摘要:
【目的】受轮机舱机械噪声、电磁噪声及轴系振动等干扰，螺旋桨推力在轴系上产生的微弱形变信号易被干扰噪声淹没，导致当前难以准确测量推力。本文旨在提出一种在低信噪比亦有很高精度的新的螺旋桨推力测量方法。【方法】傅里叶变换（FFT）、小波分析及经验模式分解（EMD）等常用降噪方法均是基于纯数据降噪，未考虑测量数据中潜藏的力学机制，导致在实船强噪声干扰环境下降噪效果不佳。不同于这类降噪方法，Kalman滤波既考虑测量数据噪声影响，又考虑数据中蕴藏的力学机制，因而有更佳的去噪效果。因此，利用Kalman滤波结合推力-形变状态方程对轴系推力测量进行研究，提出一种螺旋桨推力高精度、在线辨识方法。【结果】以恒定转速、变转速及低频波动转速三种工况为例，研究了不同信噪比下本文方法的推力辨识精度与鲁棒性。研究表明，在信噪比仅为20dB时，推力辨识最大相对误差仅为3.56%。【结论】提出的方法在低信噪比下仍有很高的辨识精度与鲁棒性。同时，该方法属于时域辨识方法，在转速突变、螺旋桨缠绕渔网等突发工况时亦能实时跟踪推力变化，因此可用于螺旋桨推力及轴系状态的在线、实时监测。
Abstract:
[Objectives] Online, real-time, and high accurate monitoring of the propeller thrust is of great significance to the hull-engine-propeller matching design, rapid prediction of the ship, and health management of the shaft. However, due to the interference of mechanical noises, electromagnetic noises and shafting vibrations in the engine room, the weak deformation signal generated by the propeller thrust on the shaft is easily drowned by these interference noises, which makes it difficult to accurately measure the thrust. [Methods] Common signal denoising methods, such as Fourier transform (FFT), wavelet analysis and empirical mode decomposition (EMD) only consider the measurement data, without considering the mechanical mechanism hidden in the measured data, resulting in poor denoising effect in the strong noise interference environment of real ships. Unlike this kind of denoising method, the Kalman filter not only considers the noise influence of the measurement data, but also considers the mechanical mechanism contained in the data, so it has better denoising effect. Given this, in this study, a high-precision online identification method of propeller thrust is proposed using the Kalman filter and strain measurement signal. [Results] Taking the three working conditions of constant speed, variable speed and low frequency fluctuating speed as examples, the proposed method's thrust identification accuracy and robustness under different signal-to-noise ratios are studied. The research shows that when the signal-to-noise ratio is only 20dB, the maximum relative error of thrust identification is only 3.56%. [Conclusions] Hence, the proposed method still has high identification accuracy and robustness at a low signal-to-noise ratio. Besides, the method proposed in this paper belongs to the time domain identification method. It can track the thrust change in real time under sudden conditions, such as the sudden change of rotation speed and the twisting of the propeller with the fishing net, so it can be used for online and real-time monitoring of the propeller thrust and shafting state.
下载:
