摘要
为了识别干扰信号,研究分析了电磁信号特征,构建了基于神经网络的识别方法。该方法利用D-S证据理论进行特征融合,提高特征提取精度;通过改进的K均值算法聚类,并结合卷积神经网络进行信号识别。改进K均值-卷积神经网络的平均识别准确率达99.3%,F1-measure为98.1%,平均识别时间为0.24 s,识别准确性和速度均较高。在信噪比为10 dB时,对直采信号和正交下变频接收机采集信号的识别准确率分别为99.8%和97.8%,优于其他算法。上述结果表明,提出的井口火焰探测器报警信号有源干扰识别方法在识别准确率和速度上均具有显著优势。
In order to achieve the recognition of interference signals,the characteristics of electromagnetic signals were studied and analyzed,and a neural network-based recognition method was constructed.This method utilizes D-S evidence theory for feature fusion to enhance the accuracy of feature extraction;Clustering is performed using an improved K-means algorithm and combined with convolutional neural networks for signal recognition.The average recognition accuracy of the improved K-means convolutional neural network reached 99.3%,F1-measure was 98.1%,and the average recognition time was 0.24 seconds.The recognition accuracy and speed were both high.When the signal-to-noise ratio is 10 dB,the recognition accuracy of the direct acquisition signal and the orthogonal down conversion receiver acquisition signal are 99.8%and 97.8%,respectively,which is superior to other algorithms.The above results indicate that the active interference recognition of the alarm signal of the wellhead flame detector proposed in the study has significant advantages in recognition accuracy and speed.
作者
穆卫巍
张瑜春
王利平
王磊
MU Weiwei;ZHANG Yuchun;WANG Liping;WANG Lei(PetroChina Changqing Oilfield Changbei Operating Company,Xi’an 710000,China)
出处
《电子设计工程》
2025年第21期58-62,68,共6页
Electronic Design Engineering
基金
2023年长庆油田公司科研项目(2023DJ0714)
2022年长庆油田公司科研项目(2022CBKJ005)。
