基于FOA-SVM的超声信号端点检测
发布时间：2018-10-30 06:29
【摘要】：在超声缺陷识别系统中,端点检测是确保缺陷准确识别的重要环节。为提高在实际探伤过程中端点检测的准确率,提出一种以果蝇算法优化支持向量机的端点检测方法。针对超声检测信号的特点,采用小波包变换提取反映该信号性质的特征向量。鉴于传统方法检出率不高及支持向量机(SVM)参数难确定的问题,利用果蝇算法(FOA)优化SVM的惩罚子和核参数,提高支持向量机建模准确度。试验结果表明:FOA-SVM模型的平均检出率达到97.5%,端点检测效果明显优于传统的双门限法、普通SVM模型和GA-SVM模型。
[Abstract]:In ultrasonic defect recognition system, endpoint detection is an important step to ensure the accurate identification of defects. In order to improve the accuracy of endpoint detection in the process of practical flaw detection, an endpoint detection method using Drosophila algorithm to optimize support vector machine is proposed. According to the characteristics of ultrasonic detection signal, wavelet packet transform is used to extract the feature vector which reflects the character of the signal. Because the detection rate of traditional methods is not high and the (SVM) parameters of SVM are difficult to determine, (FOA) algorithm is used to optimize the penalty and kernel parameters of SVM to improve the accuracy of SVM modeling. The experimental results show that the average detection rate of FOA-SVM model is 97.50.The result of endpoint detection is obviously superior to that of traditional double-threshold method, ordinary SVM model and GA-SVM model.
【作者单位】： 华北电力大学自动化系;
【分类号】：TP18;TB559
，
本文编号：2299192
[Abstract]:In ultrasonic defect recognition system, endpoint detection is an important step to ensure the accurate identification of defects. In order to improve the accuracy of endpoint detection in the process of practical flaw detection, an endpoint detection method using Drosophila algorithm to optimize support vector machine is proposed. According to the characteristics of ultrasonic detection signal, wavelet packet transform is used to extract the feature vector which reflects the character of the signal. Because the detection rate of traditional methods is not high and the (SVM) parameters of SVM are difficult to determine, (FOA) algorithm is used to optimize the penalty and kernel parameters of SVM to improve the accuracy of SVM modeling. The experimental results show that the average detection rate of FOA-SVM model is 97.50.The result of endpoint detection is obviously superior to that of traditional double-threshold method, ordinary SVM model and GA-SVM model.
【作者单位】： 华北电力大学自动化系;
【分类号】：TP18;TB559
，
本文编号：2299192
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2299192.html
