基于改进二维互补随机共振的微弱信号检测方法及应用研究

发布时间：2017-12-28 02:05

本文关键词：基于改进二维互补随机共振的微弱信号检测方法及应用研究　出处：《安徽大学》2017年硕士论文　论文类型：学位论文

【摘要】：机械设备的故障诊断对于保障人民大众的生命安全、减小不必要的经济损失以及避免社会生产的进度停滞、自然环境的保护有着极为重要的实际意义。在工业现场,对机械设备采用一系列有效的手段检测设备是否产生了故障或者是否有产生故障的趋势,这一举措很具有必要性。检测机械是否有故障需要采集机械运行状态下的声音、振动或者电流等信号,对信号处理分析。但是,从传感器采集到的信号一般情况下都会含有大量来自其他机械零部件的运行噪声和环境噪声,这些噪声成分会使处理信号的过程变得很困难。所以,微弱信号故障诊断对于抑制噪声,增强微弱信号,提高故障诊断的效果有很重要的意义。从信号处理的角度上来说,传统的一维随机共振方法(one-dimensional stochastic resonance,1DSR)是一种很特殊的非线性滤波器,它能够利用一定量的噪声增强放大非线性系统中的微弱周期信号,这种特殊的滤波去噪机制也是传统线性滤波器所不具备的。因此,1DSR对于信号频带和噪声频带重叠的微弱信号提取,拥有很广泛的应用前景。然而,1DSR由于本身性质类似于一个低通滤波器,在滤波效果上还有待进一步改进。本文研究的是一种基于利用噪声增强周期信号,被称之为二维随机共振的微弱信号检测算法,并且还有该算法在轴承故障诊断方面的应用。本文分析和讨论了离散随机共振的微弱信号提取算法,在这个基础上,提出了一种新型的集总平均二维随机共振(ensemble average two-dimensional stochastic resonance,E2DSR)算法,E2DSR方法是采用二维随机共振和集总平均方法相结合在一起,构建成了 E2DSR模型。E2DSR模型的输出信号信噪比 (signal-noise ratio,SNR)也比1DSR的输出信噪比要高,故障特征频率在功率谱中能够很明显地凸显出来。仿真试验和轴承数据实验结果都验证了 E2DSR算法的优点,这种新型的算法具有比1DSR更高效的抑制噪声的性能,具有卓越的噪声消除和微弱信号检测能力。为了进一步研究二维随机共振,本文又提出了一种自适应的二维互补随机共振(two-dimensional complementary stochastic resonance,2DCSR)方法,首先将传感器采集到的轴承信号进行带通滤波和共振解调,随后将包络信号对半分成两段信作为2DCSR的输入,利用输出信号的加权功率谱峭度(weighted power spectrum kurtosis,WPSK)自适应调节系统的参数以达到参数最优化,接着利用二维随机共振的两个输出,其中一个输出信号增强另外一个输出信号,最终得到最优输出结果及其功率谱,用来识别轴承故障类型。数值仿真和实验结果,还有输出信号的WPSK值对比都表明,2DCSR可以很有效地提高轴承故障诊断的效果。综上所述,本文研究了基于改进的二维互补随机共振方法的噪声增强微弱信号检测和在轴承故障诊断方面的应用。本文提出的两种方法都和传统的1DSR方法进行了等条件下的对比,拥有抑制噪声性强、滤波效果好、易于实现等突出的优点。与此同时,实际故障信号也证明了改进的二维互补随机共振方法的优越性和实用性。
[Abstract]:The failure diagnosis of mechanical equipment is very important for ensuring the life safety of the masses, reducing unnecessary economic losses, avoiding the stagnation of social production and protecting the natural environment. In the industrial field, a series of effective measures for machinery and equipment are used to detect whether the equipment has malfunction or whether there is a trend of failure. This measure is very necessary. It is necessary to collect sound, vibration or current signals in the state of mechanical operation to detect or analyze the signal processing. However, in general, signals collected from sensors will contain a lot of operational noise and environmental noise from other mechanical components. These noise components will make the process of signal processing very difficult. Therefore, the fault diagnosis of weak signal is of great significance to the suppression of noise, the enhancement of weak signal and the improvement of the effect of fault diagnosis. From the signal processing point of view, the traditional method of stochastic resonance (one-dimensional stochastic resonance, 1DSR) is a kind of special nonlinear filter, amplification of weak periodic signal in nonlinear systems to enhance noise it can use a certain amount of this special filtering mechanism is the traditional linear filter is not available. Therefore, 1DSR has a wide application prospect for the weak signal extraction of signal frequency band and noise frequency band overlap. However, because of its own properties similar to a low pass filter, 1DSR still needs to be further improved in the filtering effect. In this paper, a weak signal detection algorithm based on noise enhancing periodic signal, called two-dimensional stochastic resonance, is studied, and the algorithm is applied in bearing fault diagnosis. This paper analyzes and discusses the weak signal extraction algorithm for discrete stochastic resonance, on this basis, we put forward a new set of general average two-dimensional stochastic resonance (ensemble average two-dimensional stochastic resonance E2DSR) algorithm, E2DSR method is the use of two-dimensional stochastic resonance and ensemble average method of combining together, to construct the E2DSR model. The output signal to noise ratio (signal-noise ratio, SNR) of the E2DSR model is also higher than that of 1DSR, and the characteristic frequency of fault can be prominently highlighted in the power spectrum. Simulation experiments and bearing data experimental results verify the advantages of the E2DSR algorithm. The new algorithm has more efficient noise suppression performance than 1DSR, and has excellent ability of noise elimination and weak signal detection. In order to further study the two-dimensional stochastic resonance, this paper proposes a two-dimensional adaptive stochastic resonance (two-dimensional complementary stochastic complementary resonance 2DCSR) method, the band-pass filter and the resonance demodulation of bearing signal collected by the sensor, then the envelope signal is divided into two sections on the letter as the input of 2DCSR, using the weighted power output signal the spectral kurtosis (weighted power spectrum kurtosis, WPSK) parameter adaptive system to achieve optimum parameters, then the use of two output two-dimensional stochastic resonance, one output signal enhancement another output signal, then get the optimal output and power spectrum, used to identify bearing fault types. The numerical simulation and experimental results, as well as the comparison of the WPSK value of the output signals, show that 2DCSR can effectively improve the effect of bearing fault diagnosis. To sum up, this paper studies the application of noise enhanced weak signal detection and fault diagnosis in bearing fault diagnosis based on improved two-dimensional complementary random resonance (2-D). The two methods proposed in this paper are compared with the traditional 1DSR method under the equal conditions, and have the outstanding advantages of suppressing strong noise, good filtering effect and easy implementation. At the same time, the actual fault signal also proves the superiority and practicability of the improved two dimensional complementary stochastic resonance method.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.3

【参考文献】