中文摘要：

含有压缩机故障诊断模块的分布式大数据平台,已成为石化企业设备管理的未来趋势;基于粗糙集和支持向量机理论的大数据信息的快速分类方法对故障识别不理想,特别是基于典型轴类故障力学建模与诊断相结合技术的有效分类方法尚不成熟;文中提出一种基于LMD（Local Mean Decomposition,LMD）与多重分型谱信息融合的故障识别方法,对故障大数据流中难以判定的轴承类间故障程度与位置进行有效识别,利用ADAMS动力学建模分析典型故障输出振动响应,通过与实测数据对比验证,证实该方法可提高二次故障分类的准确性。

英文摘要：

Distributed big data platform, which contains the fault diagnosis module of compressor, has become the future trend of equipment management in petrochemical enterprises, fast classification method of big data information based on rough set and support vector machine theory is not ideal for fault recognition, and particularly the effective classification method based on the combination of the typical axial fault mechanics modeling and diagnosis technology is not yet mature,a fault recognition method based on LMD （Local Mean Decomposition, LMD） and multi-fractal information fusion is proposed in this paper to identify the fault location and degree of bearing with difficulty to determine in fault big data stream.ADAMS dynamic model is used to analyze the vibration response of typical fault output.By comparing with the measured data, it is proved that the method can improve the accuracy of the two fault classification.