罗汉果籽吸附氟离子效果的不同预测模型研究

Study on different predictive models for the adsorption of fluoride ions by Siraitia grosvenorii seeds

中文关键词:  罗汉果籽  BP-ANN  GA-BP-ANN  氟离子  预测模型  响应面

英文关键词:Siraitia grosvenorii seeds  back propagation artificial neural network  genetic algorithm  fluoride ions  prediction model  response surface

中文摘要:

      目的 建立不同罗汉果籽吸附氟离子预测模型。方法 以吸附量为评价指标, 筛选影响吸附效果的因素。在单因素的基础上, 通过响应面法(response surface methodology, RSM)优化吸附温度、接触时间、吸附剂投加量、氟离子初始质量浓度和溶液pH。以吸附温度、接触时间、吸附剂投加量、氟离子初始质量浓度和溶液pH作为输入参数构建基于反向传播人工神经网络(back propagation artificial neural network, BP-ANN)的吸附量预测模型。根据模型在预测集上的表现确定具体的输入参数, 将优化隐含层神经元数的BP-ANN与其他学习模型[遗传算法(genetic algorithm, GA)]优化的模型对比。结果 通过两种模型的决定系数(coefficient of determination, R2)、平均绝对误差(mean absolute error, MAE)、均方误差(mean square error, MSE)、均方根误差(root mean square error, RMSE)、平均绝对百分比误差(mean absolute percentage error, MAPE)值比较, 得出GA-BP-ANN预测模型(R2=0.92594)的预测效果较优于BP-ANN (R2=0.88498)。结论 相较于BP-ANN预测模型, 经过优化后的GA-BP-ANN预测模型对吸附量的预测精度更高。GA-BP-ANN预测模型可为罗汉果籽吸附氟离子效果提供技术参考, 去除水中氟离子效果较好。

英文摘要:

      Objective To establish a prediction model of fluoride ion adsorption by different Siraitia grosvenorii seeds. Methods The adsorption capacity as evaluation index for screened factors affected the adsorption effect. Based on single-factor experiments, response surface methodology (RSM) was used to optimize the adsorption temperature, contact time, adsorbent dosage, initial fluoride ion mass concentration, and solution pH. An adsorption capacity prediction model was constructed using back propagation artificial neural network (BP-ANN) with the input parameters being adsorption temperature, contact time, adsorbent dosage, initial fluoride ion mass concentration, and solution pH. According to the performance of the model in the prediction set, the specific input parameters were determined, and the BP-ANN which optimizes the number of neurons in the hidden layer was compared with the models optimized by other learning models (genetic algorithm, GA). Results By comparing the coefficient of determination (R2), mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE) values of the two models, it was concluded that the GA-BP-ANN model (R2=0.92594) had a better predictive effect than the BP-ANN model (R2=0.88498). Conclusion Compared with BP-ANN prediction model, the optimized GA-BP-ANN prediction model has higher prediction accuracy for adsorption capacity. GA-BP-ANN prediction model can provide technical reference for the adsorption effect of fluorine ions by Siraitia grosvenorii seeds, and the removal effect of fluorine ions in water is better.