摘要
球磨生产中影响浆料性能的因素众多,对陶瓷浆料的批量化生产的稳定性造成很大影响。本研究提出了一种基于数据驱动的球磨工艺逆向寻优模型,用于快速预测球磨性能指标以及逆向找出最优性能指标对应的工艺参数。该方法通过灰色关联分析法进行数据降维,确定了模型的10个输入变量;进而建立10输入3输出的GA-BP预测模型,并通过性能分析确定了模型的结构参数,实现了浆料性能的定量预测。开展了浆料性能的敏感性分析,揭示了工艺参数对浆料性能的影响,实现球磨工艺的优化。针对浆料性能多目标寻优的问题,利用熵权法建立了浆料性能综合评价体系,以浆料性能的综合指标最佳得分作为遗传算法的优化目标,逆向寻优得到对应球磨工艺参数组合,结果表明该方法使得浆料性能综合评价指标得分提高了5.59%。该模型具有较高的预测精度及寻优速度,实现了生产过程的降本增效,对推动陶瓷行业的转型升级具有重大的意义。
In the ball milling process,many technological parameters may influence the slurry performance,significantly impacting the mass production of ceramic slurry.This study innovatively proposes a data-driven inverse optimization model for the ball milling process,which enables to rapidly predict the ball milling performance indicators and reverse find the optimal technological parameters corresponding to the best performance indicators.The method employs grey relational analysis for data dimensionality reduction,indicating 10 input variables for the model.Then,it establishes the genetic algorithm and back-propagation(GA-BP)prediction model with 10 inputs and 3 outputs,where the model’s structural parameters are determined through the performance analysis to achieve a quantitatively prediction of slurry performance.To reveal the effect of technological parameters on the slurry performance,a sensitivity analysis is conducted to optimize the ball milling process.For solving the multi-objective optimization problem of slurry performance,a comprehensive evaluation system is established by the entropy weight method.The optimal composite score of slurry performance indicators is set as the optimization objective for the genetic algorithm,as a result,the technological parameters corresponding to optimal performance can be obtained through the inverse optimization method.Results show this method improved the comprehensive evaluation index score of slurry performance by 5.59%.This model demonstrates high prediction accuracy and optimization speed,achieving cost reduction and efficiency improvement in the production process,which provides significant importance for promoting the transformation and upgrading of the ceramic industry.
作者
高本瀚
董军乐
虞洋
谢正浩
陈雪
GAO Benhan;DONG Junle;YU Yang;XIE Zhenghao;CHEN Xue(School of Mechanical and Electrical Engineering,Guilin University of Electronic Technology,Guilin 541000,China;Guangxi Mona Lisa New Material Co.,LTD.,Wuzhou 543300,China)
出处
《中国陶瓷》
北大核心
2025年第6期51-62,共12页
China Ceramics
基金
广西重点研发计划(桂科AB204010101、桂科AB24010226)
中央引导地方科技发展专项(梧字科202301002)
国家自然科学基金(12162011)
青年人才托举工程(2021QNRC001)
桂林电子科技大学研究生教育创新计划项目(2024YCXS009)。
关键词
球磨工艺
数据驱动
遗传算法
BP神经网络
熵权法
Ball milling process
Data-driven
Genetic algorithm
BP neural network
Entropy weight method
