基于粒子群算法和小波理论的机械故障诊断

发布时间：2018-07-09 22:49

本文选题：粒子群优化算法 + 小波理论　；参考：《武汉工程大学》2013年硕士论文

【摘要】：粒子群优化算法(Particle Swarm Optimization，简称PSO)是一种基于群体智能的优化方法。算法主要利用生物群体内个体的协作与竞争等复杂的行为产生群体智能，并为工程优化问题提供高效的解决方法。在优化过程中，，首先将工程问题转化为简单的数学模型，充分了解该数学模型的具体涵义以及模型中各参数的意义和取值范围；然后对粒子群优化算法所涉及的参数进行初始化，提出基于该工程问题的适应度函数；最后根据适应度函数的精度评价优化参数，直至满足适应度函数的精度。与传统优化算法相比，粒子群优化算法在多维函数寻优方面有着算法简单，参数少，收敛速度快等优点，故在工程实践中有着广泛的应用。本文在研究粒子群优化算法理论的基础上，结合小波理论，将这种算法应用于齿轮箱的故障信号识别中。针对齿轮箱故障信号所表现出的非线性和非平稳性，本文运用自适应Morlet小波建立振动信号的数学模型，旨在提取故障信号的特征信息。自适应小波变换(Adaptive Wavelet Transform，简称AWT)确保粒子群算法所优化的模型参数和自适应小波函数的参数是一一对应的，为高精度的时频分析提供了理论依据。基于对粒子群优化算法和小波理论的应用研究，本文首先对原始信号进行希尔伯特变换，获得信号包络。再利用Morlet小波函数建立信号包络的数学模型，确定待优参数。然后运用本文提出的粒子群优化算法和最小均方误差(LeastMean Square Error，简称LMSE)求出数学模型的参数并优化，将所得参数反代入信号模型中得到模型信号。最后使用自适应连续小波变换对齿轮箱的故障状态进行诊断评估。结果表明基于粒子群算法和小波理论的方法对齿轮箱的故障诊断是有效的。本文集成了粒子群优化算法的参数优化效用和小波理论的时频分辨能力，提出了基于粒子群算法和小波理论的机械故障诊断方法，它能很好地对五种齿轮的损坏程度进行识别。通过对同一种裂纹故障类别下的不同齿轮损坏程度进行识别表明所提出的故障诊断方法是有效且可靠的。研究结果同时也对粒子群优化算法和小波理论在其他领域的应用提供了一种新的思路。
[Abstract]:Particle Swarm Optimization (PSO) is an optimization method based on swarm intelligence. The algorithm mainly uses complex behaviors such as cooperation and competition among individuals in biological communities to produce swarm intelligence and provides efficient solutions to engineering optimization problems. In the process of optimization, the engineering problem is first transformed into a simple mathematical model, and the specific meaning of the mathematical model and the significance and value range of the parameters in the model are fully understood. Then the parameters involved in the PSO algorithm are initialized and the fitness function based on the engineering problem is proposed. Finally the optimization parameters are evaluated according to the accuracy of the fitness function until the accuracy of the fitness function is satisfied. Compared with the traditional optimization algorithm, particle swarm optimization algorithm has the advantages of simple algorithm, few parameters and fast convergence speed in multi-dimensional function optimization, so it is widely used in engineering practice. Based on the theory of particle swarm optimization and wavelet theory, this paper applies this algorithm to the fault signal identification of gearbox. Aiming at the nonlinearity and nonstationarity of gearbox fault signal, this paper uses adaptive Morlet wavelet to establish the mathematical model of vibration signal, aiming at extracting the characteristic information of fault signal. Adaptive wavelet transform (AWT) ensures that the model parameters optimized by particle swarm optimization and the parameters of adaptive wavelet function are one-to-one corresponding, which provides a theoretical basis for high-precision time-frequency analysis. Based on the application of particle swarm optimization algorithm and wavelet theory, the original signal is firstly transformed by Hilbert transform, and the envelope of the signal is obtained. Then the mathematical model of signal envelope is established by using Morlet wavelet function to determine the optimal parameters. Then, using the particle swarm optimization algorithm and the least mean square error (LMSE) proposed in this paper, the parameters of the mathematical model are obtained and optimized, and the parameters are reversed into the signal model to get the model signal. Finally, adaptive continuous wavelet transform is used to diagnose and evaluate the fault state of gearbox. The results show that the method based on particle swarm optimization and wavelet theory is effective for gearbox fault diagnosis. In this paper, the parameter optimization utility of particle swarm optimization algorithm and the time-frequency resolution of wavelet theory are integrated, and a mechanical fault diagnosis method based on particle swarm optimization and wavelet theory is proposed, which can identify the damage degree of five gears well. It is proved that the proposed fault diagnosis method is effective and reliable by identifying the different gear damage degree under the same crack fault category. The results also provide a new idea for the application of particle swarm optimization and wavelet theory in other fields.
【学位授予单位】：武汉工程大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TH165.3;TP18

【参考文献】