摘要
针对卷积神经网络(Convolutional neural network,CNN)模型在对工业数值型数据分类方面存在特征使用不充分、模型分类性能不佳等问题,提出了一种基于自适应卷积核的改进CNN(Improved CNN based on adaptive convolution kernel, ACK-ICNN)算法。该算法为了增加特征的重复使用率,构建了一种多尺度卷积核的模型结构,通过融合处理卷积核提取的不同特征来实现,增强了模型的适应能力;为了进一步提升该算法的性能,利用网格搜索算法自适应选取CNN中最优的卷积核大小,使得模型能够提取出最优的特征。采用TE过程的故障数据对其进行测试,并与支持向量机、极限学习机、最近邻等典型的数据驱动方法进行对比,测试结果表明,该算法能有效提升各类故障的分类精度。
To tackle the problems of insufficient use of features and poor performance of model classificationof convolutional neural network(CNN) model in the classification of industrial numerical data, an improved CNN based on adaptive convolution kernel(ACK-ICNN) algorithm is proposed. In the algorithm, a multi-scale convolution kernel model structure is constructed in order to increase the reuse rate of features, which is realized by fusion processing of different features extracted from convolution kernel to enhance the adaptability of the model. To further improve the performance of the algorithm, the grid search methodis used to select the optimal convolution kernel size in CNN in adaptive way, so that the model can extract the optimal features.The fault data of the TE process is used to test the performance of the proposed algorithm, accompanied with comparison with typical data-driven methods such assupport vector machine, extreme learning machine and nearest neighbor. The test results show that the proposed algorithm is helpful to largely improve the classification accuracy of various faults.
作者
程诚
任佳
CHENG Cheng;REN Jia(Faculty of Mechanical Engineering & Automation,Zhejiang Sci-Tech University,Hangzhou 310018,China)
出处
《浙江理工大学学报（自然科学版）》
2019年第5期657-664,共8页
Journal of Zhejiang Sci-Tech University(Natural Sciences)
基金
国家自然科学基金项目(61203177)
浙江省自然科学基金项目(LY17F030024)
关键词
卷积神经网络
数值型数据
自适应卷积核
网格搜索
convolutional neural network(CNN)
numerical data
adaptive convolution kernel
grid search
