摘要
为解决空气处理机组在故障检测过程中难以获得大量带有类标记样本,且故障样本数据标记代价较高的问题,本文结合支持向量机与半监督学习方法,提出了针对空气处理机组故障检测的半监督学习算法.首先利用序列前向选择选出重要的特征作为分类依据,将半监督学习方法引入支持向量机的学习过程中,并使用遗传算法寻找支持向量机的最佳参数.然后选择类标记置信度高的未标记样本加入训练样本集,利用未标记样本中有利于支持向量机的信息,提高学习性能.实验表明,提出的混合算法能够在故障标记样本比较少的情况下达到较高的故障诊断率.
To solve the problems of obtaining a large number of labeled samples and the high cost of marking fault data in the process of fault detection for air processing units, this paper presented a semi-supervised learning algorithm for air handling unit fault detection by combining support vector machines and the semi-supervised learning method. Firstly, the important features were selected by sequence forward selection as the classification basis. The semi-supervised learning method was introduced into the learning process of support vector machines, and the genetic algorithm was used to find the best parameters of support vector machines. Then, the unlabeled samples with high confidence in class markers were added to the training sample set, and the information in the unlabeled samples that was beneficial to the support vector machine was used to improve the learning performance. Experiments show that the hybrid algorithm proposed can achieve a higher fault diagnosis rate with fewer fault marker samples.
作者
钟超文
花君
严珂
陆慧娟
叶敏超
ZHONG Chaowen;HUA Jun;YAN Ke;LU Huijuan;YE Minchao(College of Information Engineering,China Jiliang University,Hangzhou 310018,China)
出处
《中国计量大学学报》
2018年第3期311-316,344,共7页
Journal of China University of Metrology
基金
国家自然科学基金项目(No.61602431)
关键词
故障检测
半监督
遗传算法
支持向量机
特征选择
空气处理机组
fault detection
semi-supervised
genetic algorithm
support vector machine
feature selection
air handling unit
