机制砂空隙率软测量模型构建及试验研究  

Constuction and Experimental Study of Soft Sensing Model for Void Ratio of Manufactured Sand

作　　者：林文华 房怀英[1] 范璐璐 杨建红[1] LIN Wenhua;FANG Huaiying;FAN Lulu;YANG Jianhong(College of Mechanical Engineering and Automation,Huaqiao University,Xiamen 361021,China;Shenzhen Municipal Engineering Corporation,Shenzhen 518000,China)

机构地区：[1]华侨大学机电及自动化学院,福建厦门361021 [2]深圳市市政工程总公司,广东深圳518000

出　　处：《自动化仪表》2022年第1期55-59,64,共6页Process Automation Instrumentation

基　　金：福建省科技重大专项基金资助项目(2020YZ017022);福建省高校产学合作基金资助项目(2020H6012);深圳市科技攻关基金资助项目(JSGG20201103100601004)。

摘　　要：机制砂的空隙率是衡量混凝土性能的重要指标。空隙率的在线检测能够提升混凝土性能。现有的测量方法无法对机制砂空隙率进行在线检测。因此,提出一种通过动态图像法建立软测量模型,进而实现空隙率在线检测的方法。首先,采用基于动态图像法原理构建的机制砂形态测量平台来采集机制砂图像。然后,计算机制砂的关键形态参数,选择合适的软测量模型算法。最后,构建并比较不同软测量模型的预测性能。对比结果显示,随机森林模型的准确率最高,预测值和试验值最大误差为0.6%。相较于传统方法,该方法可在机制砂生产线中在线检测空隙率,有效提升混凝土性能。Void ratio of manufactured sand is an important index to measure the performance of concrete.Its online detection can improve the quality for concrete construction.But it is difficult to realize the on-line detection of void ratio at present.A void ratio on-line measuring method for establishing soft sensing measuring model based on dynamic image method is proposed.Firstly, the sand morphology detection platform based on dynamic image method is used to collect particle image.Then, the key morphological parameters of manufactured sand are calculated, and the appropriate soft sensing model is selected.Finally, the prediction performance of different soft sensing models is constructed and compared.The compared results show that random forest model is the best prediction, and its maximum absolute error is 0.6%.Compared with traditional methods, this method can realize on-line detection of void ratio in sand production line and effectively improve the performance of concrete.

关 键 词：机制砂 空隙率 机器视觉 粒径 粒形 软测量建模 在线检测 随机森林模型 

分 类 号：TH865[机械工程—仪器科学与技术]