Extreme Learning Machine Soft-Sensor Model With Different Activation Functions on Grinding Process Optimized by Improved Black Hole Algorithm

W Xie,M W Guo,S S Guo,C Xing,J S Wang,L F Zhu

doi:10.1109/access.2020.2970429

Abstract

Aiming at predicting the key economic and technical indicators (Granularity and Ore content)in the grinding production process, the extreme learning machine (ELM) soft-sensor model with different activation functions on grinding process optimized by improved black hole (BH) algorithm was proposed. Based on the selected auxiliary variables for the soft-sensor model of the grinding, the KPCA method is used to reduce the dimensionality of the high-dimensional data. In order to investigate the influence of different activation functions on the prediction accuracy of the ELM model, seven continuous function (Arctan, Hardim, Morlet, ReLu, Sigma, Sin and Tanh) are used as the activation function of the ELM neural network to establish soft-sensor models respectively. For the shortcomings that ELM model weights and offset values are arbitrarily given so as to result in the low prediction accuracy and low reliability, an improved BH algorithm based on the golden sine operator and the Levy flight operator (GSLBH) was used to optimize the parameters of the ELM neural network. Simulation results show that the model has better generalization and prediction accuracy, and can meet the real-time control requirements of the grinding process.

Highlights

In the process of processing iron ore, the grinding and classifying process is a very important step in the industrial process of the concentrator
The global search ability of the pure black hole (BH) algorithm is poor, and it is easy to fall into the local optimal solution, which leads to the poor fitting effect of some BH-extreme learning machine (ELM), and some results have worse prediction accuracy than the original ELM model, such as the black hole algorithm (BH-ELM) model where the activation function is sin
Some phenomena can be observed, the global search ability of the pure BH algorithm is poor, and it is easy to fall into the local optimal solution, which results in many BH-ELM fitting effects are poor, and the prediction accuracy is even worse than the original ELM model, such as the BH-ELM model where the activation function are sin and thah

Summary

INTRODUCTION

In the process of processing iron ore, the grinding and classifying process is a very important step in the industrial process of the concentrator. When the activation function g(x) is infinitely differentiable, the parameters of the ELM model do not need to be fully adjusted, w and b can be randomly selected before training, and remain unchanged during the training process, and the connection weight β between the hidden layer and the output layer can be obtained by solving the least squares solution of the following equations, that is to say: min H β − T (19). W. Xie et al.: ELM Soft-Sensor Model With Different Activation Functions on Grinding Process Optimized by Improved BH Algorithm TABLE 6.

SIMULATION EXPERIMENT AND RESULT ANALYSIS

Findings

CONCLUSION