Data driven-based model for predicting pump failures in the oil and gas industry

Abstract

Equipment reliability is a critical aspect of the oil and gas industry because the failure of such equipment can lead to catastrophic outcomes on different levels. In these industries, the need for safe, accurate, and cost-effective equipment operation requires reliable failure estimation and identification in order to improve performance and ensure that the maintenance strategy is efficient and effective. Thus, failure prediction is vital in the oil and gas industries for extending part lifetime and reducing unexpected equipment failures, thereby avoiding costly plant shutdowns and equipment damage. This paper proposes a simple and easy-to-implement data-driven model based on multiple linear regression for predicting the failures of a seawater pump in the oil and gas industry. Real-world historical data from equipment and process sensors mounted on the selected pump is collected and analyzed to develop the proposed model. The factors used to build the proposed model have been identified through a review of the literature and expert interviews. In addition, the correlation between the input factors is investigated. The results revealed that the proposed model has the capability to predict pump failure with more than 97% accuracy. Furthermore, sensitivity analysis revealed that temperature and pressure are the most important factors influencing pump failure. The findings of the proposed model can assist maintenance managers and decision-makers in properly detecting and predicting failure. It also helps to monitor and control the input factor levels in order to prevent failures. This study contributes to the development of failure prediction alerts, which will serve as a maintenance decision support system for operatives to avoid potential incoming failures.