A Back-Propagation Neural Network Model Based on Genetic Algorithm for Prediction of Build-Up Rate in Drilling Process

In the process of directional well drilling, the control of build-up rate of the drilling tool assembly with bottom hole stabilizer is the key to the control of the drilling trajectory, and the prediction of the build-up rate is the precondition of the build-up control. Because the prediction of build-up rate is affected by many factors, it is a complicated nonlinear problem influenced by many variables. It is difficult to accurately describe the specific relationship between build-up rate and many influencing variables with quantitative relations. Therefore, this paper proposes a new method to predict the build-up rate based on back-propagation (BP) neural network optimized by genetic algorithm. Firstly, the influence factors of build-up rate are selected as the input of the model, and the output is the actual build-up rate. Then, the model seeks the global threshold weight of the neural network through genetic algorithm and constructs the prediction model. The numerical test results of drilling data in an oilfield show that this method is 0.9176 in MSE evaluation index, which is better than BP neural network, support vector regression, ridge regression, radical basis function neural network and other machine learning methods. This method is not restricted by the guide structure and principle of the tool, making that the build-up rate may be calculated according to the measured real-time data. It can be used as a model for popularization and application in engineering applications, which can effectively improve drilling efficiency and reduce drilling costs.