带钢卷取温度高精度预报的遗传神经网络方法

Genetic-neural network method for high-accuracy prediction of coiling temperature of hot rolled strip

热轧带钢卷取温度是影响成品带钢性能指标的重要工艺参数之一,其层流冷却控制系统具有高度非线性。影响卷取温度的因素多而且复杂,采用传统的温度预报模型难以达到较高的精度要求。为了满足卷取温度高精度的要求,提出了一种基于数据挖掘技术的遗传神经网络方法。充分发挥数据挖掘的关联分析能力、神经网络的泛化映射能力和遗传算法的全局搜索能力,将三者结合起来,建立了卷取温度预测模型。运用实际现场数据进行测试表明:它能准确地预报卷取温度,具有在线应用的前景。

Hot strip coiling temperature is one of the important parameters of performance index in hot rolled strip,and its control system of highly nonlinearity.Traditional coiling temperature prediction model is difficult to get the high precision of prediction because of multi-factors.In order to satisfy the demands of high precision of the coiling temperature,a genetic-neural network method to predict coiling temperature based on data mining is put forward.To make full use of the association-analysis capability of data mining,the approximation capability of neural network and the liability to fall into the global optimum solution of genetic algorithm,a forecast model is established by using the method above.Checked with actual production data,the result indicates that the method could predict precisely the strip coiling temperature,and be used on-line afterwards.