A Strip Thickness Prediction Algorithm Using Extreme Learning Machine with Improved PSO
Li Zhang · Journal of Information and Computational Science · 2014
The accuracy of strip thickness is important to measure the quality of the finished product. This paper proposes a strip thickness prediction algorithm using Extreme Learning Machine (ELM) with Improved Particle Swarm Optimization (IPSO). Due to the randomly selecting of the input weights and hidden biases in ELM, the generalization performance may be influenced as well as result in illcondition problem. In this paper, IPSO is used to choose the input weights and hidden biases, and Moore-Penrose generalized inverse is applied to analytically determine the output weights. IPSO applies jumping behavior to reduce the impact of the bad particles, and considers not only the RMSE but also the 2-norm condition number of the hidden output matrix, so the output weights with smaller norm are obtained. Finally, the optimized ELM is applied to predicting the strip thickness. The results of the comparison experiment show that the proposed algorithm reduces the prediction error, and improves the prediction accuracy and fitting degree.