并联式六维加速度传感器的混合解耦及误差自补偿算法

针对惯性仪器普遍存在的误差累积问题,以并联式六维加速度传感器为例,提出了一种新的解决方案。通过融合位形空间和相空间内的解耦算法,推导出关于8个Hamilton转动参量的常微分方程组以及广义加速度关于转动参量的一次多项式,共同构成了新的混合解耦算法。通过剖析混合算法的基本结构及其中间参量的影响因素,找到了产生误差累积效应的根本原因。选取角速度、角位移的方向交替点为系统的振动状态特征点,推导了特征点与Hamilton变量的映射关系,并据此定义了2个状态观测量。基于观测量的局部小值和特征阈值,给出了特征点的判据,进而构建了一种半闭环结构的误差自补偿算法。样机试验结果显示:混合算法的计算精度明显高于纯空间算法;加入误差自补偿算法后,关键参量的相轨迹得到明显收敛,且未破坏整个算法的实时性;随机扰动为±30%时,特征点的误判率、漏判率分别为2.0%和3.5%,且该指标值与扰动正相关;实验室条件下,并联式六维加速度传感器在1 min内的综合误差仅为3.05%,表明新的解决方案有效缓解了误差累积效应。

Hybrid decoupling algorithm and error auto compensation algorithm of the parallel type six axis accelerometer

According to the widespread error accumulation problem of inertial instrument, a new scheme is proposed for example of the parallel type six axis accelerometer. By fusing two types of complete decoupling algorithms in configuration space and phase space respectively, ordinary differential equations about 8 Hamilton rotation parameters and one degree polynomial of generalized acceleration about rotation parameters were derived, which constituted the new decoupling algorithm in hybrid space. By analyzing the basic structure of the new algorithm and the influencing factors of intermediate quantities, the basic reason of error accumulation effect was sought out. By selecting directional alternative points of angular velocity and angular displacement as vibrational state feature points, mapping relationships between feature points and Hamilton parameters were derived, and then two state observed quantities were defined. Based on local minima and feature threshold of observed quantities, the criterions of feature points were presented, and then an error auto compensation algorithm of semi closed loop structure was established. Experimental results indicate that the computational accuracy of hybrid algorithm is higher obviously than that of before. With error auto compensation, the phase trajectories of key parameters are converged obviously, which keep the real time performance of whole algorithm. When the random disturbance is ±30%, misjudgment rate and leaving out rate of state feature points are 2.0% and 3.5%, respectively, and there is a positive correlation between these two index values and noise disturbance. In the laboratory, the composite error within a minute of parallel type six axis accelerometer is 3.05%, which verifies that the proposed scheme is correct and feasible.