摘要
利用核独立成分分析(KICA)进行非线性特征提取,然后用最小二乘支持向量机建立故障分类模型。研究表明,不同核函数对模型的性能有很大影响。利用已有核函数构造混合核函数,提出基于混合核函数的KI-CA-LSSVM故障分类方法,并应用到某石化企业的润滑油生产过程。实验结果表明该方法具有很高的分类和泛化能力。
A kernel independent component analysis （KICA） was used for nonlinear feature extraction. A fault classification model based on LSSVM was proposed. The research showed that the performance of a classification model depended on different kernel functions. A combined kernel function could be developed by existing kernel functions. A fault classification method of KICA-LSSVM based on the combined kernel function was also represented, and the method was applied to a lubrication oil process. The implementation shows that the method has a good classification accuracy and generalization capability.
出处
《化工自动化及仪表》
CAS
北大核心
2010年第3期14-18,共5页
Control and Instruments in Chemical Industry
基金
广东省自然科学基金重点项目(07117421)
广东省自然科学基金重点项目(8351009001000002)
