Application of variable-metric chaos optimization neural network in predicting slab surface temperature of the continuous casting

A slab surface temperature prediction model of the continuous casting based on the variable-metric chaos optimization neural network is presented to solve the problem which the slab surface temperatures can not be measured continuously directly for plentiful inhalator, water film and ferric oxide on the slab surface in the secondary cooling zone. The model is shown to fit the actual data precisely and to overcome several disadvantages of the conventional BP neural networks, namely: slow convergence, low accuracy and difficulty in finding the global optimum. A series of tests have been conducted based on the inputs of the continuous casting in a steel factory. It has been shown that the error is less than 1% between the predicted surface temperatures with the model and the actual temperatures, and the error is less than 2% between the predicted slab thicknesses with the model and the actual slab thicknesses. The model has yielded highly desirable results.