摘要
为了缩短印制线路板产业(PCB)废水处理的调试周期,控制化学试剂用量,节约能源,采用反向传播(BP)神经网络训练并建立了线路板废水处理的神经网络模型。以混凝沉淀水处理工艺的5个主要影响因素作为输入层参数,以出水水质指标作为输出层参数,设置单隐含层。将10组调试数据作为训练样本,网络运行得到的系统误差为0.000 999 96,将3组调试数据作为预测样本,网络预测值与实际数据值吻合较好。说明该网络具有较好的泛化能力,能很好地对在不同水质参数下线路板废水的处理效果进行预测,在达到所要求的水处理效果的基础上,降低进水水量及水质变化系数较大等不利因素的影响,合理投加化学试剂,使水处理系统在最优的状态下安全、稳定、低成本及高效率运行。
In order to shorten the debugging cycle of wastewater treatment in Printed Circuit Board （ PCB ） production, control the chemical reagent consumption and save energy sources, Back Propagation （BP） neural network is trained and built for PCB wastewater treatmerit. The five key influential factors on coagulation water treatment technology are regarded as characteristic input vectors, and the effluent quality index as output vectors. The debug data are divided into train group and prediction group. Running the BP neural network, the system error is 0.000 999 96 and the network prediction is in good agreement with the actual data values, showing the precision and the generalization of network is good. Based on the wastewater treatment efficiency, the BP neural network provides a window to reduce the effect of unfavorable factors such as the influent water quantity and quality, adding the chemical reagent reasonably, ensuring the smooth operation of the system, minimizing the olaeration costs and improving the treatment efficiency.
出处
《计算机技术与发展》
2015年第8期194-198,共5页
Computer Technology and Development
基金
广东省"十二五"规划课题(2012JK312)
2013年广东省教指委教改项目(K0155206)
阳江职业技术学院教改课题(2013jgyb02)
广东大学生科技创新培育专项资金项目
关键词
BP神经网络
线路板废水
化学试剂
处理
训练
预测
BP neural network
PCB wastewater
chemical reagent
treatment
training
prediction
