【作者】 张敬芬；

【导师】 赵德有； 马骏；

【作者基本信息】 大连理工大学 ， 船舶与海洋结构物设计制造， 2003， 博士

【摘要】 在实际生产中，机械设备重要零部件的故障往往造成生产停工和产出废品，严重的情况甚至造成人员伤亡。板、梁是船舶与海洋结构物的基本组成单元，而这些结构物的损伤多以板、梁的裂纹表现其故障形式。随着大型结构物的不断出现，由于裂纹未能及时发现而造成的裂纹扩展的结果往往造成整个结构的破坏，等到事故发生的时候一切已不可挽救。船舶与海洋结构物既包括发动机这样的机械设备，又有甲板、舱壁这样的工程结构，结构和设备的故障都有可能造成重大事故，因此，对结构和设备都需要及时准确的故障诊断，防患于未然。在对结构裂纹的诊断中，本文主要致力于板裂纹的诊断研究。由于对裂纹的诊断是基于振动的宏观理论，故不考虑裂纹尖端的应力情况。 基于振动理论的频谱分析方法在工程中较为实用，对螺杆压缩机的转子故障进行了频谱分析诊断，确定了其故障发生的频率范围是转子工作基频开始的高阶频段，对转子故障进行了初步诊断，得到有效的诊断结果。 对悬臂板不同的位置、长度、深度和角度的各种裂纹情况进行了有限元数值仿真分析。在梁结构的振型斜率变化率基础上，得出了板结构两个方向的振型切角变化率。 船舶与海洋工程结构物，如海洋平台等，结构比较复杂，所受载荷和环境激励具有随机性，加上水下部分难以探测，因此选择适当诊断指标是很必要的。由于水下结构贴应变片测量应变不方便，工程实际中一般通过在水上平台测量裂纹损伤造成的模态变化，根据振型、固有频率的振动信息的变化来诊断裂纹情况。本文给出了诊断裂纹的五种指标。经过对不同裂纹情况的研究，得出在各诊断指标中，振型切角变化率指标对裂纹损伤最敏感。分析了不同情况裂纹对固有频率和振型切角变化率裂纹诊断指标的影响。得到结论是裂纹位置越靠近固定边，对诊断指标的影响越大；裂纹越长，对诊断指标的影响越大；裂纹角度越大，对诊断指标的影响越大；裂纹越深，对诊断指标的影响越大。 螺杆压缩机故障常常表现为阴阳转子的各种故障模式。但是，到目前为止对螺杆压缩机阴阳转子的故障模式尚未明确提出。本文综合考虑斜齿轮以及旋转机械转子的故障模式，结合具体的工程实际建立了螺杆压缩机阴阳转子的故障模式。对螺杆压缩机转子的故障诊断中，参考模糊神经网络在低于5倍基频的低阶频段的故障诊断模式，将模糊神经网络的故障诊断技术应用到从转子基频到28倍基频的高阶频段上。解决了类似压缩机这样故障信号主要表现在高阶频段的模糊神经网络故障诊断问题。对裂纹的模糊神经网络故障诊断中，将五种裂纹诊断指标分成三组进行模糊神经网络诊断，对悬臂板裂纹进行了包括位置、长度、深度和角度的全方位诊断。 本文提出了利用噪声信号对螺杆压缩机转子进行了模糊神经网络故障诊断的方法。利用振动信号和噪声信号的非同源数据的模糊神经网络数据融合的螺杆压缩机转子故障诊断模型，对螺杆压缩机转子进行了振动信号和噪声信号的非同源模糊神经网络数据融合技术研究。 本文在多传感器的数据融合技术基础上，提出了诊断板裂纹特征的同源数据大连理工大学博士学位论文不同诊断指标的模糊神经网络数据融合诊断模型。对悬臂板裂纹的诊断采用的是一个传感器采集信号，提取不同裂纹诊断指标的同源数据不同指标融合诊断，这给裂纹的诊断技术提供了一个新的诊断模式。融合后的诊断精度明显好于未融合的诊断精度。 进行了悬臂裂纹铝板的模态试验，得到了完好板和17种不同裂纹板、多裂纹板的试验模态数据库，为以后同类问题的研究提供了有价值的参考。并从中提取裂纹诊断识别指标，对裂纹板进行了基于模糊神经网络数据融合技术的故障诊断，并与实际裂纹情况进行了比较，诊断结果表明同源数据不同诊断指标的模糊神经网络数据融合诊断模型对裂纹可以达到位置、长度、深度和角度的全面刻画。更多还原

【Abstract】 The fault of mechanical main parts usually results in production breaking and waster, the severe fault even do harm to person. Plate and beam is the main element of naval architecture and ocean engineering. Crack damage of plates and beams is the main fault forms of structure. With more and more large structure appearing in our lives, if the crack were not diagnosis in time, the expanding crack would damage the whole structure in the end. Once the accidence occurred, it cannot be recovered. Naval architecture and ocean engineering includes both mechanical equipment and engineering structure, such as motor engine and deck. Fault of both structure and equipment might bring severe accident. To avoid accident in advance, fault diagnosis in time to both structure and equipment is necessary. Crack diagnosis to beam is more perfect than that of plate. Crack diagnosis to plate will be emphasized in this dissertation. The diagnosis to crack is based on vibration without considering the stress influence from the crack tip.Spectral analysis based on vibration theory is practical in engineering. Fault of screw compressor rotor is analyzed based on vibration spectrum analysis. The fault frequency range is at a higher range from the fundamental rotating frequency. Diagnose the rotor primarily and get an effective result.FEM numerical simulation to cracks with different location, different length,different depth and different angle is taken. Mode tangent angle variance ratio ( Atga)of the cantilever plate in two directions is gotten from the beam mode slope variance ratio.Naval architecture and ocean engineering, such as oceanographic platform, is a complex structure with random load and actuator. Its underwater part is difficult to be inspected. It is necessary to make a good choice with appropriate diagnosis index. Because it is difficult to measure the strain by pasting strain gauge under the water, mode variance for cracks is measured on the platform. Diagnose the crack using the vibration information of mode and frequency variances. Five diagnosis indexes areused in this dissertation. &tga is the most sensitive index after researching ondifferent cracks. Influence both to natural frequency and Atga by different cracks isanalyzed. Closer to the clasped edge, longer crack, lager angle crack, and deeper crack can all bring bigger influence to the dynamic features.The fault mode of screw compressor is usually shown as the faults of rotors. There have no specific fault modes of rotor by far. Consult with the fault modes of rotating mechanics and bevel wheel, the fault modes of screw compressor rotor are founded together under the practical engineering. Consult the fuzzy neural networksfault mode under five times fundamental frequency, apply fuzzy neural networks fault diagnosis to a higher nine frequencies range, which begins from the fundamental frequency to the 28 times frequency. When diagnose crack, divide five diagnosis feature into three groups. Make three groups diagnosis features as the FNN input to diagnose the crack’s location, length, depth and angle.Fuzzy neural networks (FNN) diagnosis to screw compressor using noise signal is established. FNN&DF diagnosis mode with vibration signal and noise single is presented. Vibration signal and noise signal from different data resource are data fused when diagnose the screw compressor rotor.Based on the multi sensor data fusion model and established FNN&DF diagnosis model to diagnose cracks. Different crack diagnosis features from the same data resource are fused when the cantilever plate crack is diagnosed. The FNN&DF diagnosis precision is much more higher that of FUZZY NEURAL NETWORKS.Modal experiment of cracked cantilever is carried out. A database of the relation between 17 kinds of cracked plates and mode and natural frequency is established. So did the intact plate. FNN&DF diagnosis to the actual cracked cantilever aluminum plate. Compared the diagnosis result with the actual cracked plates. It is improved that FNN&DF diagnos更多还原