中文摘要：

为提高动态过程异常模式的监控效率，提出基于多特征混合与多分类支持向量机的动态过程质量异常模式识别模型。采用离散小波变换提取原始数据的低频近似系数和重构数据特征；抽取重构数据的形状特征并与低频近似系数进行混合，形成质量模式的混合特征；采用粒子群优化的多分类支持向量机进行异常模式识别。仿真实验表明，所提出的识别模型比采用单一类型特征或融合特征的整体识别精度均有显著提高，且大大降低了模型训练时间。

英文摘要：

To improve the monitoring efficiency for abnormal patterns in dynamic process, a novel quality abnormal pattern recognition model based on multi-features hybrid with Multi-class Support Vector Machine（MSVM）was pro- posed. The low frequency approximation coefficients and the wavelet reconstruction data were extracted from origi- nal data by discrete wavelet transform. The shape features extracted from wavelet reconstruction data were com- bined with the low frequency approximation coefficients to form the hybrid feature of quality pattern. The quality abnormal pattern was recognized by multi-class support vector machine optimized particle swarm optimization. Com- pared with the recognition model with a single class of feature and fusion feature, the simulation results illustrated that the recognition accuracy and training time of the proposed model had a remarkable improvement.