【作者】 梁海波；

【导师】 袁赣南；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2011， 博士

【摘要】 微机电系统(Micro Electro Mechanical System, MEMS)惯性器件具有成本低、体积小、功耗低、抗冲击能力强等优点,在航天航空、汽车工业、生物医学工程、移动通信等领域,发挥了重要的作用。由于MEMS惯性器件与传统器件相比精度较差,因此,在充分发挥其优点的同时,可以采用冗余容错和信息融合技术,来提高由其构成的测量单元的测量精度和可靠性,进而提高整个微惯性系统的测量精度和容错能力,达到高可靠性导航、定位及控制的目的。论文依托实验室在研项目,开展基于MEMS陀螺冗余的微惯性航姿系统的研究,为MEMS惯性器件在低成本航姿系统中的进一步应用奠定基础。论文根据航向姿态测量精度和航姿系统可靠性的要求,从微惯性航姿系统的总体设计、冗余陀螺配置方式、冗余陀螺测量单元在线故障诊断方法、冗余陀螺误差的实验室标定与在线标定方法等方面开展了研究。(1)针对微惯性航姿系统的技术要求,提出了由MEMS惯性器件、电子罗盘和GPS构建微惯性航姿系统的方案。其中,MEMS陀螺系统采用冗余设计,在对冗余陀螺配置方式进行论证分析的基础上,为了改善冗余陀螺测量单元的角速率测量精度和可靠性,提出了一种由九个单轴MEMS陀螺构成的新型冗余配置方式,并对其测量精度、可靠性和故障后可重构性能进行了定量分析,结果表明该冗余配置方式比实验室原有冗余陀螺配置方案具备更高的角速率测量精度和容错能力。(2)针对冗余陀螺系统的在线故障诊断方法,对目前广泛使用的最优奇偶向量法进行了理论和试验分析。针对最优奇偶向量法存在的缺陷,将支持向量机理论应用到冗余陀螺系统的在线故障诊断中,并提出了基于支持向量机的故障诊断方法。在对支持向量机的核函数和参数进行深入研究的基础上,采用三步搜索法对参数进行寻优。通过试验表明,基于支持向量机的故障诊断方法可以弥补最优奇偶向量法的不足,且比最优奇偶向量法具有更高的故障识别率,更低的漏检率和虚警率。(3)开展了针对冗余陀螺误差实验室标定方法的探索性研究,提出了采用Kalman滤波对冗余陀螺的常值误差、标度因数误差和安装误差进行估计的实验室标定方法。通过理论和试验分析,不断改进冗余陀螺的误差测量模型,采用小角度旋转向量对安装误差进行描述,并设计了标定试验的编排方式。在对标定滤波器的量测方程进行设计时,采用零空间扩增法来改善滤波器中各个陀螺误差状态量的可观测性。通过对滤波器进行可观测度分析和仿真试验,表明该实验室标定方法能够对陀螺误差模型中的所有误差状态量做出准确的估计,从而证明了该方法的有效性。(4)在对陀螺常值漂移在线标定方法的研究中,以组合导航理论为基础,以惯导系统的误差方程为核心,引入GPS的位置和速度信息作为参考值,采用Kalman滤波对陀螺的常值漂移进行在线标定。为了改善滤波器中随机漂移的可观测性,设计了飞行器的机动方式。通过试验表明,对陀螺常值漂移估计的最大相对误差低于7%,能够满足飞行器的日常使用需要。(5)针对陀螺随机漂移的在线标定问题,提出了基于自适应Kalman滤波的在线标定方法。首先,在对陀螺的随机漂移建立自回归滑动平均(ARMA, Auto Regressive Moving Average)模型的基础上,从理论和试验两个方面将ARMA模型转化为状态空间模型的三种方式进行了分析,给出了适用性较强的模型转化方式；然后,采用“当前”概率密度模型对载体角速率建模,来适应载体机动方式的不确定性；最后,考虑到陀螺随机漂移的时变特性,采用虚拟噪声法来抵消由固定ARMA模型参数带来的标定误差。在以上三个方面研究的基础上,提出了完整的陀螺随机漂移在线标定方法。试验结果表明,经过随机漂移补偿后的航向解算误差均值和标准差最大降幅达到53%,从而证明了该方法的有效性。更多还原

【Abstract】 MEMS (Micro Electro Mechanical System) inertial sensors, with the advantages of low cost, small size, fine shock resistance, play an important role in the field of automobile industry, biomedical engineering, aerospace, mobile communications, national defense science and technology. However, comparing with conventional inertial sensors, the measurement accuracy of MEMS inertial sensors is so poor. Therefore, the technique of fault-tolerance with redundant sensors and information fusion can be adopted to improve its measurement accuracy and reliability, and to improve the performance of Micro inertial measurement system with navigation, positioning and control information. Relying on the going project in the laboratory, several aspects of Micro inertial attitude heading reference system (AHRS) are pre-researched, which lay the foundation for the further application.According to the characteristics of MEMS sensors and accuracy requirements of AHRS, this thesis puts the emphasis on the issues of AHRS overall design, redundant configuration, on-line fault diagnosis, laboratory and on-line error calibration of redundant gyroscopes.Considering the specifications of the MEMS AHRS, an overall scheme including MEMS inertial sensors, electronic compass and GPS is set forth. The technique of redundant configuration is brought into the MEMS gyroscope measurement unit. In the view of measurement accuracy and reliability, a new configuration with nine MEMS gyroscopes is proposed. Then its accuracy, reliability, and performance of reconstruction after failure are analyzed theoretically. The analysis results show that this redundant configuration can meet the design requirements.For the on-line fault diagnosis method of redundant gyroscopes, the performance of widely used method, optimal parity test (OPT), is analyzed by both theoretical and experimental aspects. Taking the limitation of the OPT into consideration, the support vector machine (SVM) is brought forward to the diagnosis method. Firstly, the effect of the SVM kernel function and parameters are studied in depth. Secondly, a simple and efficient parameter optimization algorithm is addressed. At last, a comparison test between OPT and SVM shows that the SVM method has higher recognition rate of failure, lower missing rate and false alarm rate.The exploratory study of laboratory calibration for redundant gyroscope system is carried out. By experiments and theoretical analysis, the redundant gyroscope error measurement model is improved continuously, and the arrangement of turntable test is designed. When designing the measurement equations of calibration filter, a method of zero-space amplification is proposed to improve the observability of filter states. Through observability analysis and simulation test, it indicates that all the states in gyroscope error model can be estimated correctively, which proves the effectiveness of this calibration method.Since the constant drift and random drift of MEMS gyroscope are larger than traditional ones, it is necessary to study how to calibrate them online. When calibrating the constant drift, the integrated navigation theory is recommended. Based on the error model of inertial navigation system, the position and velocity information from GPS is chose as reference value, and then the constant drift of gyroscopes is estimated real time by Kalman filter. In order to improve the observability of states, the maneuver of aircraft is designed. The simulation test shows that the maximum relative estimation error is less than 7%, which can meet the usual needs of flying task. As for the online calibration of random drift, the Kalman filter is used again. Firstly, the model of random drift is established in accordance with autoregressive moving average (ARMA) model, then 3 kind of transformation from ARMA model to state space model is analyzed by experimental and theoretical aspects. The effective transaction is obtained which is the foundation of follow-up study. Secondly, the angular velocity model is established by the method of maneuver target tracking, which can describe all kinds of angular velocity in the aircraft maneuver. Thirdly, in accordance with the time-variable characteristic of the random error, the Kalman filter with time-variable noise estimator is adopted for data fusion between random drift and angular velocity. At last, an experiment is designed to evaluate the performance of the models and the filter. The experiment result shows that the mean and standard deviation of the calculated heading error decreased by 53% after compensating the random drift.更多还原