Novel physical network algorithm for indirect measurement of polished rod load of beam-pumping unit

作     者：Wang, Ke Gong, Guoliang Shen, Rongxuan Wang, An Yao, Zhaolin Mao, Wenyu Lu, Huaxiang 

作者机构：Chinese Acad Sci Lab Artificial Networks Inst Semicond Beijing Peoples R China CAS Ctr Excellence Brain Sci & Intelligence Techn Shanghai Peoples R China Univ Chinese Acad Sci Beijing Peoples R China Beijing Key Lab Semicond Neural Network Intellige Beijing Peoples R China 

出 版 物：《JOURNAL OF ENGINEERING-JOE》 (J. Eng.)

年 卷 期：2019年第2019卷第10期

页      面：7287-7292页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 

基　　金：Beijing ST planning task [Z181100001518006] S&T innovation Project of The Chinese Academy of Sciences [CXJJ-17-M152] Leading Project of The Chinese Academy of Sciences [XDA18040400] Lab of Semiconductor Neural Network Intelligent Perception and Computing Technology 

主　　题：rods (structures) dynamometers neural nets mechanical engineering computing electric drives spatial variables measurement graph theory down-hole conditions data-driven models energy transfer process load sensor drift problem working state monitoring horse head position two-dimension graph dynamometer card indirect polished rod load measurement physical meaning economical measurement novel neural network algorithm beam-pumping unit novel physical network algorithm 

摘      要：A novel neural network algorithm for indirect measurement of the polished rod load of the beam-pumping unit is proposed. The dynamometer card is a two-dimension graph of polished rod load versus horse head position, and it is widely used for working state monitoring of pumping unit. The drift problem of load sensor makes the direct measurement of polished rod load not economical and practical, so indirect measurement of polished rod load is crucial. Through the analysis of the mechanical characteristics of the pumping unit, a neural network with physical meaning (physical network) is designed. The experiment shows that the physical network can simulate the energy transfer process properly with only 18 training data items, which is about 1% of other data-driven models, and maintains good performance under different down-hole conditions.