Artificial Neural Network for the Prediction of Unconfined Compressive Strength of Ultra Lightweight Cement for Oil and Gas Wells

This study demonstrates the use of Artificial Neural Network (ANN) modeling techniques to estimate the Unconfined Compressive Strength (UCS) of lightweight oil and gas well cement. At varied simulated wellbore temperatures, 172 cement samples were reinforced with variable doses of strength-enhancing ceramic cenospheres. An Isothermal water bath was used to cure the cement samples for 24 hours and the samples were crushed using a compressive load frame. Multiple samples comprising the same properties were crushed to confirm the consistency of the dataset. The model was trained with 70% of the data set, validated with 15%, and tested with the remaining 15%. The proposed ANN model has four layers: a three-neuron input layer, twenty-five neurons in the first hidden layer, five neurons in the second hidden layer, and a single-neuron output layer. The two hidden layer ANN architecture was determined to be the optimum configuration to prevent overfitting the data. After many rounds of hyperparameter tuning, optimum performance was achieved, and the model parameters were retained and utilized. With a Root Mean Squared Error (RMSE) score of 144 psi, coefficient of determination (R2) exceeding 0.9 and accuracy of approximately 90%, the ANN produced a highly accurate model and signified a good model fit when evaluated on the test set. Similarly strong results were observed on the validation set indicating that the model generalized well to unseen data. These results demonstrate that the ANN approach is adept at producing high fidelity models for UCS prediction of lightweight cement for use in Oil and Gas well cementing operations. This paper discusses a new UCS prediction method as an alternative to rigorous experimental testing for lightweight cement systems where achieving optimum strength is highly important for the proper zonal isolation of the wellbore. 1. INTRODUCTION Global energy demand has been and is projected to increase steadily in the coming decades. To match this demand, the oil and gas industry continuously expands its capabilities by operating in extreme and challenging environments (Canbaz, et al., 2021). Since many of the easy-to-reach reservoirs are maturing, it is not uncommon to drill multi-mile wells, both vertically and horizontally, to maximize reservoir contact (Anya, 2018). Many of these new wells are drilled into high-pressure and high-temperature zones (Arbad & Teodoriu, 2020), which require differing cement properties from those traditionally used.