【作者】 舒浩华；

【导师】 于德介；

【作者基本信息】 湖南大学 ， 机械工程， 2008， 博士

【摘要】 压缩机作为流程制造业的重要构成装备,其安全高效运行对流程制造业生产经营至关重要。维护作为保障压缩机安全稳定高效运行的重要技术手段得到了学术界和企业界的广泛关注。如何构建先进的压缩机维护策略和相应维护支撑工具是设备维护领域的重要研究课题。本文在国家自然科学基金项目、中国石油化工股份有限公司科技攻关项目和长沙市科技计划项目的支持下,对压缩机的非线性故障机理、故障诊断方法、故障智能诊断技术和状态预测技术进行了深入系统地研究,并结合背景企业典型压缩机组的预知维护需求设计开发了相应的监测系统,主要研究工作和创新性研究成果有:1、在建立转子-轴承系统非线性动力学模型的基础上,通过计算机仿真,分析了系统参数变化对系统响应形式的影响规律以及对系统响应稳定性的影响,发现了如下规律:(1)、压缩机转子-轴承系统的激励频率、偏心量和轴承支座质量对系统的分叉与混沌运动有决定性的影响,在满足一定条件下,系统响应将出现周期运动、概周期运动和混沌等复杂的运动形式,转子系统通向和离开混沌的途径主要有倍周期分叉、阵发性分叉以及概周期运动;(2)、滑动轴承的油膜力对系统的运动状态有本质影响,系统的稳态周期响应存在着倍周期分叉和阵发性分叉等有可能使各系统的运动发生本质变化的分叉现象,倍周期分叉和概周期运动是系统通向和离开混沌的道路之一;(3)、从系统的分叉图可以看出,系统响应对参数的变化敏感,参数的微小变化可能引发系统响应形式的巨大变化。上述规律为转子系统的故障机理与特征分析提供了理论基础和依据。2、针对分形维数对噪声敏感的特点,提出了关联维数与EMD降噪相结合的转子系统故障诊断方法,通过对仿真信号的分析诊断,验证了该方法的有效性,为压缩机松动、碰摩和松动碰摩耦合故障的诊断提供了新思路。针对广义解调时频分析方法特别适用于处理多分量调幅-调频信号的特点,提出了基于广义解调时频分析的压缩机齿轮故障诊断方法,通过对齿轮启动过程仿真和实验信号的分析诊断,结果表明该方法较小波等时频分析方法能有效性地提取齿轮振动信号的调制特征,为齿轮故障诊断提供了新方法。3、研究了基于案例推理的压缩机智能诊断技术,考虑到案例推理技术在智能诊断应用中存在的案例无法共享交换与集成、推理过程不具有重用性和推理检索方式单一等不足,设计了基于本体的案例表示方法和相应检索机制,实现了基于案例推理的设备故障智能诊断,并以某石化企业的压缩机组为实施对象,将基于案例推理的智能诊断技术应用于工程实践,取得了良好的应用效果。4、针对压缩机状态预测问题,研究提出了支持向量机与EMD结合的状态预测方法,该预测方法不仅适用于小样本统计机器学习,而且在预测步数较长的情况下仍然具有很好的预测效果,具有很强的推广能力。对压缩机实际监测数据的预测表明,该方法较单一的支持向量机预测方法在预测精度上有显著改进。5、以大机组的预知维护实施为目标,研究了大机组的群监测技术,该技术利用设备机组群在结构特性、运行工况和技术参数等方面的相似性提升了设备监测的准确性和经济性。以背景企业K5403压缩机为具体监测对象,借助软件组态技术、虚拟仪器技术和面向对象技术,设计开发了压缩机群监测系统,该系统运行稳定、功能完善、结构可扩展性强,可有效应用于流程企业压缩机群监测。更多还原

【Abstract】 As an essential consisting facility in process industry, compressor’s safe and efficient operation plays a critical role in the production and running in process industry. Maintenance, as an important technique measure to ensure the safe, stable and efficient operation of compressor, has been widely concerned by academic and business circles. How to construct advanced maitenance strategy of compressor and the corresponding maintenance supporting tools is an important research subject. Sponsored by National Natural Science Foundation of China, SINOPEC Scientific and Technologic Project and Science and Technology Plan of Changsha, the present paper did a profound and systematic research on nonlinear fault machenism, fault diagnosis methods, intelligent fault diagnosis technique and condition prediction technique. Consisdering demand of predictive maintenance of typical compressor unit in the cooperation enterprise, a corresponding monitoring system has been developed. The main research work and original research achievements are:1. Based on the establishment of nonlinear dynamic model of rotor-bearing system, the influence of system feedback forms and the stability caused by parameter variation is analyzed, which would thanks of computer simulation. The following rules have been revealed:a. The pumping frequency, eccentric quantity and quality of bearing spider in compressor’s rotor-bearing system have a determining influence on the system’s bifurcating and chaotic motion. In given conditions, system feedback will be periodic motion, approximate periodic motion, chaos and other complex motions. The ways to and off chaos for rotor system are mainly period doubling bifurcation, popping bifurcation and approximate periodic motion;b. The oil membrane force of sliding bearing has an essential influence on the system motion status. There are period doubling bifurcation, popping bifurcation and other bifurcations, which may change the motion nature of each system in system steady-state period feedbacks. Period doubling bifurcation and approximate periodic motion is one of the ways for the system directing to and off chaos;c. From the system bifurcation figure, we can see that the system feedback is sensitive to parameter variation. A tiny parameter variation may arise a great change in system feedback forms.The above rules provide theoretical principal and foundation for rotor system fault mechanism and feature analysis.2. Aiming at the sensitive feature of fractural dimension to noises, a fault diagnosis method for rotor system is proposed in this paper, in which correlation dimension and EMD denosing were combined. The analysis results from simulation signals show the validity of proposed method, which provide a new idea for the diagnosis of compressor loose, compressor collision and friction, and compressor loose combined with collion and friction. Aiming at the feature that general demodulation time frequency analytical method is suitable for frequency-modulated and amplitude-modulated signals of multiple components, a compressor gear fault diagnosis method was put forward based on general demodulation time frequency analysis. The analysis results from simulation and experiment signals in gear starting procedure show that this method is superior to that of the wavelet method in extracting modulation characteristics from gear vibration signals, which provides a new method for gear fault diagnosis.3. Intelligent compressor diagnosis based on case reasoning was studied, as well. Consisdeing that cases from case reasoning techniques in intelligent diagnosis application cannot share and integrate, the reasoning process is not reusable, lacking in reasoning searching and other disadvantages, case presenting method based on ontology and corresponding searching mechanism have been designed, which realized intelligent plant fault disgnosis based on case reasoning. Take a compressor unit from certain petrol business as an object, the intelligent diagnosis technique has been applied to real industry and has achieved favorable result.4. Aiming at the prediction of compressor condition, the present paper proposed a condition prediction method to combine supporting vector machine and empirical mode decomposition. This method not only applies to statictical machine learning of small samples, but also has remarkable prediction results in the condition of multi-step prediction, which is applicable to lots of fields. The prediciton on real monitoring data of compressor shows that this method, compared with supporting vector mechine prediction method, is much more accurate.5. In order to accomplish the predictive maintenance application of plant key units, group monitoring technique of plant key units has been studies. Making use of plant units group’s similarity in structure features, operation situation and technical parameter, the technique improves the accuracy and economy of condition monitoring. Take K5403 compressor from the cooperation enterprise as a monitoring object, with the help of software configurating technique, virtual instrument technique, and object-oriented technique, a compressor group monitoring system has been designed and developed. This system enjoys stable operation, complete function and advanced structure expandability, which can be effectively applied to compressor group monitoring system in process industry.更多还原