捷联惯导系统(SINS)是一种可用于煤矿井下采煤机定位的自主导航传感器。采煤机初始姿态角的确定作为导航定位的前提，后续导航的性能很大程度上取决于初始姿态矩阵的精确程度。采煤机截割煤壁时产生的剧烈震动会使得SINS敏感到高频噪声，高频噪声会严重降低传统动基座SINS初始对准的性能。为了克服煤矿环境下传统动基座SINS初始对准算法收敛速度慢及精度低等问题，提出一种基于频域分离算子的采煤机动基座初始对准算法。首先，根据采煤机的结构特点，运用质量块法构建采煤机截割煤壁时的动力学模型。在此基础上，利用MATLAB提供的基于Runge-Kutta变步长算法的Ode45函数分析和研究了采煤机截割煤壁时机身的运动状态，之后通过傅里叶变换将时域信息转化到频域，进而确定采煤机截割煤壁时机身的运动频率范围。其次，在传统动基座SINS初始对准算法的基础上，通过分析干扰加速度和采煤机截割煤壁时机身晃动的频率特点，进而引入频率分离算子的概念，并由此利用等纹波最佳逼近法精心设计了有限长单位冲激响应(FIR)低通滤波器，从而有效地抑制SINS敏感到的高频噪声。实验结果表明，所提出的算法可以有效地提高采煤机初始对准的收敛速度以及精度，俯仰角和横滚角的稳态误差<0.1°，航向角稳态误差<0.7°，满足后续精对准的要求。

The strap-down inertial navigation system (SINS) is a commonly-used sensor for autonomous underground navigation that can be used for shearer positioning at a coal mine working face. The determination of the initial attitude angle of the shearer is the premise of navigation and positioning,and the performance of the subsequent navigation depends largely on the accuracy of the initial attitude matrix. The severe vibration induced by the cutting coal wall of shearer will make the SINS more sensitive to high frequency noise,which will degrade the performance of the initial alignment for the shearer. In order to overcome the low precision and slow convergence rate of the existing initial alignment algorithm with strong vibration conditions in underground coal mines,a novel initial alignment method based on frequency domain isolation is introduced. Firstly,according to the structural characteristics of shearer, the dynamic model of shearer when cutting the coal wall is constructed based on the mass block method. In this respect,the motion state of the shearer fuselage when cutting coal wall is analyzed and investigated by using the Ode45 function based on Runge-Kutta variable step size algorithm provided by MATLAB. Then,Fourier transform is used to transform the time domain information into frequency domain,and the frequency range of the fuselage when the shearer cutting the coal wall can be determined. On the basis of the traditional initial alignment algorithm,the concept of frequency domain isolation operator is incorporated and Finite Impulse Response (FIR) filter is designed to suppress the high frequency noise. The experiment results show that the proposed method can effectively suppress the high frequency vibration,and improve the convergence rate and accuracy of the initial alignment. The steady state error of pitch angle and roll angle is less than 0. 1°,and the steady state error of heading angle is less than 0. 7°,which meets the requirements of subsequent fine alignment.