Development of a Novel Soft Sensor with Long Short-Term Memory Network and Normalized Mutual Information Feature Selection

Abstract

In this paper, a novel soft sensor is developed by combining long short-term memory (LSTM) network with normalized mutual information feature selection (NMIFS). In the proposed algorithm, LSTM is designed to handle time series with high nonlinearity and dynamics of industrial processes. NMIFS is conducted to perform the input variable selection for LSTM to simplify the excessive complexity of the model. The developed soft sensor combines the excellent dynamic modelling of LSTM and precise variable selection of NMIFS. Simulations on two actual production datasets are used to demonstrate the performance of the proposed algorithm. The developed soft sensor could precisely predict the objective variables and has better performance than other methods.

Input Selection Methods for Soft Sensor Design: A Survey

“…Industrial SSs are generally designed with Artificial Neural Networks (ANN), mainly with the Multi-Layer Perceptron (MLP) structure [7,11,63,64], Convolutional Neural Networks (CNN) [65,66], Generative Adversarial Networks (GAN) [67,68], Radial Basis Networks [69], Wavelet Networks [70], Hinging Hyperplanes [71], Deep Belief Networks [72,73], Stacked Autoencoders [74], Long Short-Term Memory Networks [75], Support Vector Regression [76], Gaussian Processes Regression [77], Extreme Learning Machines [78], Fuzzy Systems and Neuro-Fuzzy Systems [79,80], just to mention a few [5,[81][82][83][84][85][86][87][88]. In some cases, the designer can choose to create more linear or nonlinear models for the same system, each one for a different working point, instead of a single one covering all of the system dynamics.…”