摘要
水泥强度的预测具有多变量、非线性和大时滞特性,因此传统线性回归方法的结果不准确。除此之外,传统的神经网络预测可能对少量样本不够精确。本文建立灰色BP模型,以此来预测水泥的强度。建立一个多因素灰色模型GM(1,N)用于水泥化学成分的样本数据进行预处理,得到新的数据来作为建立预测模型的样本数据,通过BP神经网络建立预测模型。最终通过建立的灰色BP神经网络预测模型来预测28天水泥强度。仿真结果表明:灰色BP预测模型的效果比BP预测的要准确。
The cement strength prediction has characteristics of multi-variable, nonlinearity and large time delay and the traditional linear regression method results in a poor prediction accuracy; in addition, the con- ventional BP neural network may not be accurate enough for a few samples. In this paper, the grey BP model was estab}ished to predict cement strength. Having multi-factor grey model GM (1 ,N) used to preprocess the sample data of cement' s chemical component so as to get new data for the prediction model established through the BP neural network. Simulation results of applying BP network model to predict 28-day cement strength show that, the effect of grey BP forecasting model is more accurate than that of BP.
出处
《化工自动化及仪表》
CAS
2017年第10期925-928,932,共5页
Control and Instruments in Chemical Industry
基金
国家自然科学基金项目(61163051)
云南省教育厅科学研究基金项目(2015Y071)