全部数据进行了处理，并比较了各种算法的特点。 本文中的所有算法都用线路试验实测数据进行了仿真，对优选出的的算法用标准C语言编制了相应的可移植到DSP系统上的程序，该程序已在专为摆式列车开发设计的DSP系统上运行通过，并且通过了线路试验的初步检验。更多还原

【Abstract】 The research of tilting train in our country has been in the stage of practicality. As one part of the project, the hardware configuration of testing system and software algorithms are very important. The main content of this paper is about sensors’ type-selection in testing system, the sensors’ fault-detection and fault-tolerance, and the correlative signal processing algorithms, providing the control system with accurate and reliable uncompensated signal and filtering-delay-compensation signal.Based on a great deal of testing-on-line data of tilting train, the performance of sensors is analyzed, and the direction of sensors’ type-selection and the sensors’ improvement are put forward. The sensors’ fault-detection and fault-torlerance algorithms are studied, so do the three variance-caculating methods for sensors’ fault-detection, and the fault-tolerance algorithms of parallel connection system and 2/3(G) vote system for hardware redundancy. All the derivation and realization processes of these algorithms are discussed in detail. A lot of classical LPFs are designed for the filtering of acceleration signal, then the direction of filters’ optimization is put forward in terms of the performance analysis of the filters and the filtering results. Two adaptive filtering algorithms based on gratitude search are also discussed. The principles of filtering, wavelet-denoise and integral algorithms to pick up the tilting pulse (filtering-delay-compensation signal) are discussed. Especially, a "high-in & low-out" method is presented to determine the thresholds in the three algorithms; it is very robust and adaptable. In the end, the performance of the three algorithms is compared to each other based on the process of total data of one test-on-line.All the algorithms presented in this paper are simulated through the data gained in the test-on-line. The C program to realize the optimal algorithms can be easily transferred to DSP system, it performs well in the DSP system designed for tilting train, and it has also passed the preliminary test in the test on line.更多还原