A new, self-learning optimization algorithm improves the tube thickened end control of the atrac® smart control for stretch-reducing mills.
In the case of frequently unavoidable larger wall thickness deviations in the area of the shell ends, the setting of the thickened end control must almost always be actually adapted. In the past, the operator had to perform this task until the automatically adapting thickened end control was introduced. This operation mode carried out the operator's working method, using data storage for the tube sequence and a defined optimization algorithm to achieve a better result that was above all independent of the operator.
In fact, the task of adapting the thickened end control to the current rolling situation is a complex optimization task: by selectively varying several interdependent parameters, an optimum (shortest possible end losses) must be achieved as quickly as possible under specified boundary conditions.
A new improved optimization method solves this task perfectly. For this purpose, the currently valid relation of the parameters (= the formulation of their dependence on each other) is used to execute the variation of several parameters simultaneously, instead of separately and successively, thus minimizing the number of optimization steps. The relation of the parameters is first specified and then automatically optimized with help of practical results. The relation (depending on the rolling situation) is stored and reused together with successful parameter combinations. In this way, a faster, self-learning optimization algorithm for multiple optimization parameters is implemented, another step in the permanent further development of the atrac® system.