大型锻件超声检测方法及信号处理算法研究

发布时间：2018-11-24 14:25

【摘要】：大型锻件是重大装备的关键部件,已广泛应用于电力、航空航天、船舶、重型机械等领域,其质量直接影响到装备的整体水平和运行可靠性,需要对其可能存在的裂纹、气孔和夹杂等缺陷进行检测。超声检测具有穿透能力强、缺陷检测准确率高、灵敏度高、检测成本低、对人及环境无害等优点,适合大型锻件的缺陷检测。目前,国内大型锻件的超声检测大多是手动扫描、人工判读,易出现漏检和误判,检测效率低、可靠性差,需要研制一种大型锻件自动超声检测系统,其中,检测方法和信号处理是系统涉及的关键技术,但是,目前使用的技术都有一定的局限性,本文针对这些技术进行理论和实验研究,主要研究内容如下：1.提出了大型圆筒型锻件及大型中厚钢板的多通道自动超声检测方法。对于大型中厚钢板,主探头组(兼作横边探头组)放置在钢板横向中间部位,纵边探头组放置在钢板两个纵边,检测横边时,钢板不动,主探头左右摆动,检测其它部位时,钢板沿压延方向匀速直线运动,主探头左右摆动,纵边探头不动,形成板边为矩形而板内为正弦或余弦的扫描轨迹；对于大型圆筒型锻件,直探头组和斜探头组分别纵向放置在直径方向上筒壁的两个外侧,检测时,圆筒型锻件绕其轴线原地旋转,两组探头同时沿轴线平移,形成空间螺旋线的扫描轨迹。将超声探头分组放置在不同位置,同时对检测工件按指定路径进行自动全方位扫描,提高了检测效率和可靠性。2.对超声反射回波信号降噪的理论和算法进行了研究,提出了基于小波变换和独立分量分析的超声反射回波信号降噪算法(WICAW)。利用小波变换将原始信号分解,对分解的系数进行独立分量分析,并对分离出的独立分量进行阈值评估,滤除噪声,再通过小波重构得到降噪后的超声信号。仿真和实验结果表明,该算法既不丢失有用信息又提高了信噪比,性能优于小波软阈值降噪算法。3.对缺陷超声信号的特征提取与识别技术进行了研究,提出了基于小波系数聚类和SVM的缺陷超声信号特征提取与识别算法。利用小波变换对降噪的超声回波信号进行分解,然后以概率统计聚类的方法将分解得到的小波系数聚类,计算每个聚类的小波系数能量值并作为SVM分类器的输入特征向量,实现缺陷识别。通过对典型缺陷试块进行检测,实验结果表明,该算法减少了分类器的计算量,提高了小样本缺陷识别的准确率。4.设计了基于以太网的大型锻件自动超声检测系统,搭建了实验平台,进行了实验研究,验证了本文提出的检测方法和信号处理算法的有效性。
[Abstract]:Large forgings are the key parts of heavy equipment, which have been widely used in electric power, aerospace, ship, heavy machinery and other fields. The quality of forgings has a direct impact on the overall level and operational reliability of the equipment. Defects such as pores and inclusions are detected. Ultrasonic detection has the advantages of strong penetration, high accuracy, high sensitivity, low detection cost and harmless to human and environment. It is suitable for defect detection of large forgings. At present, the ultrasonic inspection of large forgings in China is mostly manual scanning, manual interpretation, easy to be missed and misjudged, low detection efficiency and poor reliability, so it is necessary to develop an automatic ultrasonic detection system for large forgings, among which, Detection methods and signal processing are the key technologies involved in the system. However, the technologies used at present all have some limitations. In this paper, theoretical and experimental research on these technologies is carried out. The main research contents are as follows: 1. A multi-channel automatic ultrasonic detection method for large cylindrical forgings and large medium-thick steel plates is presented. For large medium and thick steel plates, the main probe group (both as the horizontal side probe group) is placed in the transverse middle part of the steel plate, and the longitudinal side probe group is placed in the two longitudinal sides of the plate. When the horizontal edge is detected, the plate does not move, the main probe swinging from side to side, and the other parts are detected, The plate moves in a uniform speed straight line along the calendering direction, the main probe swinging from side to side, and the longitudinal side probe does not move, forming the scanning track with rectangular edge and sinusoidal or cosine in the plate. For large cylindrical forgings, the straight probe group and the oblique probe group are placed longitudinally on the two sides of the upper cylinder wall in the diameter direction, respectively. In the detection, the cylindrical forgings rotate around the axis of the forgings, and the two groups of probes move along the axis at the same time. The scanning trajectory that forms the helix in space. The ultrasonic probe is grouped in different positions, and the inspection workpiece is scanned automatically according to the specified path, which improves the efficiency and reliability of the detection. 2. In this paper, the theory and algorithm of ultrasonic echo signal denoising are studied, and the denoising algorithm of ultrasonic reflection echo signal based on wavelet transform and independent component analysis (WICAW).) is proposed. The wavelet transform is used to decompose the original signal, and the coefficients of decomposition are analyzed by independent component analysis, and the separated independent component is evaluated by threshold to filter the noise, and then the ultrasonic signal after noise reduction is obtained by wavelet reconstruction. Simulation and experimental results show that the proposed algorithm not only does not lose useful information but also improves signal-to-noise ratio (SNR), and its performance is better than wavelet soft threshold denoising algorithm. Based on wavelet coefficient clustering and SVM, the feature extraction and recognition algorithm of defect ultrasonic signal is proposed. The wavelet transform is used to decompose the noise reduction ultrasonic echo signal, and then the wavelet coefficients are clustered by the method of probability and statistics, and the wavelet coefficient energy of each cluster is calculated and used as the input eigenvector of the SVM classifier. The defect recognition is realized. The experimental results show that the algorithm reduces the computation of the classifier and improves the accuracy of small sample defect recognition. An automatic ultrasonic detection system for large forgings based on Ethernet is designed. The experimental platform is built and the experimental research is carried out to verify the effectiveness of the detection method and signal processing algorithm proposed in this paper.
【学位授予单位】：天津工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TG316.193;TB559

【参考文献】