针对复杂工业过程样本集中的类不平衡、样本标注代价昂贵和样本孤点的问题,研究基于委员会投票选择(MQBC)和代价敏感支持向量机(CS-SVM)的故障检测方法.给出未标注样本信息度的定义,提出改进的委员会投票选择算法.主动代价敏感学习通过MQBC选择信息度高的未标注样本对其标注并添加到训练集.CS-SVM将不同类样本的误分类赋予不同的误分类代价,从而提高CS-SVM的故障检测率.最后,以铜闪速熔炼过程为例,实验结果验证了所提出方法的有效性. In order to solve the problem of class imbalance, sample annotation cost and sample loneliness in complex industrial process sample sets, a fault detection method based on commission voting (MQBC) and cost-sensitive support vector machine (CS-SVM) The definition of sample information degree and the improved voter selection algorithm are put forward.The active cost-sensitive learning selects and adds to the training set the unseen sample with high informationality by MQBC.CS-SVM gives different classification of misclassification of different types of samples Misclassify the cost, so as to improve the fault detection rate of CS-SVM.Finally, taking the copper flash smelting process as an example, the experimental results verify the effectiveness of the proposed method.