【摘要】 研究了用小波包分析方法从炉口音频信号中提取AOD炉喷溅预报特征信息的方法。采用db10(小波基函数)小波对喷溅发生前的特征信号进行4层小波包分解,结合快速傅里叶变换法及小波尺度谱进行时频特征分析,并研究了其各频带分解信号的能量比例特点。结果表明,喷溅前40s信号的主频值较正常信号有明显降低,0~312Hz与312~625Hz频段信号能量值比例变化显著。而且低频重构信号可以极好地滤除多种现场干扰,说明该时频特征可以作为准确预报喷溅的特征向量。最后,通过实验确定了8个特征向量值并分别与喷溅或正常信号的特征向量进行相关性比较,验证得出相关度0.95可作为喷溅预报的判定阈值。从而实现了喷溅预报特征信号的准确提取并可转化为计算机容易识别的数值特征。更多还原

【Abstract】 Wavelet packet analysis method was used to study the extraction of splash prediction for AOD furnace from audio signal of furnace opening.Using db10 wavelet completed 4layers wavelet packet decomposition from characteristic signal before splash,combined with fast Fourier transform method and wavelet scale spectrum to analyse the time-frequency characteristic,the energy ratio characteristics of the signal frequency band decomposition.The test results show that the frequency value of the 40 ssignal before splash was significantly lower than normal signal,the signal energy ratio between 0-312 Hz and 312-625 Hz frequency band changed significantly.Additionally,the low frequency reconstruction signal was excellent in filtering a variety of scene interference signal,which shows that the time-frequency characteristics can forecast splash feature vector.Finally,8feature vectors was determined by the experiment and compared with the feature vector of splash or normal signals,and verified that relevance 0.95 can be used as a threshold of splash forecast decision.So that realizing the accurate feature extraction of splash prediction;the prediction result can be converted into computer numerical characteristics easily recognized.更多还原