论文部分内容阅读
为了满足某厂1580热连轧机宽度控制精度需求,提高宽展模型的广泛适用性,利用ANSYS/LS-DYNA有限元软件,对热轧粗轧区立轧--平轧过程进行了模拟.根据模拟数据,系统地分析了轧件宽度、厚度、轧辊直径、立辊侧压量和厚度压下量对“狗骨”宽展、自然宽展和绝对宽展的影响规律.利用模拟数据并结合现场数据构造了FES(finite element simulation)“狗骨”宽展模型和自然宽展模型,并建立了PSO-BP神经网络(粒子群BP神经网络).最后,FES宽展模型与PSO--BP神经网络相结合预报第1、3和5道次的宽展,其预报值与实测值误差在1mm以内的均达到了99%以上,达到了宽度控制的精度要求.
In order to meet the demand of width control precision of 1580 hot strip mill in a factory and improve the wide applicability of wide strip model, the process of vertical rolling and flat rolling in hot rolling rough area was simulated by ANSYS / LS-DYNA finite element software.According to the simulation Data were used to systematically analyze the effects of rolling width, thickness, roll diameter, roll pressure and thickness reduction on the “dog bone” width, natural width and absolute width. Using the simulated data Based on the field data, a finite element simulation (FES) model and a natural wide-spread model of the dogbone are constructed and a PSO-BP neural network (PSO-BPNN) is established.Finally, - BP neural network prediction of the 1st, 3rd and 5th pass wide spread, the forecast value and the measured value error within 1mm reached more than 99%, reaching the accuracy of the width control requirements.