摘要
运用反向传播(BP)神经网络模型对已知沉淀池的建设费用进行函数逼近,以已知沉淀池的计算表面积S及计算体积V为网络的输入,沉淀池的建设费用为网络的输出,获取BP网络模型。对目标沉淀池采用回归正交设计方法,以沉淀池的长宽比k、沉淀时间t、效率η为设计因子,编制组合设计表,结合BP网络模型建立平流式沉淀池的费用函数,以该费用函数最小作为优化模型的目标函数,以沉淀池的各水力要素作为约束条件,建立沉淀池的优化设计模型。采用连续型的Hopfield网络对模型进行优化求解。通过实例验证,得出的优化设计方案与常规设计相比可节约费用8%。
Back propagation(BP) neural network was adopted to fit the known horizontal sedimentation tank construction cost with the calculation surface area( S ) and volume(V) of the tank as input data. Regression orthogonal method was used to design the aim sedimentation tank. The combined design table was organized based on the design factors: the proportion of length to width( k ), detention time (t) and the efficiency (η). And the horizontal sedimentation tank cost function was gained by the combined design table and the BP network. Then the integration optimal design model was set up on the basis of the objective function with the minimum cost and the restriction hydraulic factors. The results of the model were calculated with continuous Hopfield network optimal computation theory. A case study shows that the obtained optimal design can cut the charges by 8 % compared with the conventional design.
出处
《同济大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2008年第5期631-635,684,共6页
Journal of Tongji University:Natural Science