IEEE Transactions on Instrumentation and Measurement · 2019 · 73 citations · 36 references
ComputingComputer ScienceIndustrial EngineeringMachine LearningOptimizationSensorsFeature SelectionNeural ModelingNeural NetworksOptimization For Machine LearningRepresentation LearningSoft SensorsInput VariablesMultilayer PerceptronNeural Network ModelingNeural Network PredictionPerceptronPerformance PredictionSoft SensingSoft Sensor Modeling
In nonlinear industrial processes, some product qualities or key variables are usually difficult to measure online automatically due to the lack of sensors. In this work, a novel data-driven soft sensor technology based on a multilayer perceptron (MLP) neural network with a double least absolute shrinkage and selection operator (dLASSO) approach, named dLASSO-MLP, is developed with a two-step procedure. First, an MLP model is constructed through the process data set. Second, a dLASSO algorithm is integrated into the model to solve two redundancy problems, i.e., the input variable redundancy and the model structure redundancy. The proposed method not only selects input variables that are most sensitive to the model, but also simplifies the MLP structure by deleting redundant hidden nodes to avoid the model overfitting. In addition, the method is validated by data from simulation examples as well as an industrial application. Compared to other neural network methods, the proposed method requires fewer neurons and presents better prediction performance.
36
Dropout: a simple way to prevent neural networks from overfitting
Nitish Srivastava, Geoffrey E. Hinton, Alex Krizhevsky et al. · 2014 · 34.3K citations
Model Selection and Multimodel Inference
Technometrics · 2003 · 5.6K citations
Data-driven soft sensor development based on deep learning technique
Chao Shang, Fan Yang, Dexian Huang et al. · Journal of Process Control · 2014 · 542 citations