CAJ | 학술논문

示功图分析是目前比较常用的油井故障诊断方法,基于神经网络的示功图分类识别要求准确地提取示功图的特征值,特征值的质量直接关系到示功图识别的效率和可靠性。传统的示功图特征值提取方法计算量很大,与油井现场的实时性要求相悖。为了解决这一问题,提出了用Freeman链码来表达示功图特征,对示功图的识别进行研究。分析了示功图Freeman链码的提取方法以及典型工况链码特征,建立示功图链码特征样本库,给出了示功图识别的方法步骤,在MAT-LAB下进行仿真验证。结果表明,Freeman链码特征值能够有效地分类出各种典型工况示功图,神经网络具有更快的收敛速度和更高的识别效率。
시공도분석시목전비교상용적유정고장진단방법,기우신경망락적시공도분류식별요구준학지제취시공도적특정치,특정치적질량직접관계도시공도식별적효솔화가고성。전통적시공도특정치제취방법계산량흔대,여유정현장적실시성요구상패。위료해결저일문제,제출료용Freeman련마래표체시공도특정,대시공도적식별진행연구。분석료시공도Freeman련마적제취방법이급전형공황련마특정,건립시공도련마특정양본고,급출료시공도식별적방법보취,재MAT-LAB하진행방진험증。결과표명,Freeman련마특정치능구유효지분류출각충전형공황시공도,신경망락구유경쾌적수렴속도화경고적식별효솔。
The analysis of indicator diagram is a commonly-used method for diagnosing oil well faults. The classification and identifica-tion of indicator diagram based on neural network requires the accurately extracted eigenvalues,the quality of which is directly related to the recognition rate and recognition reliability of the indicator diagram. However,the traditional method for extracting eigenvalues needs a great amount of calculation,so it runs counter to the real-time requirements of the well sites. To solve this problem,attempt to illustrate features of indicator diagram by the Freeman chain-code and then research its identification. Firstly,analyze the extracted methods of in-dicator diagram Freeman chain-code and the typical features of working condition chain code. Then,try to establish a sample library for the indicator diagram chain code features. Finally,provide a practicable method and procedure for the identification of indicator diagram and meanwhile carry out the simulation validation under MATLAB. The results reveal that the Freeman chain-code eigenvalues can sort out all kinds of typical working condition indicator diagrams. Therefore,the neural network will have faster convergence speed and higher recognition efficiency.