Abstract: Economic life determines whether electromechanical equipment can continue to generate economic benefits.Predicting economic life from the perspective of equipment performance degradation can provide support for developing an economic and efficient operation and maintenance plan for the equipment in service.However,the economic life is affected by both physical and technical degradation of the equipment,leading to an uncertainty of the key degradation factors and a difficulty in predicting the economic life.To this end,applying the object-field theory and deep learning technology were applied to propose an economic life prediction method based on electromechanical equipments degradation characteristics.The key physical-field performance factors that affected the economic life of the equipment were screened based on the mutual information method in physical entity performance and intangible technical field performance degradation characteristics.Considering the complexity of the degradation mechanism,the timing and coupling of multiple degradation features,a data-driven deep separable convolutional network prediction model was developed.The validity of the proposed method was verified using the example of a planetary reduction.

Key words: economic life prediction, object-field analysis, performance degradation, deep learning

摘要： 经济寿命是衡量机电装备能否继续产生经济效益的重要指标。基于性能退化预测其经济寿命可为制定经济、高效的运维方案提供支撑。然而,装备物理实体与技术性能退化的双重作用导致退化特征复杂多样且动态变化,使关键退化因子不明确、经济寿命预测难。为此,借鉴物场理论并融合深度学习技术,提出一种基于退化特征的经济寿命预测方法。首先,基于互信息法筛选物理实体退化和无形技术场性能退化特征中影响经济寿命的关键物场性能退化因子;其次,考虑退化机理的复杂性、多类型退化特征的时序性与耦合性,建立时序数据驱动的深度可分离卷积网络预测模型。最后,以行星减速器为例,验证了所提方法的有效性。

关键词: 经济寿命预测, 物场分析, 性能退化, 深度学习