【作者】 刘付强；

【导师】 赵琳；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2008， 博士

【摘要】 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。随着器件性能的不断提高，其应用领域不断扩大，但MEMS陀螺仪的精度较低，尚未达到船舶惯性导航设备的应用要求。论文依托实验室在研项目，开展基于MEMS惯性器件的船用卫星天线微型姿态测量系统研究，在实现船用天线稳定系统的低成本的同时，为MEMS惯性器件在船舶惯性设备中的进一步应用奠定基础。论文结合MEMS惯性器件的性能特点与姿态测量系统的精度要求，对姿态测量系统的总体设计、MEMS陀螺仪信号处理、系统捷联矩阵的更新算法等几项关键技术开展了研究。针对MEMS陀螺仪姿态系统误差积累较快的缺点，利用加速度计与磁强计构成的测量系统与陀螺仪系统组合，使组合姿态测量系统能够同时具有稳定的长期精度和较好的动态性能。提出了将微型测量系统与天线基座固联的安装方案，提高了磁强计的标定精度，并保证系统在天线跟踪丢失时能够连续工作。推导了此方案下的姿态系统到稳定平台的角度转换公式，证明方案可行。鉴于MEMS陀螺仪较大的随机误差对系统动态精度影响较大，论文在对其误差分量特点分析的基础上开展了MEMS陀螺仪的随机误差辨识技术研究。在利用陀螺信号的Allan方差分析结果辨识误差系数时，针对目前常用的图解法与最小二乘拟合法的不足，提出了对陀螺仪数据分频采集、对Allan方差结果分段拟合的实验方法，得到了对陀螺仪随机误差参数的较好的辨识结果。开展基于小波阈值收缩方法的MEMS陀螺仪输出信号降噪技术研究，结合姿态测量系统的工作要求，对影响降噪效果的主要因素进行了深入分析。针对通用阈值准则及普通软、硬阈值函数的不足，根据MEMS陀螺输出噪声特点，提出了一种基于自适应双曲阈值的小波降噪方法对MEMS陀螺仪信号进行降噪。仿真试验表明，与普通小波阈值降噪法相比，采用此方法能够对陀螺仪信号更有效地降噪。以MEMS陀螺仪测量信息为状态向量、以加速度计及磁强计测量信息为量测向量构建Kalman滤波方程。推导出两种基于四元数误差模型的系统基本方程，针对此类非线性模型的不足，提出了采用伪量测向量模型对天线姿态测量系统建模的设计方案。推导出具有双伪量测向量的系统方程和状态依赖条件下的系统状态及量测噪声协方差阵，并对四元数归一化的合理实施进行了分析，最终推导出具体的Kalman滤波算法方程。仿真试验表明，此算法可较好地实现陀螺仪、加速度计和磁强计的信息融合，并能克服大的初始对准误差影响，适合于本系统使用。开展了针对变化的系统状态及量测噪声特性的自适应姿态算法的研究。根据MEMS陀螺仪误差特性变化幅值较小、随机性强的特点，以Kalman滤波系统残差协方差模型误差最小为自适应目标函数，推导出系统的自适应Kalman滤波方程，仿真试验表明该自适应算法能够有效克服MEMS陀螺仪误差模型变化对系统的影响。针对船舶机动运动引起的加速度计测量误差的幅值较大的特点，设计径向基函数神经网络对姿态算法进行学习。并根据扰动加速度的可预知性，提出了基于神经网络与Kalman滤波的组合姿态测量系统，克服了船舶机动运动对系统的恶劣影响。更多还原

【Abstract】 MEMS inertial sensors are characterized by low cost,small size,low powerand fine shockproof capability.With the continual advancements of the elementperformance,the application domains of MEMS inertial sensors keep enlarging,however,up to now,MEMS gyros couldn’t meet the practical needs of shipborneinertial navigation equipment due to low precision.The paper relies on the goingproject assumed by the laboratory,makes a study on shipborne satellite antennamicro attitude measurement system based on MEMS inertial sensors to get alow-cost shipborne antenna stabilized system,and form a basis simultaneously forfurther applications of MEMS inertial sensors in shipborne inertial equipment.In view of the performance characteristics of MEMS inertial sensors and theprecision requirements of attitude measurement system,the paper puts emphasison the study of several key techniques,i.e.,the whole design of the attitudemeasurement system,MEMS gyro signal processing,updating algorithm forsystemic strapdown matrix etc.Aiming at the disadvantage that errors of MEMS gyro attitude systemaccumulate quite fast,the measurement system comprising accelerometers andmagnetometers is combined with the gyro system to form an integrated attitudemeasurement system,thereby ensuring both a steady long-term precision and abetter dynamic performance.The installation scheme is presented that micromeasurement system is fixedly connected with the antenna base,to enhance thecalibration precision of magnetometers,and simultaneously ensure the systemuninterrupted when the antenna tracking signal is lost.The formula of angletransformation from attitude system to stabilized platform is derived for thepresented scheme which is demonstrated feasible.Since the large MEMS gyro random errors have a great effect on thesystemic dynamic precision,the identification technique for MEMS gyro randomerrors is studied based on the characteristic analysis of MEMS gyro error components.When identifying error coefficients by the use of Allan varianceanalysis results of gyro signal,due to the weakness lying in the common usedillustration method and least square fitting method in practical applications,theexperimental means of frequency-division collection of gyro data and segmentedfitting of Allan variance results are put forward to implement a betteridentification of the gyro random error parameters.Taking working requirements of attitude measurement system into account,the means to denoise MEMS gyro output signal are studied based on waveletthreshold shrinking,and the key factors for denoising is further analyzed.For theshortages of universal threshold criterion,conventional soft threshold and hardthreshold,an adaptive hyperbola threshold based wavelet denoising approach ispresented to denoise the MEMS gyro signal according to the characteristics ofMEMS gyro output noise.Simulation experiment has been performed to provethe effectiveness of the presented approach in gyro signal denoising incomparison with the conventional wavelet threshold denoising means.The Kalman filter function is constructed with MEMS gyro measurementinformation as the state vector and measurements from accelerometers andmagnetometers as the observation vector.Since both of the fundamental systemequations derived based on two distinct quaternion error models are nonlinearequations,the design scheme is brought forward by adoptingpseudo-measurement vector to model the antenna attitude measurement system.The system equation with double pseudo-measurement vectors,as well as thesystem state covariance matrix and the observation covariance matrix under thecircumstance of state dependency are firstly derived.Then the rationalimplementation of the quaternion normalization is analized,and the Kalman filterequation is finally derived in detail.The simulation shows that the algorithmmakes a pretty good information integration of gyros,accelerometers andmagnetometers,and is immune to large initial alignment error,thereby applicableto the designed system in this paper.The study of adaptive attitude algorithm for the varying systemic state and observation noise characters is developed.According to the small change inamplitude and strong randomness for the MEMS gyro error model,the systemicadaptive Kalman filter equation is derived by taking the minimization of theresidual covariance model error of Kalman filter as the adaptive target function,and the efficiency of the adaptive algorithm to overcome the bad effect of thevarying MEMS gyro error models on the system is demonstrated via simulatedexperiments.For the large amplitude of the accelerometer measurement errorcaused by ship maneuvering motion,a Radial Basis Function neural network isdesigned to learn the attitude algorithm,and an integrated attitude measurementsystem based on neural network and Kalman filter is put forward according to theforeseeable disturbance acceleration,thereby overcoming the bad effect of theship maneuvering motion on the system.更多还原