基于灰色-神经网络的往复泵状态监测和趋势预测研究-《中国安全生产科学技术》

文章信息/Info

Title:: Rearch on condition monitoring and trend prediction of reciprocating pump based on greyneural network

作者:: 刘文才¹; 陈铁²; （1中国石油安全环保技术研究院安全技术研究所，北京〓100083;2 龙源风电工程技术有限公司，北京〓100034）

Author(s):: LIU Wen cai¹; CHEN Tie²; (1CNPC Research Institute of Safety＆Environment Technology, Beijing 100083, China;2LongYuan Wind Power Engineering Technology Company Limited,Beijing 100034, China)

摘要:: 往复泵是油田钻井、注水和压裂等工艺中的重要设备，其工作条件十分恶劣，该设备能否正常运转对油田安全生产十分重要，因此对其易损件，如泵阀、活塞-缸套副、柱塞-密封副等的状态监测和趋势预测，成为往复泵故障诊断的关键问题。主要研究了灰色-神经网络预测方法在往复泵故障趋势预测中的应用，结合实际案例，根据往复泵故障发展趋势，针对其故障诊断与预测的难点，提出采用组合预测模型进行趋势预测，相比采用单一模型预测方法，该组合模型具有较高的精度，对状态监测工作有非常好的应用价值和实际意义。

Abstract:: The reciprocating pump is an important equipment in the processes of drilling, water injection and fracturing. The working condition is bad, so the condition of monitoring and trend prediction of its wearing parts, such as pump valve, pistoncylinder liner and plungerseal pair become the key problem to the safe operation of reciprocating pumps. Whether the device works in the normal state is very important for the safety production of oil field. At present, domestic and foreign technical staff usually do the rearch on condition monitoring and trend prediction using a single model ,but each model has its own limitation, which often leads to larger error between predicted results and real values. According to the fault development trend and difficulties of fault diagnosis and prediction of reciprocating pumps, the prediction model combining with grey and neural network was selected, which had higher prediction accuracy and the effective degrees of prediction. The fault trend prediction method of reciprocating pump was studied based on grayneural network, which is very valuable for the condition monitoring and trend prediction.