摘要
采用机器学习算法中的支持向量机模型(SVM)对水滑石粒度分布进行预测,以超声衰减信号为主要输入变量,以马尔文激光粒度仪测量的粒度分布作为输出。研究首先对超声衰减信号特征进行提取和分析,结合支持向量机的强大学习能力,引入主元分析(PCA),建立了粒度分布预测模型,采用网格搜索(grid search)优化的方式对模型的超参数进行优化。实验结果表明,支持向量机模型能够有效捕捉超声衰减信号与水滑石粒度分布之间的复杂关系,预测值与测量值峰型相似,均方误差MSE为0.1373,模型拟合优度R^(2)=0.9758,累计粒度分布绝对误差最大值为0.0405,预测精度显著高于传统统计方法。研究表明,该方法在处理非线性复杂数据方面具有很大的优势,为水滑石粒度分布的无损检测和实时监测提供了新思路,具有较高的应用价值和推广潜力。
The support vector machine(SVM)model in the machine learning algorithm was used to predict the size distribution of hydrotalcite,with ultrasonic attenuation signal as the main input variable and the size distribution measured by Malvern laser particle size analyzer as the output.The study firstly extracted and analyzed the features of the ultrasonic attenuation signal,combined with the strong learning ability of SVM,and introduced principal component analysis(PCA)to establish a size distribution prediction model.The model’s hyperparameters were optimized using grid search optimization.The experimental results showed that the SVM model could effectively capture the complex relationship between the ultrasonic attenuation signal and the size distribution of hydrotalcite,with the predicted peak shape similar to the measured value.The mean square error(MSE)was 0.1373,the coefficient of determination(R^(2))was 0.9758,the maximum absolute error of the cumulative size distribution was 0.0405,and the prediction accuracy was significantly higher than traditional statistical methods.The study showed that this method had great advantages in dealing with nonlinear complex data and provided a new idea for the nondestructive detection and real-time monitoring of the size distribution of hydrotalcite.It had high application value and potential for promotion.
作者
吴博
马琳萱
张明峰
曹丽娟
周蕾
王学重
WU Bo;MA Linxuan;ZHANG Mingfeng;CAO Lijuan;ZHOU Lei;WANG Xuezhong(College of New Materials and Chemical Engineering,Beijing Institute of Petrochemical Technology,Beijing 102617,China;Beijing Key Laboratory of Biomass Refining and Fine Chemicals,Beijing 102617,China)
出处
《化工进展》
北大核心
2025年第8期4365-4374,共10页
Chemical Industry and Engineering Progress
基金
国家自然科学基金(52102208)
北京石油化工学院交叉科研探索项目(BIPTCSF-002)
State Key Laboratory of Chemical ResourceEngineering,Beijing University of Chemical Technology(CRE-25-C003)。
关键词
模型
粒度分布
主元分析
优化
预测
model
particle size distribution
principal component analysis
optimization
forecast
