科氏流量计故障检测数据清洗方法研究  

Research on Cleaning method of Coriolis flowmeter fault detection data

作　　者：赵德军 秦怡欣 杨鹏[2] ZHAO Dejun;QIN Yixin;YANG Peng(Sinopec Qilu Company Ltd.Zibo,Shandong 255400,China;Northwestern Polytechnical University,Xi’an,710077,China)

机构地区：[1]齐鲁石化,山东淄博255400 [2]西北工业大学,西安710129

出　　处：《自动化与仪器仪表》2023年第7期257-259,264,共4页Automation & Instrumentation

基　　金：国家自然科学基金资助项目(60974109);陕西省重点研发计划重点产业创新链(群)-工业领域(2019ZDLGY17-05-02)。

摘　　要：科里奥利质量流量计(简称科氏流量计)是一种直接式质量流量测量仪表,可以同时测量流体的质量流量、密度等参数。科氏流量计在使用过程中易出现挂壁故障,影响测量精度,因此,需要对挂壁故障进行定期检测。检测所采集的数据质量会影响科氏流量计挂壁故障检测的精度,针对此问题,必须对脏数据进行清洗,进一步提高科氏流量计挂壁故障检测的准确率,提出了一种基于强化学习的数据清洗算法。将数据与后端机器学习模型相结合,使用Q-Learning算法对数据清洗流程进行搜索,寻找能够使后端机器学习模型达到最佳性能的数据清洗策略。最后对科氏流量计挂壁实验数据进行清洗,对算法进行了验证,检测准确率达92%。实验结果表明,该算法在提高数据清洗效率的同时可以有效提高机器学习模型的性能,使挂壁故障检测准确率得到提高。Coriolis mass flowmeter(Coriolis flowmeter for short)is a direct mass flow meter,which can simultaneously measure the mass flow and density of the fluid.In the process of using Coriolis flowmeter,wall failure is easy to occur,which affects the measurement accuracy,so it is necessary to detect the wall failure regularly.The quality of data collected by detection will affect the accuracy of Coriolis flowmeter wall failure detection.To solve this problem,dirty data must be cleaned to further improve the accuracy of Coriolis flowmeter wall failure detection.A data cleaning algorithm based on reinforcement learning is proposed in this paper.Combining the data with the back-end machine learning model,Q-Learning algorithm is used to search the data cleaning process and find the data cleaning strategy that can make the back-end machine learning model achieve the best performance.Finally,the experimental data of Coriolis flowmeter were cleaned and the algorithm was verified,and the detection accuracy reached 92%.Experimental results show that the proposed algorithm can improve the performance of machine learning model and the accuracy of wall failure detection.

关 键 词：科氏流量计 数据清洗 强化学习 机器学习 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]