作　　者： ; ; ; ;

机构地区： 北京理工大学计算机学院

出　　处： 《西南科技大学学报》 2013年第3期81-85,共5页

摘　　要： 对于许多工业产品的质量评估,特别是进行可靠性评估的常用方法之一是基于退化轨迹的方法。该方法通过分析并建立产品退化数据与时间的关系(即退化轨迹模型),进而得到产品的伪失效寿命,最终得到产品的可靠性模型。但现有的退化轨迹模型难以准确地建模退化数据和时间的关系,并且模型的泛化能力比较差。提出一种基于Adaboosting的退化轨迹分析方法,即采用Adaboosting对多个退化轨迹模型进行组合,最终得到动态加权的可靠性模型。实验证明,基于Adaboosting的退化轨迹分析方法可以更准确地建立产品的退化数据与时间的关系,不但能够有效提高产品可靠性预测的准确性,而且有效提高模型的泛化能力。 Degradation trail is one of the usual methods for the evaluation of the quality, especially the re- liability, of many industrial products. This method usually obtains degradation trail model by establishing the relationship between degradation data and time firstly, then builds the model of product reliability. However, each available degradation trail model cannot provide a high precision of the relationship be- tween degradation data and time with the better generalization ability. As combination of several models can establish the degradation trail much better, Adaboosting method is applied to model combination in this paper. With the combination of several degradation trail models by using Adaboosting, a dynamic weighted reliability model can be established. Experiment results demonstrate that combining several mod- els can establish the relationship between degradation data and time better, consequently, improving the precision of reliability prediction as well as the model generalization ability.

关 键 词： 退化轨迹模型 模型组合

领　　域： [理学] [理学] [理学] [一般工业技术]