Fuzzy Neural Network Control For Shape Of UC Rolling Mill

Our country's steel rolling industry has developed at full speed in recent years, and our country has become one of the countries whose output of steel are most in the world. But currently sizable disparity exists in our country's steel production as for steel rolling technology compared with international advanced level. Shape precision is one main quality index of cold rolled strip and is also an important factor which determines its competitiveness in the market. The urgent demand of improving the quality of cold rolled strip has made shape control be an important problem that faces our country's steel enterprises. The shape is an important quality criterion of cold rolled strip. In order to improve product quality and productivity, shape controlling become increasingly an important subject in steel enterprise.UC (Universal Crown) rolling mill is a strong shape control which has many means. Such as: inclining roller, bending roller and moving the intermediate roller. The shape control system of UC rolling mill is a multivariable control system. In control strategy, it is simplified into several single variable then control it. On decoupling control, the global shape control system can decompose three loop on line, which are supporting roller is inclined, working roller is bending and intermediate roller is bending. By model recognition and model decomposition, shape detection information provides corresponding control variable for each loop.The shape control system of UC rolling mill has the characteristics of inertia and lag, and the mathematical model of the control loop of bending the intermediate rolls has time-variability and uncertainty. The routine PID algorithm can't be responsive to the transformation of system parameters rapidly and accurately. The precision of the shape can't be assured easily. To realize accurate control of bending of the intermediate rolls, a fuzzy neural network is presented to control the loops of bending of the intermediate rolls in which fuzzy logic control and neural network are combined. To avoid the minimal value and influence of the initial weight on the final solution, the genetic algorithms gradually narrow the searching range. Then BP network to solve precision. It adopts a new combined method of genetic algorithm and BP algorithm with momentum factor. After taking result to compare with only BP algorithm in the train network, it shows that the control effect is excellent , with small ultra regulated and good robustness, reaching the goal of the intelligence control for shape on line.