针对石化大型机组在复杂环境下，振动监测信号往往存在大量的非线性、随机、不可遍历的信息，导致复合故障难分离的问题。本项目构建一套全新的理论：首先，在申请人团队前期研究基础下，针对传统的无量纲指标数量不足的问题，本项目提出新无量纲指标的构建方法；然后，针对目前所用的匹配算法不能有效准确区分复合故障信号的问题，本项目探讨在热噪声环境下以新无量纲指标为基础，提出一种全新的曲线匹配的方法；其次，以人工免疫、双样本K-S检验、改进证据理论相结合的数据处理为基本分析方法，构建人工免疫及改进证据理论相结合的复合故障分离方法，完善复合故障诊断新理论；最后，将本项目提出的新理论、新方法，借助广东石化装备安全技术协同创新中心、广东省石化装备故障诊断重点实验室的石化大型旋转机械故障诊断实验平台，进行实验室验证；通过理论证明以及实验室验证，把新成果应用于石化大型机组进行现场验证，并验证研究成果的可适用性和可扩展性。

In view of the large petrochemical unit in a complex environment, vibration monitoring signal tend to have a large number of nonlinear, random, not traverse information, resulting in compound fault difficult to separate. This project presents new dimensionless parameter theory, improved evidence theory, genetic programming method and two-sample K-S testing algorithm to deal with the compound fault isolation problem for large-scale petrochemical unit under complex environment. First of all, on the base of summarizing the previous research results of the applicant's team, new dimensionless parameter construction method is proposed aiming at the lack of dimensionless parameters. Secondly, duo to the current algorithm could not distinguish the compound fault efficiently and precisely, new curves matching method is proposed based on new dimensionless parameter in the thermal noise environment. Then, a new analysis method fused artificial immune, two-sample K-S test and improved evidence theory is constucted for compound fault isolation. Finally, to test the reliability of the new theory and new method, the rotating machinery fault diagnosis experiment platform of Guangdong provincial key laboratory of petrochemical equipment fault diagnosis and Guangdong provincial petrochemical equipment safety technology of collaborative innovation center is applied. Through theoretical proof and validation in the laboratory, the new achievements are applied to large-scale unit in petrochemical field to verify and validate the the applicability and scalability of the research results.