随着网络技术的发展和新型设备的应用，EPON（以太无源光网络）以其低成本、高带宽、扩展性强和灵活重组的特点，成为接入网的最佳选择。EPON作为PON（无源光网络）技术和以太网技术结合的产物，可以使用SNMP（简单网络管理协议）实现网络管理，但是EPON又具有不同于传统网络的特点，对EPON的管理成为新的课题和研究对象。论文主要研究EPON管理系统中故障管理模块的设计与实现。 论文从分析EPON故障管理模块的技术需求和实际应用需求入手，完成对EPON故障管理模块的流程设计和功能设计，并将EPON故障管理模块划分为告警采集、告警过滤、告警显示、告警处理四个子模块。通过对EPON告警信息的分析完成告警数据格式和数据库表的设计，在数据库层采用Hibernate框架实现对多数据库的支持；采用TRAP和主动轮询机制实现对告警数据的采集；在服务器端实现告警显示子模块、告警过滤子模块和告警处理模块功能；利用WEB浏览器实现人机交互界面。在分析数据挖掘、人工智能等告警相关性分析方法优缺点的基础上，提出使用基于规则推理和基于事理推理相结合的告警相关性分析方法，采用Drools规则引擎技术实现对告警信息的相关性分析。最终使用JAVA开发语言及SSH（Struts-Hibernate-Spring）开源框架完成EPON故障管理模块的系统实现。经过系统测试与试运行，本系统具有完备的功能、良好的稳定性，并且能够准确、实时地捕获网络设备的故障信息，故障响应时间低于电信故障管理10s的标准，系统具有良好的性能。 最后，对论文所做工作进行了总结，提出可行的改进方案，并对改进方案进行了初步探讨和展望。

With the development of network technology and application of new equipment, EPON (Ethernet Passive Optical Network) has become the best choice of Access Network, which has the advantages of low-cost, high-bandwidth, scalability and flexible to restructure. EPON is a combination of PON (Passive Optical Network) technology and Ethernet technologies; we can use SNMP (Simple Network Management Protocol) to achieve network management. But EPON also have some characteristics different from the traditional network, so EPON management has become a hot research topic. The paper is mainly about the design and implementation of Fault manage module in EPON network. This paper starts from the analysis of technology needs and actual application requirements of Fault manage module in EPON network, and then complete the process design and functional design. Fault manage module is divided into four sub-modules: alarm acquisition, alarm filtering, alarm, alarm handling. According to the analysis of EPON alarm information, alarm data format and database tables were well designed, using Hibernate framework to achieve the multi-database support. We use TRAP and active polling mechanism to collect the alarm data; the functions of alarm display sub-module, alarm filtering sub-module and alarm processing module were fulfilled at server-side; WEB browser was used to achieve interactive interface. Based on the analysis of several alarm correlation analysis method, such as data mining, artificial intelligence, we propose a alarm correlation analysis new method which combine rule-based reasoning and common sense-based reasoning ,using rules engine technology Drools to achieve the correlation analysis of alarm information At last we use Java development language and open-source framework SSH (Struts-Hibernate-Spring) to complete the fault management system. According to the testing result, we know the system has fully functional, good stability; it also can capture the network equipment failure accurately and in real time. The failure response time of system is less than 10s, which is the standard of telecommunications fault management, the system has good performance. Finally, all the work of this paper was summary, and the possible improvement was proposed and further discussed.