【作者】 宋小春；

【导师】 康宜华； 武新军；

【作者基本信息】 华中科技大学 ， 机械制造及其自动化， 2005， 博士

【摘要】 在目前电力供应紧张的情况下,作为供电大户的火力发电厂每年提供了全国用电量的63%。但是,由于制造质量“先天不足”和运行维护“后天失调”,我国大型电站锅炉管道爆漏停用事故严重影响着火力发电厂机组的安全经济运行,因此,研究快速高效的无损检测方法,并开发相应无损检测装备,对电力企业安全生产具有十分重要的现实意义。本学位论文提出了主磁通超声波融合检测水冷壁管壁厚和漏磁检测水冷壁管缺陷的方法,研制出一套机械化无损检测装备。首先,针对竖直水冷壁的检测要求和现场特点,在完成爬行装置结构设计的基础上,研制出水冷壁管检测爬行装置、磁性检测传感器、超声波探头定位抬放装置、远距离数据采集系统以及信号分析处理软件。实现了管壁大面积快速扫查和定量化评价以及检测过程的自动化,提高了检测效率和可靠性。其次,根据局部缺陷、随机干扰、内螺纹等噪声与壁厚腐蚀信号的不同特征,讨论了基于小波分析的磁性检测信号预处理方法。研究了基于奇对称小波变换的模极大值信号识别方法、小波分解与重构去噪方法以及自适应小波滤波器方法,实现了多种背景信号的分类和消除,从而有效提取出壁厚减薄信号,获得了较高的壁厚信息融合精度。第三,从提高检测效率、降低误判和漏判率的实际出发,提出了主磁通与超声测厚融合检测壁厚的方法,通过主磁通普查和超声波定点测量实现壁厚的精确评定。在比较分析了多项式函数、指数函数和对数函数拟合精度的基础上,选择多项式函数拟合对管壁厚度与主磁通检测信号峰峰值之间的关系建模,运用超声波测厚值对模型参数进行现场标定,以实现壁厚的定量化评定。实验结果表明该方法的评价精度优于统计关系模型,更适合实时检测与评价。由于不需要太多的实验数据,因此该方法具有更好的实用性。最后,在分析现有反演方法特点的基础上,将小波多尺度分析的思想引入到漏磁非线性反问题求解中,研究了漏磁检测信号的小波多尺度反演方法,分别建立了漏磁信号的小波多尺度迭代反演模型和小波神经网络的正演、反演模型,讨论了漏磁反问题多尺度求解的一般过程。反演结果表明,小波多尺度反演比广义线性反演的收敛速度快,而且迭代收敛性好,反演精度高; 而基于小波神经网络的反演方法也具有一定的可靠性,其中,小波基函数和分解尺度数的选择对网络的反演性能有直接影响。另外,对小波变换后,内外表面缺陷信号的不同特征进行了分析,初步研究了内外伤分类的小波方法。更多还原

【Abstract】 In the strained circumstance of the current electricity supply, as a big door of power supply, the thermal power plant annually provides about 63% electricity that the whole nation used every year. However, because of the poor manufacturing quantity and the unscientific management, the safety and cost-effective of the thermal power plant is influenced seriously by the tube explosion and leakage. In order to decrease and prevent the incidents happening,and to improve the boiler tube inspection quality,it is very important for us to search for an quantitative inspection method and develop an effective apparatus for testing boiler tube. In the light of the status that there exists faults in water-wall tube and there are no effective inspection and evaluation means, in this dissertation, a novel fusion testing method using the main magnetic flux (MMF) technique and ultrasonic method is studied, and an automatic testing device is developed. The main research contents, results and new ideas are as follows: First, according to the spot inspection environment of the erecting water-wall, a water-wall inspection apparatus, which includes the adhesion and crawling mechanism and safety,the magnetic transducer,a fixer for ultrasonic probe, remote data acquisition, the control system and the signal processing software, is designed. Comparing with other inspection methods for water-wall tube, this system not only have the fast examination speed and high sensitivity, but also can realize automatic examination and quantitative evaluation. And it can increase the examination efficiency and the reliability, guarantee the boiler for the thermal power plant running safely and cost-effectively, and have considerable economic benefit and the extensive market applied foreground. Second, because of the different characteristics of the local defect, system noise, the inside screw thread and the wall loss corrosion signal, a signal pre-processing method based on wavelet analysis is discussed. The signal processing method based on wavelet transform module maximum value, the wavelet decomposition and reconstruction denoise and an adaptive wavelet noise canceller is used to cancel the various noises. And wall loss signal is picked up from the original. Third, in order to meet the need of fusion inspection and avoid the wrong evaluation, a new method combined the MMF technique with ultrasonic method is proposed, and the magnetic technique is used to do full inspection and locate the flaws, and the ultrasonic is employed to implement further quantitative inspection accurately. After comparing the precision of the polynomial, exponential and logarithmic function, the first function is selected to fit the relations between the thickness and the peak value of MMF signal. Thus, the data measured by ultrasonic probe is used to calibrate the model parameters, and the defect profile can be evaluated quickly. The experimental result demonstrates that the model used in this system has better accuracy than the statistics relation model obviously, and it is suitable for defect evaluation real-time. Moreover, it is unnecessary to have much more experimental data for the curve fitting technology, so it has the better practicability than the other methods. Finally, after analyzing of the characteristics of inversion method existed today, the wavelet multi-resolution analysis method is employed to solve the nonlinear inversion problem for MFL, and the theoretical framework about multi-resolution solving for MFL inversion problem is discussed. In addition, the iterative model of multi-resolution inversion, wavelet neural network (WNN) forward and inversion model are set up. And the choice of wavelet basis function, parameters initialization, the number of hidden layer nodes and parameters regulating are discussed, the learning algorithm for wavelet network is also studied. The inversion result shows that the WNN has certain reliability to inverse the MFL signal, and it is very important to choose the wavelet basis function and resolutions for WNN. Thus, some methods to improve the inversion precision are discussed in this dissertation. In addition, the difference between the outside defect and the inside defect is studied, and the flaw recognition and classification method is discussed based on wavelet analysis.更多还原