Adjusting neural networks for accurate control model tuning

In this paper, we propose adjusting neural networks (AJNNs), which are an extended model of conventional multilayered neural networks (CNNs), for accurate model tuning with small tuning numbers. The AJNN consists of two multilayered neural networks, namely, a CNN and an error calculation neural network (ECNN) which is added in parallel to the CNN. The ECNN calculates the output error of the CNN and subtracts it from the output of the CNN, to obtain accurate tuning values. A training method for the AJNN is also proposed, where the modified back-propagation developed for reducing the error of the AJNN and the conventional back-propagation for decreasing the output of the ECNN, are applied to the AJNN alternately. The AJNN is evaluated with model tuning of temperature control for reheating furnace plants and is demonstrated to be effective to improve the accuracy of tuning and decrease tuning numbers