【作者】 王昕；

【导师】 魏国；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2008， 硕士

【摘要】 在众多的模型参数估计算法中,卡尔曼滤波以其较强的噪声抑制能力正受到越来越多的重视。本文将卡尔曼滤波应用于模型参数估计中,主要有两个应用领域:一个是用于超声波回波渡越时间(Time Of Flight,TOF)的估计。由于超声波易受外界噪声的影响,从而导致其回波前沿到达时间难于精确确定。本文采用已有的单回波包络经验模型作为超声波回波的包络模型,并以此推导出适用于双回波重叠的超声波包络经验模型。分别利用扩展卡尔曼滤波和无迹卡尔曼滤波在单回波和双回波重叠两种情况下对渡越时间进行估计,分析估计误差,比较两种方法的估计结果;另一个是用于多功能传感器模型的参数估计。由于多功能传感器同时敏感多个物理量,从而导致其输入输出关系难以用现有的物理理论加以描述。本文采用Takagi-Sugeno(T-S)模糊模型作为数学模型,在非线性较大和较小两种情况下,利用减法聚类生成初始参数和结构,利用扩展卡尔曼滤波对传感器模拟网络进行逆向建模和参数估计。本文简要介绍了传统卡尔曼滤波算法(Kalman Filter,KF)及其在非线性系统中的两种推广算法,即扩展卡尔曼滤波(Extended Kalman Filter,EKF)和无迹卡尔曼滤波(Unscented Kalman Filter,UKF),并从原理上比较了两种算法的优劣。仿真实验和误差分析表明,卡尔曼滤波算法具有优良的噪声抑制能力,能够精确的估计出模型的参数。同时,由于其本身是一种递推算法,因而对处理器和存储空间要求较低,易于推广和应用。更多还原

【Abstract】 Among the various model parameters estimation algorithms, Kalman filtering is drawing more and more attention with strong noise suppression capability. There are two main application areas which Kalman filtering is used in model parameters estimation. One is used for estimating Time Of Flight of the ultrasonic echo. Ultrasonic is vulnerable to the impact of noise, so its echo beginning is difficult to be estimated accurately. This dissertation will adopt the single echo envelope experience model as the model of the ultrasonic echo envelope, and educe the envelope experience model of two overlapped echoes. Then Kalman Filtering is used in the estimation of TOF under the single echo and two overlapped echoes condition. The estimation errors are also be discussed and compared. The other is used in parameters estimation for multifunctional sensor. The multifunctional sensor is sensitive a number of physical variables, so it is difficult to describe the relations between the input and output using existing physical theories. This dissertation will use T-S fuzzy model as a mathematical model, Subtractive Clustering were used to create the initial parameters and the structure and EKF to construct the inverse model and estimate the parameters of the analog sensor network in two situations which nonlinearity is high and low.This dissertation will introduce the traditional Kalman filtering algorithm and its two extended algorithms in non-linear system, which are, Extended Kalman Filtering (EKF) and Unscented Kalman filtering (UKF) and compares the merits of the two algorithms from principle. The simulation experiments and the error analysis show that the Kalman filtering algorithm has strong noise suppression capability, can estimate model parameter precisely. At the same time, because of itself is one kind of recursion algorithm, it requires low in processor ability and storage space, and is easily to be spread and applied.更多还原