【作者】 刘怀喜；

【导师】 张恒；

【作者基本信息】 武汉理工大学 ， 材料学， 2005， 博士

【摘要】 本文针对复合材料高速储能飞轮的现状和发展要求,在查阅和分析了大量国内外相关文献的基础上,针对复合材料储能飞轮的损伤与断裂问题进行了系统的研究。首次将声发射方法应用于飞轮损伤的检测,用人工神经网络对损伤模式进行识别,并获得了成功。 本文的研究内容包括以下几方面:一、复合材料试件的制备及损伤的检测;二、飞轮模拟结构的设计、制造及损伤的检测;三、飞轮的制造、高速破坏实验及损伤的检测与模式识别;四、复合材料中损伤的有限元模拟及飞轮损伤与断裂的有限元分析。主要创新点及相关研究结果如下: 1.在国内首次用声发射法对高速旋转构件一飞轮进行损伤检测。通过前期的试样损伤检测、模拟结构损伤的检测建立实验资料数据库,并在检测过程中采取一系列有效措施,可使飞轮的损伤信号能顺利地实时采集并被合理区分。 2.在国内首次用人工神经网络的方法对飞轮的损伤模式进行识别。通过反复试验,发现:用含有一个隐层、29个隐层节点的tansig-logsig BP神经网络,用‘trainlm方法经1500步训练后能达均方差为0.0192,具有较高的仿真精度,能很好地识别本研究中的碳纤维/环氧树脂复合材料飞轮各种损伤类型。 3.在国内首次用有限元分析方法中的“单元生死”法模拟复合材料单向层板中的脱层缺陷。以实验数据为目标不断修正模拟模式,最后能合理地模拟带缺陷的试样在承受拉伸载荷时的破坏形式,较为逼真地再现损伤过程 4.玻璃纤维/不饱和聚脂、碳纤维/环氧树脂两种复合材料在承载过程中主要的损伤类型是:基体开裂、纤维/基体界面开裂、纤维断裂。研究发现损伤在载荷远低于实际破坏载荷时就已开始,前期的损伤形式主要为基体开裂,而后期的损伤形式主要是纤维断裂。 5.用声发射特征曲线中的时间.持续时间(Time-Duration)、时间一频率(Time-Frequency)、时间-能量(Time-Energy)、时间-振铃计数(Time-Counts)等关联图能比较清晰地区分长纤维增强的树脂基复合材料的损伤各阶段。研究发现各种损伤所对应的各特征参量有明显变化规律。 6.在纤维剧烈断裂时,出现大“能量”(Energy)、大“振铃计数”(Counts)、长“持续时间”(Duration)的信号,上述参量均较其它两种损伤阶段时的损伤信号的各对应值大一个数量级以上。临界承载状态可以以纤维开始剧烈断裂为判据来确定。更多还原

【Abstract】 Based on the current demand of the high-speed flywheels for storage-energy, this thesis focuses on the systematic study of the damage and fracture of the high-speed flywheels for storage-energy. For the first time the acoustic emission method (AE) is used successfully to measure the damage of the high-speed flywheel for storage-energy. The artificial neural network technology is also used successfully to the damage-type recognition.The following issues have been studied in the research: First, the preparation of the samples of composite material and the detection of their damage; second, the design, manufacture of the simulation-structure of flywheels and the detection of their damage; third, the manufacture of flywheels, the experiment on high spin-speeded flywheels is carried out and detection of their damage; lastly, the finite element method is used on the simulation of the crack propagation of composite material and the analysis of the damage and fracture of flywheels. The major conclusions of the research and the creative ideas are as following.1. It is the first time in China that the acoustic emission technology is used to detect the damage of the structure spinning in high speed—flywheels. Based on the damage-detection on the sample and simulation-structure, a set of effective procedures are applied during the testing process so that the damage signals of flywheels can be successfully collected and distinguished.2. For the first time in China, the artificial neural network technology is used to recognize the damage pattern of flywheels. After repeated testing, we found that the artificial neural network of tatansig-logsig with one hidden layers and 29 hidden units can reach low variance (0.0192) and accurate simulation after it is trained by trainlm-method. It can accurately distinguish the different types of damages of carbon/epoxy composite material used in our testing.3. For the first time in China, the "element birth and death" in finite element method is used to simulate debonding in unidirectional laminates composite. Through continuous adjustment of the model of the simulation, finally the types of damage to defect-sample under tensile load can be simulated reasonably, thus the process of the damage can be recreated realistically.4. During the load-testing, the major types of the damage of two composite materials of glass/unsaturated polyester and carbon/epoxy, are matrix-crack、 crack in the interface of fiber and matrix and fiber-fracture. The damages start when the load is much less than the real fracture-load. The early-stage damages are mostly matrix-crack, while the late-stage damages are mostly fiber-fracture.5. The different damage-stage in polyester-composite reinforced by long fiber can be distinguished by following correlation diagrams: Time-Duration, Time-Frequency, Time-Energy, Time-Counts of the character- curve in acoustic emission test.6. When fiber is fracturing dramatically, high-energy, large-counts and long-duration signals appear. And the energy, counts and duration mentioned above is one magnitude greater than those appeared with signals caused by matrix-fracture and fiber/matrix interface-crack. The critical load is determined when the fiber start to fracture dramatically.7. I designed and developed a static testing structure which simulates perfectly the damages of flywheels.8. Dye penetrant method can be used fairly well to detect the damages of flywheels. For tangential-winding flywheels, which are studied in this paper, the common types of damages are matrix-crack > fiber/matrix interface-crack.9. When steady state propagation and unstable extension of the cracks appear in flywheels, the character-curve of AE- signals have obvious characters.10. When steady state propagation and unstable extension of the cracks appear in glass/unsaturated polyester flywheels, the Felicity’s ratio is less than 0.93.11. We learned from finite element analysis that both maximal radial stress and maximal tangential stress locate close to the middle of flywheels, and the location does not change with the rotate speed. The maximal radial stress and tangential stress increase with the rotation speed. The relation between maximal radial stress and rotate speed, and the relation between tangential stress and rotate speed, fit second order polynomial. According to the finite element analysis: the failure of flywheels appears when radial stress in middle of flywheels exceeds the critical strength of the material, and for glass/unsaturated polyester composite flywheels, the critical rotation speed is 21840r/min.12. The ANSYS analysis of location and style of damages to composite flywheels well agrees with experiment, but the critical rotation speed we obtained through testing is 18655~20865r/min, which is a little smaller than what the ANSYS predicts.The work of this thesis was supported by the project of national natural science foundation of China (19972063).更多还原