遗传神经网络对RPC流动度预估的研究

Prediction of RPC fluidity using back-propagation network based on genetic algorithm

流动度是RPC配制的一个关键指标,它直接反映了其工作性的优劣,但其影响因素复杂,难以用统一的数值关系直接描述RPC流动度与其影响因素的量化关系,目前还没有合适的计算方法,为此,提出控制RPC流动度的数值方法,即引入遗传神经网络对RPC的流动度进行预测控制。在建立网络模型后,选取适当的参数,进行训练仿真分析。结果表明,该遗传神经网络模型是有效的,对RPC的流动度预测有较高的精度和稳定的预测结果,与单纯的BP神经网络模型相比,具有精度高、训练速度快、工作性能稳定等优点。

To RPC, the fluidity was usually a key index, because it directly reflected the work-ability. The effect factors were complex, however, the quantity relation between the RPC fluidity and its effect factors could not be expressed with a uniform numeric formula. There was no effective method for this problem at present. In accordance, a numeric method which used for controlling the RPC fluidity was advanced. The neural network model based on genetic algorithm was introduced for a aptitude method consequently. After founding the network model and selecting the appropriate parameters, the network model was trained and simulated. The results indicate that the neural network model based on genetic algorithm for controlling the RPC fluidity is effective, and the precision is high enough. Comparing with the single BP neural network, it has higher precision, faster train speed, more steady performance.