Prediction Method of Cutting Settling Velocity in Gas-liquid Two-phase Flow Based on Multi-gene Genetic Programming

Abstract

The cuttings settling process becomes erratic in the complex gas-liquid mixture under gas kick, it is very difficult to accurately describe the migration process of cuttings. Meanwhile, the traditional empirical formula based on fitting experimental data is difficult to accurately predict the complex cuttings settlement. Neural network and other intelligent models with high prediction accuracy are difficult to be popularized and applied due to their black box properties. Multi-gene genetic programming can accurately describe complex nonlinear problems and automatically optimize the structure and parameters of the mathematical model, so as to effectively reduce the complexity of the model. Based on the multi-gene genetic programming algorithm, this study used a variety of input parameters to predict the settling velocity, explored the relationship between the input variables and the result, and established an explicit mathematical model of settling velocity with RMSE of 0.0896 in test set and R2 of 0.9292, which breaks the accuracy limits of traditional empirical model and the inexplicability of neural network model. This new method for predicting cuttings settling velocity in gas-liquid mixture can provide theoretical guidance for efficient cuttings migration in wellbore during gas kick.