【作者】 何昆鹏；

【导师】 吴简彤；

【作者基本信息】 哈尔滨工程大学 ， 精密仪器及机械， 2009， 博士

【摘要】 微机电系统（MEMS）惯性传感器具有成本低、体积小、测量范围大、可靠性高和易于实现数字化等优点,基于MEMS惯性器件的导航、制导技术得到了迅速发展,并被广泛应用到汽车工业、生物医学工程、航天航空、精密仪器、移动通信、国防科技等领域。但是MEMS惯性器件存在测量精度低、噪声大等缺点,需要采取一些必要的措施以提高其精度,除了优化机械结构设计,提高电子线路的性能,以及采取屏蔽外部电磁干扰措施之外,另一种有效的途径是:从应用的角度对MEMS惯性器件和微惯性测量单元（MIMU）进行误差分析及补偿,提高系统的测量精度。为此,全文进行了以下几个方面研究:（1）提出了虚拟陀螺技术具体的实施方案,即在MIMU的每个测量轴上放置了3个单独的MEMS陀螺,组成一个阵列测量同一个角速度,然后通过多个同类传感器的信息融合技术,实现优于单个传感器的性能的目的。首先设计陀螺信号仿真器,仿真产生相关性很小、较小和很大时3种情况下的陀螺信号,验证所设计的虚拟陀螺系统的正确性。然后采集实际系统中MIMU的陀螺信号,分析得到其相关系数约为0.07,再用陀螺信号仿真器产生该相关强度的陀螺信号,7组数据的Monte-Carlo仿真分析发现,虚拟陀螺的精度提高了2.1～3.6倍,而实测数据融合的虚拟陀螺的精度最大提高2.7倍,最小1.9倍。因此,在比较仿真和试验结果后,找出了影响虚拟陀螺的精度的因素,并提出了提高虚拟陀螺精度的具体措施。（2）在实验室标定时,由于MIMU中的加速度计和陀螺均存在较大的测量噪声,尤其是陀螺敏感到的地球自转角速度ω_ie几乎全部被噪声淹没,因此标定中的两个基准量ω_ie和g（重力加速度）受到严重干扰。另外,MIMU的3个测量轴之间不正交的失准角非常大,有时达到5°以上,传统的标定模型和方法不再完全适用。课题中针对这两个问题提出了一种工程实用性强的高精度标定方法,分两级、两步完成对MIMU的标定。首先在元件级标定中,设计Kalman滤波器估计出加速度计和陀螺的零位、刻度因数和二次项系数等。其次进行IMU级标定,在该级标定时,先采用传统的标定方法对不正交角和安装误差角进行粗标定;然后在详细地推导出精确标定模型基础上,以粗标定的结果作为初值进行参数寻优,这样减少了参数寻优时发散的风险,加快了寻优的速度。在设计标定编排试验方案时,采用回归D-最优设计,对试验进行优化,这样准确地估计出MIMU中的陀螺、加速度计的刻度因数误差、不正交角和在三轴转台上的安装误差角等。最后通过试验验证该标定方法,试验结果表明新方法快速、有效,而且不需要额外的设备和繁琐的解耦计算,具有较强的通用性和简便性。（3）MIMU的动态误差是影响系统精度的主要因素。在分析引起系统动态误差的原因后,将动态误差分为刻度因数不对称性误差、非线性误差和由角加速度引起的误差三类,针对这三类误差提出了相应的补偿模型和方法,在通常的陀螺测量模型中增加角速度绝对值项|ω|,以补偿不对称性误差,然后对该改进的模型进行自适应分段,以补偿其非线性误差,最后利用三轴转台速率试验一次性辨识出角加速度项误差系数。补偿前后的试验结果表明,所设计的动态误差补偿方案将系统的姿态测量精度提高约5倍。（4）温度对惯性器件精度的影响主要与环境温度、温度变化速率、温度梯度有关。从惯性器件结构机理定量分析温度的影响,建立误差机理模型比较困难,因此尝试从试验的角度建立误差补偿模型,并针对传统的温度误差标定方法中的不足,提出了一种全温范围内改进的标定方法,新方法最大的特点是将零位和标度因数统一标定,建立以器件温度、器件温度变化率、环境温度、环境温度变化率和输出电压等5个量为输入,角速度、加速度测量为输出的MIMU全温范围内温度误差补偿模型,然后采用逐步回归分析方法,优化MIMU温度分段误差补偿模型,试验结果表明,经过补偿后的加速度计和陀螺在全温范围内误差分别保持在10mg和0.04°/s以内,这样有效地降低了温度非线性对MIMU测量精度的影响,从而提高了MIMU在全温度范围内的测量精度,达到了工程实用的目的。（5）研究了MEMSINS与倾角传感器（SPS003）、电子罗盘（EC）进行组合导航的方式,首先推导出MEMS INS和SP5003的姿态测量模型,然后采用频域辨识技术辨识出SP5003的测量模型中的参数和采用椭圆假设罗差补偿方法补偿EC的罗差,再将MEMSINS、SP5003、EC和陀螺等四者的测量模型进行组合,得到完整组合导航系统模型,最后设计一个最优状态观测器,对3个陀螺的常值漂移进行在线估计和补偿,提高了系统的航姿测量精度。组合系统在三轴转台上的摇摆试验和跑车试验都证明了这种异类传感器信息融合方法正确、有效。更多还原

【Abstract】 Due to the advantages of MEMS inertial sensor, such as low cost, small size, high reliability, large measurement range and easy to digitations, the navigation technology using MEMS inertial sensors has been developed rapidly and widely used in such field as auto industry, biomedicine, aerospace, precision instruments, national defense, etc.However, MEMS inertial sensors have shortcomings like low measurement precision and high noise that need to be overcome. Besides optimizing the mechanism structure, enhancing the circuit performance and shielding external electromagnetic disturbance, another effective approach to fabricate moderate accuracy system with low precision inertial instrument is to analyze, modeling and filter the information of MEMS inertial sensors, to exploit new high calibration method, and to fuse information with alike or unlike sensors.To achieve the purpose what has been discussed above, the dissertation includes several works as follows:Contrive a great deal experiments to analyze the signals of gyros and accelerators in MIMU thoroughly. Those analyses method are spectrum analysis, autocorrelation function analysis, PSD analysis, Allan variance and wavelet analysis, and so forth. The results of experiments were used to know the performance index detail and then estimate the limit navigation precision of MEMS INS on the one hand. On the other hand, the noise statistics of gyros and accelerators provide gist for the parameter design of Kalman filter later, and so help avoiding emanation of Kalman filter due to the inaccurate statistics of model and observation noise."Virtual Gyro" technology was presented to improve the sensor performance. The "Virtual Gyro", which has the same meaning with the JPL lab, was realized by using three MEMS gyros on each measure axes to get the same angular velocity and fusing the information by information fusion technology. This paper brings forward the implementation detail of "Virtual Gyro" technology after introducing its principle. Through simulating gyro signal under situations of awfully small, small and large relativity, the correctness of virtual gyro Kalman filter was validated. And then, practical gyro signals of MIMU was collected and processed. The experiment result shows that virtual improve the whole precision by about 1.9 times minimum and 2.7 time maximum. Besides, the dissertation put forward three methods to improve the relativity of gyro signal.Discussed high precision calibration means when there exists large misalignment in MIMU. This method includes two levels and two steps: component level calibration and MIMU level calibration. Component calibration analyzes the measurement error reason of MEMS inertial components according to the principle of gyro and accelerator first. And design Kalman filter to estimate the error index. In this level, the practical effect of Kalman filter was not as effective as simulation. The reason for this instance was due to the inaccuracy of model. In the end, advanced Kalman filter which absorbed process virtual noise had an excellent performance. MIMU calibration was a must because of the non-orthogonal angle and turntable installation error. An precision measurement model of MIMU was built first based on reasonable hypothesis for one thing and seeking parameters with restriction using least square method for another thing. The first step of calibration is coarse calibration using conventional method. In this step, three axis point to sky and ground for once separately and compute the coarse value of non-orthogonal angle. The second step is to design an optimal experiment project according the D-optimal principle for MIMU precision measurement model.Established dynamic error model and identified the coefficients to improve navigation system precision. Through testing method to establish the dynamic error measurement model, then parameter identification to calibration for MIMU and a mass of experiments to validate the model finally, the effectiveness of model and error compensation were proved.Cut down the influence of temperature on the inertial instruments. The influence of temperature on inertial instruments precision mainly related to environment temperature, temperature variation rate and temperature gradient, this paper develop compensation model due to the difficulty of mechanism model and put forward an advanced calibration method in whole temperature range to solve the defect of conventional temperature error calibration. This method takes component temperature, temperature variation rate, environment temperature and its rate, and output voltage for input, and takes angular velocity and accelerator measurement for output. It uses stepwise regression analysis to get optimized segment error compensation model. Significance analysis and experiment show that this advanced method cut down the influence of temperature variance to MIMU.Discussed integration navigation of MEMS INS and EC. From the angle of combination model, an optimized observation was designed to combine the output of gyros, accelerators, tilt sensor and magnetic compass, and so on. This integration can get relatively precision estimation of inclination and heading angle as well as the constant drift of three gyros.更多还原