摘要
针对长距离矿浆管道临界流速计算难度大、传统经验公式预测精度低且适用性差等问题,提出了一种基于麻雀搜索算法(SSA)-卷积神经网络(CNN)的矿浆管道临界流速预测模型。首先,分析矿浆管道临界流速的影响因素,选取4个主要影响因素作为模型特征;然后,利用SSA算法对CNN模型中的8个超参数进行迭代寻优,消除人为设置参数的不确定性;最后,将优化后的CNN模型用于临界流速的预测,以某一水平矿浆管道试验段为例进行实证研究。结果表明,SSA-CNN模型残差平方和为0.0283,平均绝对百分误差为4.19%,平均绝对误差为0.0540,与LSSVM、SSA-BP和CNN模型相比,该模型的预测精度更高,学习和泛化能力更强,为矿浆管道输送研究提供了一种新思路。
In order to solve the problems of the difficulty in calculating the critical velocity of long-distance pulp pipeline and the low prediction accuracy and poor applicability of the traditional empirical formula,a prediction model of the critical velocity of pulp pipeline based on Sparrow Search Algorithm(SSA)optimized convolutional neural network(CNN)was proposed.Firstly,the influencing factors of the critical flow velocity of the slurry pipeline were analyzed,and four main influencing factors including solid material density,slurry volume fraction,conveying pipe diameter,and particle size were selected as model characteristics.Then,the experimental data were normalized to build a CNN model that includes an input layer,2-layer convolution layer,2-layer activation layer,2-layer full connection layer,and an output layer.SSA algorithm was used to optimize the number of iterations,vector,the first layer of convolution kernel size and quantity,the second layer convolution kernel size and quantity,and the number of neurons in two full connection layer of the CNN model to avoid the influence of parameter value on model regression performance.Finally,the optimized CNN was used to predict the critical velocity of pulp pipelines.Taking a long-distance pulp pipeline test section as an example,an empirical study was carried out,and the prediction results were compared with those of other models.The results show that:The average relative error of the SSA-CNN prediction model is lower than that of other models,and the sum of squares of residual error is 0.0283,the average absolute percentage error is 4.19%,and the average absolute error is 0.0540.Compared with LSSVM,SSA-BP and CNN,the prediction accuracy of this model is higher.It is proved that the model design scheme is correct and provides a new idea for the study of pulp pipeline transportation.
作者
张新生
贺凯璐
ZHANG Xin-sheng;HE Kai-lu(School of Management,Xi'an University of Architecture and Technology,Xi'an 710055,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2022年第5期2524-2531,共8页
Journal of Safety and Environment
基金
国家自然科学基金项目(41877527)
陕西省社科基金项目(2018S34)。
