中文摘要：

针对传统的故障预警预测方法存在误差较大的问题，提出一种基于数据分组处理（GMDH）模型的故障预测方法对滚动轴承的潜在故障进行预警。该方法利用模型选定准则选择最优的预警模型，发出故障预警信息，并设置停机阈值，可为设备的预知性维护研究提供支持。对滚动轴承加速疲劳寿命试验所得的数据进行分析，分析结果表明，利用GMDH模型对滚动轴承故障的预测结果与实际值的拟合程度高，相对误差仅为3．1％，比传统的基于BP神经网络模型的预测精度提高了0．51％。这说明基于GMDH模型的故障预测方法为油气设备的安全运行提供了更可靠的保障。

英文摘要：

There exists remarkable error in the traditional prediction method of fault early warning. Therefore, a fault prediction method on the basis of the group method of data handling （GMDH） was proposed to conduct an ear- ly warning of the hidden fault of rolling bearing. The method selects the optimum warning model according to the model selection criterion. It gives fault warning information and establishes the shutdown threshold, providing the support for predictable maintenance research on equipment. The analysis of the data obtained from the accelerated fatigue life test of rolling bearing shows that the prediction result of rolling bearing fault with the GMDH Model has a high fitting degree with the practical value, with a relative error of 3.1%. Compared with the traditional prediction precision on the basis of BP neural network model, its precision is improved by 0.51%. This shows that the GMDH- based fault prediction method improves the precision and effect of fault prediction.