基于卷积神经网络算法判定稀土总量滴定终点技术研究  

Research on Titration Endpoint Determination of Total Rare Earth Based on Convolutional Neural Network Algorithm

作　　者：吴永明[1] 陈思琦 陈吉文 WU Yongming;CHEN Siqi;CHEN Jiwen(University of Science and Technology Beijing,Beijing 100080,China;North China University of Technology,Beijing 100143,China)

机构地区：[1]北京科技大学,北京100080 [2]北方工业大学,北京100043

出　　处：《中国无机分析化学》2024年第5期567-574,共8页Chinese Journal of Inorganic Analytical Chemistry

基　　金：国家重点研发计划项目(2021YFF0700102)。

摘　　要：由于稀土产业发展迅速,市场需求越来越大,应用范围越来越广,需建立一个长期有效且适用于测定各类稀土的方法。针对目前传统稀土总量检测方法效率低、准确度低、滴定终点差异大等问题,难以满足在线检测的需要。卷积神经网络算法通过摄像头记录样品在滴定过程中溶液的颜色变化,对溶液进行图像特征提取和学习,从而有效、准确地实现化学反应过程中溶液颜色的自动化识别,配合步进电机和注射泵等部件实现自动滴定过程,建立了一种基于卷积神经网络的稀土总量浓度在线分析方法。图像识别本质上是对图像信息进行特征提取,而卷积神经网络(CNN)有着传统识别方法不具备的优点,比如能够自行训练、识别速度更快、所需特征更少等,将自动滴定与神经网络相结合,实现了滴定流程的自动化和样品前处理、滴定、终点判定等过程的一体化,且设备内可同时进行五个样品滴定,提高了滴定效率。Due to the rapid development of the rare earth industry,increasing market demand,and wider application range,it is necessary to establish a long-term effective and suitable method for determining various types of rare earths.In response to the problems of low efficiency,low accuracy,and large differences in titration endpoint of traditional total rare earth detection methods,it is difficult to meet the needs of online detection.An online analyzer of total rare earth concentration based on convolutional neural network was proposed.The convolutional neural network algorithm recorded the color changes of the solution during the titration process of the sample through the camera,extracted and learns the image features of the solution,so as to effectively and accurately realized the automatic recognition of the solution color during the chemical reaction process,and realized the automatic titration process with the components such as the stepping motor and the injection pump.In essence,image recognition was to extract features from image information,while convolutional neural network(CNN)had advantages that traditional recognition methods did not have,such as self training,faster recognition speed,fewer features required,etc.This device combined automatic titration with neural networks to achieve automation of the titration process and integration of sample pretreatment,titration,endpoint determination,and other processes.Additionally,the device could simultaneously performed five sample titration tests,improving titration efficiency.

关 键 词：卷积神经网络 自动滴定 稀土总量 终点判定 

分 类 号：O657[理学—分析化学] O655[理学—化学]