基于模糊神经网络的焊缝缺陷识别的研究

焊缝缺陷在X射线设备下成像转成数字图像后，分析其图像的特点，进行缺陷的定位与边缘检测，结合人工识别焊缝缺陷的经验选取对焊缝缺陷分类影响因子较大的特征参数。用模糊集合的概念描述特征参数，建立特征参数的模糊规则库，构建以模糊化后的特征参数为输入层，以模糊规则为隐含层，缺陷预知识别分类为输出的模糊神经网络模型。分析实验结果，成功定位缺陷在数字图像中的大概位置与边缘检测。该方法提高了集合交叉较大的焊缝缺陷的识别率，能有效地对缺陷进行识别分类。

Research on Weld Defects Distinguishing Based on Fuzzy Neural Networks

After weld defection is converted to digital images under the X-ray equipment,analyze the characteristics of the images for lo-cating defects and detecting edges,select the characteristic parameters,which have larger impact on weld defect classification combined with the experience of artificial identification. Describe the characteristic parameters in the concept of fuzzy set,and establish the fuzzy rule base of characteristic parameters,building the fuzzy neural network model with the fuzzy characteristic parameters as the input layer, and fuzzy rules as the implied layer,and the identification and classification of defecting prediction as output. Analysis of test results show it can successfully locate the defects about position and edge detection in digital images. The method improves the recognition rate of the larger crossed set of weld defection,which can identify the classification effectively.