Study of the hamming network efficiency for the sucker-rod oil pumping unit status identification

The oil extraction process requires continuous monitoring of the oil well equipment operation. One of the most effective methods of the on-line control of sucker-rod pumps operation is obtaining information from the force sensor in the polished rod or the current sensor of the pump jack driving motor. In many cases, timely troubleshooting and preventive repair allow saving large costs. Therefore, studies in the area of developing diagnostic systems and, on their basis, creating automated control systems for sucker-rod pumping units (SRPU) are of topical value.

The paper discusses the neural-network-based approach to solving the tasks of forecasting the technical status of jack pumps. The modified Hopfield network (Hamming network) was used as a neural network, for which an SRPU status identification algorithm was devised. Due to it, the identification process outputs not the sample curve itself, but only its number, which results in the faster neural network and smaller computing resources and memory required.

For testing the performance of the proposed identification algorithm, a laboratory bench simulating the operating SRPU status diagnostic system was created. The obtained experimental data show that the Hamming-network-based identification system can perform real-time diagnosing of the current status of the downhole equipment with the minimum error.