连铸结晶器拉坯阻力信号特征提取及识别

在板坯连铸生产过程中,为实现结晶器拉坯阻力信号的自动识别,采用小波包变换技术,对MDF信号进行降噪预处理,并提取结晶器在不同状态下的阻力信号特征。针对故障样本相对较少的情况,采用虚拟样本技术对其进行扩充,以实现样本的平衡性。最后,采用支持向量机方法,对提取的结晶器拉坯阻力信号特征进行识别研究。实验结果表明,采用小波包变换技术提取的阻力信号特征明显,支持向量机对提取的阻力特征识别精度高、实时性强。

Module Friction Feature Extraction and Recognition

To realize automatic recognition of the Mould Friction(MDF) signal,in slab continuous casting,the wavelet packet decomposition was applied to noise reduction and feature extraction,under different working status of the mould.Generally,fault sample sets are fairly small.According to this condition,the samples were augmented using virtual sample technology that can make sample sets balance.Then,a new machine-learning method––Support Vector Machine(SVM) was applied to recognize the feature of friction signal.The results show that SVM can correctly identify MDF features.