Abstract：The coating weight control of continuous hot dip galvanizing has the characteristics of high-dimensional, nonlinear and time-varying. The traditional mathematics model and shallow neural network have limited ability to predict the relationship of complex variables, while the deep learning model adopts a multi-layer nonlinear network structure and good approximation of complex functional relations can be achieved. Kernel learning is powerful tool to deal with complex nonlinear data. So, a deep map multiple kernel learning algorithm based on multi-layer information of deep neural network was proposed. By making nonlinear product of this deep map kernel and multi-scale Gaussian base kernel, new improved kernels with high expression ability were obtains, which contained the deep feature information of data. A large number of benchmark data sets and actual industrial data show that the algorithm can achieve higher precision prediction of coating weight and improves the classification accuracy and generalization ability by combining the advantages of deep learning and multiple kernel learning, solves the control difficulties of strong nonlinearity, time-varying large lag and multivariable in the galvanizing process, and realizes higher precision prediction of coating thickness. The average absolute prediction error of coating weight is reduced from 3.04 g/m2 to 1.22 g/m2.
张岩, 翟景峰, 王军生, 孙瑞琪. 基于深度映射多核的热镀锌镀层厚度预测[J]. 中国冶金, 2023, 33(8): 112-117.
ZHANG Yan, ZHAI Jingfeng, WANG Junsheng, SUN Ruiqi. Coating weight prediction for hot dip galvanizing based on deep map multiple kernel[J]. China Metallurgy, 2023, 33(8): 112-117.
AZIMI S M,BRITZ D,ENGSTLER M,et al. Advanced steel microstructural classification by deep learning methods[J]. Scientific Reports,2018,8(1):2128.
[8]
GEORGIOS N K. SVM kernel based on particle swarm optimized vector and Bayesian optimized SVM in atmospheric particulate matter forecasting[J]. Applied Soft Computing,2020,93:106410.
[9]
PORIA S,PENG H,HUSSAIN A, et al. Ensemble application of convolutional neural networks and multiple kernel learning for multimodal sentiment analysis[J]. Neurocomputing,2017,261:217.
[10]
AGETHEN S, HSU W H. Deep multi-kernel convolutional LSTM networks and an attention-based mechanism for videos[J]. IEEE Transactions on Multimedia,2020,22(3):819.
[11]
SONGY H,THIAGARAJANZ J J,SATTIGERI P,et al. Optimizing kernel machines using deep learning[J]. IEEE Transactions on Neural Network and Learning System,2018,29(11):5528.
[12]
GRETTON A,BOUSQUET O,SMOLA A,et al. Measuring statistical dependence with Hilbert-Schmidt norms[C]//International Conference on Algorithmic Learning Theory. Berlin:Springer,2005:63.
AIOLLI F,DONINI M. Easymkl:A scalable multiple kernel learning algorithm[J]. Neurocomputing,2015,169:215.
[21]
LAURIOLA I,POLATO M,AIOLLI F. The minimum effort maximum output principle applied to multiple kernel learning[C]//ESANN 2018 Proceedings. Bruges,Belgium:[s.n.],2018:1121.
2023, Vol. 33 
