论文部分内容阅读
利用BP神经网络模拟了钻井液的粒度分布与钻井液处理剂种类及加量之间的关系,并采用神经网络集成的方法提高模型的泛化能力。利用激光散射粒度分析仪测定了NaCl、CaCl_2、钻井液用高温抗盐降滤失剂SPNC、磺甲基酚醛树脂(SMP-2)4种处理剂对钻井液粒度分布的影响,选用60组实验数据作为试验样本,建立了钻井液粒度分布模型,并利用19组实验数据验证其精度。结果表明:该模型具有良好的预测精度,集成网络输出结果的平均误差率和最大误差率均小于单个神经网络子网,表现出良好的泛化能力;并利用该模型研究了单一处理剂对钻井液粒度的影响。
The relationship between the particle size distribution of drilling fluid and the type and amount of drilling fluid treatment agent was simulated by BP neural network. The neural network integration method was used to improve the generalization ability of the model. The effects of NaCl, CaCl2, high temperature anti-salt fluid loss additive SPNC and SMP-2 on the particle size distribution of drilling fluid were determined by laser light scattering particle size analyzer. Sixty experiments Data as a test sample, the establishment of the drilling fluid particle size distribution model, and the use of 19 sets of experimental data to verify the accuracy. The results show that the model has a good prediction accuracy. The average error rate and the maximum error rate of the output of the integrated network are less than those of a single neural network, which shows a good generalization ability. Using this model, Effect of liquid particle size.