基于遗传算法优化神经网络的齿轮箱故障诊断研究

发布时间：2018-05-05 14:38

本文选题：齿轮箱 + 故障诊断　；参考：《中北大学》2011年硕士论文

【摘要】：齿轮作为机械设备的传动部分,担负着传递动力和运动的重要使命,齿轮故障不仅会损坏齿轮本身,还直接关系着整个设备的运转,甚至危及到人身安全,所以对齿轮传动系统进行状态监测及故障诊断具有重大意义。遗传算法是模仿自然界生物进化机制发展起来的随机全局搜索优化方法,它可在搜索过程中能够自动获取和积累有关搜索空间的知识,并且自适应地控制搜索过程以求得最优的方案,具有良好的全局收敛能力和收敛速度快、效率高的优点,还具有全局搜索性,能够克服神经网络固有的局部极小值问题,结合利用神经网络的非线性逼近能力,可以实现优势互补,同时避免各自缺点。本文在深入研究齿轮箱的结构与工作过程的基础上,建立了13-4-5型的遗传算法优化神经网络的齿轮箱故障诊断系统,实验系统以齿轮箱为实验对象,将选好的传感器安装到已选测点上,通过测量到齿轮箱振动信号对其进行消噪处理、提取反映齿轮箱运转特征信息的特征值以体现齿轮箱运转的状况,提出改进遗传算法适应值的一种新方法,并将其应用于遗传算法优化的BP网络模型应用于齿轮故障诊断,建立改进后的遗传算法优化神经网络的两种模型并且进行仿真,通过一系列训练和测试,结果表明:通过采用两种GA优化算法分别在18步和44步达到收敛,传统的未经改良的遗传算法需要71步,收敛速度分别提高了74.6%和38.0 %,极大的提高了诊断的可靠性和准确性,本文最后与BP神经网络比较,更加验证了改进后遗传算法的优越性。
[Abstract]:As the transmission part of mechanical equipment, gear takes on the important mission of transmitting power and motion. Gear failure will not only damage the gear itself, but also directly affect the operation of the whole equipment, and even endanger the personal safety. So the condition monitoring and fault diagnosis of gear transmission system is of great significance. Genetic algorithm (GA) is a stochastic global search optimization method which imitates the evolutionary mechanism of natural organisms. It can automatically acquire and accumulate knowledge about search space in the process of searching. It has the advantages of good global convergence ability, fast convergence speed, high efficiency, and global search ability, which can overcome the inherent local minimum problem of neural network. Combined with the nonlinear approximation ability of neural network, the advantages of the neural network can be complementary and their shortcomings can be avoided at the same time. On the basis of deeply studying the structure and working process of the gearbox, this paper establishes a gearbox fault diagnosis system based on the genetic algorithm optimization neural network of model 13-4-5. The experimental system takes the gearbox as the experimental object. The selected sensor is installed on the selected measuring point, the noise is removed by measuring the vibration signal of the gearbox, and the characteristic value reflecting the characteristic information of the gearbox operation is extracted to reflect the condition of the gearbox operation. A new method to improve the fitness of genetic algorithm is proposed. The BP neural network model optimized by genetic algorithm is applied to gear fault diagnosis, and two models of improved genetic algorithm to optimize neural network are established and simulated. Through a series of training and testing, the results show that the two GA optimization algorithms can converge at 18 and 44 steps respectively, and the traditional unimproved genetic algorithm needs 71 steps. The convergence rate is increased by 74.6% and 38.0% respectively, which greatly improves the reliability and accuracy of diagnosis. Finally, compared with BP neural network, the superiority of the improved genetic algorithm is verified.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3;TP183

【参考文献】