基于表观堆栈遗传编程的回转支承寿命预测

Life prediction of slewing bearing based on epigenetic stack genetic programming

由于回转支承的损伤机制不清楚,传统的寿命预测模型不能找出其寿命与振动信号的数学关系。本文改进传统的遗传编程算法(GP),根据表观遗传学的最新研究和堆栈结构,提出了基于表观堆栈遗传编程(ESGP)的寿命预测方法。先从时域、时频域中提取多个特征值,再用动态等距离映射算法从高维特征值中提取能够反映回转支承退化趋势的单一特征值,最后用ESGP进行寿命状态识别。结果表明,该算法成功地找出回转支承的寿命与振动信号的数学关系,为回转支承的损伤机制研究提供理论基础。其寿命预测精度、模型简洁度要高于传统GP。

The damage mechanism of slewing bearing is still not clear and traditional life prediction model cannot find the mathematical expressions between the life and vibration signal of slewing bearing. According to the latest research results of epigenetics and the idea of “stack structure”,a multi domain feature method based on epigenetic stack genetic programming （ESGP） was proposed for slewing bearing life prediction by the modification of the traditional genetic programming （GP）. The features were extracted from time domain and time-frequency domain, and isometric mapping （ISOMAP） was used to reduce high-dimensional features to a single feature which could better reflect degeneration of slewing bearing. The proposed method had the mathematical relationship between the life of slewing bearing and vibration signal, which provided theoretical basis for the research of damage mechanism of slewing bearing. The precision of life prediction was higher than that of traditional GP and the life prediction model was simpler.