On the Impact of Temporal Resolution on Nonlinear Model Accuracy for Predicting Wear Due to Solid Particle Erosion in Digital Twins of Oil and Gas Equipment

Abstract

Abstract The use of mathematical models in conjunction with sensor data in digital twins provides a powerful tool to optimize performance, improve efficiency, and reduce costs. Nevertheless, the reliability of model estimations depends on a careful consideration of the specific requirements of the system, the quality of the sensor data, and the level of technical expertise required to implement and validate those models. In this sense, two issues play a relevant role: the input data sampling frequency and the model estimation frequency. Since there are several data sources, like sensors, test and project data, each one with its own acquisition configuration, the input parameters are obtained with different time intervals. Some of them, such as process data, can be acquired in millisecond intervals, while laboratory data, in intervals of several months. On the other hand, slow dynamic phenomena such as erosion and corrosion do not require high model estimation frequency, which may demand a huge amount of computational resources for storage and processing. Hence, the implementation of digital twins demands the conditioning of the input data with algorithms like averages, resampling and interpolation, which may lead to different model estimations, according to the parameters used in the estimation. This paper explores the influence of different frequencies of partial calculations on the accuracy of a nonlinear model, the DNV erosion model, used to predict wear of oil and gas equipment caused by solid particle erosion. The DNV erosion model is based on empirical data and considers several input parameters such as pressure, temperature, flow and sand content. To investigate the impact of different frequencies of partial calculations, the model was run using several temporal resolutions ranging from daily to yearly calculations, using real production data from the Volve field. The outcomes suggest that temporal resolution can have a significant impact on the accuracy of the cumulative wear predictions, mainly, due to the nonlinearity of the applied model and high variability of process parameters. This paper presents important insights into the use of nonlinear models in predicting wear due to solid particle erosion, and highlights the importance of considering the temporal resolution of partial calculations when developing and employing such models. These findings have important implications for the development and optimization of oil and gas equipment used in harsh environments where solid particle erosion is a significant concern.