小子样机械系统可靠性分析与预测研究

ANALYSIS AND PREDICTION OF RELIABILITY ON A MACHINERY SYSTEM OF SMALL SAMPLING

描述了以小子样机械系统寿命序列为代表的时间序列系统的浑沌特性,阐述了灰色关联分析和灰色模型在机械产品可靠性分析与预测中可开拓贫信息的工程背景,论述了GM 和BP预测方法在机械系统寿命预测时各自的特长和缺陷,以及作者创立的GM+ BP预测方法具有扬长避短、优势互补的优良特性,同时提出BP神经网络的分形问题和短序列BP网络的构造过程,并以3 个工程实例进一步验证了灰色关联分析和GM+ BP方法的优良特性。本文旨在应用浑沌、分形理论对重大机械产品可靠性分析与预测的小样本开发有所突破。

A time serie system with chaotic properties is described and the life series of a few sample size machinery system are taken as an example of the time series. It is expounded that the engineering background of gray relational analysis and gray models can exploit poor information in analysis and prediction of machinery products' reliability. The advantages and shortcomings of GM and BP predicting methods in machinery systems' life predicting are discussed, and GM+BP predicting methods, which are developed in this paper, carry forward advantages and avoid shortcomings. At the same time, fraction questions of BP neural networks are put forward and the short series BP networks are built, and the fine characteristics of gray relational analysis and GM+BP methods are verified further by three engineering examples. It is aimed at some improvement in a few sample size exploitation of magnitude machinery products' reliability analyzing and forecasting by chaos and fraction theories.