非线性量化小脑模型神经网络在溶出循环母液配比模型中的应用

Application of Nonlinear Quantization CMAC to the Proportion Mixture Model of Digesting Recycled Liquor

采用自适应算法对小脑模型神经网络的概念映射进行设计,提出了非线性量化小脑模型神经网络算法,提高小脑模型神经网络的计算速度和精度以满足复杂动态环境下的非线性实时控制的需要.提出了基于非线性量化小脑模型神经网络的溶出预脱硅系统时间序列预测模型,用于准确实时地预测循环母液加入量,在此基础上进行循环母液投放措施优化.试验说明了该模型在对化工软计算的预测精度和快速性上具有明显的优越性,本模型已应用于某氧化铝厂工艺优化系统中,动态调节循环母液投放量以节省原料.

The adaptive algorithm is adopted to design the concept mapping of cerebella model articulation controller （CMAC） neural network. The nonlinear quantization CMAC is presented to improve the speed and accuracy of calculation to meet the complex and dynamic demand under the nonlinear and realtime controlling environment. A proportion mixture time series prediction model of digesting recycled liquor by the nonlinear quantization CMAC is presented to forecast the quantity of recycled liquor accurately and fast. The quantity of recycled liquor is optimized on the basis of the model. The test shows that the accuracy and speed of the time series prediction model has obvious advantages. The model has been applied to certain alumina olant to ootimize dvnamicallv the recycled liquor quantitv to save the raw materials.