Bit Wear Prediction Using Adaptive Data Analytics: Dynamic Scenario for Real Time Well Engineering

Abstract

Abstract As the wells become more complex, identifying the locations of critical bit wear becomes increasingly difficult. The analytical models currently used in the drilling industry are not capable of completely realizing the downhole dynamics and thus incapable of bit wear prediction – they either overestimate or are too conservative, while some just require too many parameters, constants and coefficients. There is no acceptable universal mathematical model that describes bit wear and thus this paper proposes not to use a model at all. This study focuses on harnessing the impact of real time downhole dynamics on drilling parameters to predict bit wear using adaptive data analysis techniques. Real time rate of penetration (ROP) is identified as a parameter that would demonstrate the impact of downhole dynamics with a high degree of functionality. Though other parameters can also be used, ROP is the basic parameter always available. Depending on the parameters available, different parameters can be convoluted together and fed to the algorithm. For instance, ROP and Gamma ray data can be convoluted in an appropriate way to predict bit wear and also predict whether the bit wear is due to formation change or not. The WITSML ROP and drilling data are continuously fed to an adaptive data analytics algorithm which decomposes the incremental ROP data at each depth into its intrinsic mode functions (IMF) using Empirical Mode Decomposition. The trend of the final IMF at each depth is continuously monitored to predict bit wear. For practical purposes of making the whole process fast and straightforward, the energy of the final IMF, referred as IMF energy, is also continuously computed. In either case, the depth and/or depth interval where the IMF energy becomes negative and/or where the trend of the final IMF plot flips is the depth and/or interval where the bit wear is high and may lead to failure. The proposed workflow has been tested on historical wells and the results are compared with analytical models (B. Rashidi et al 2010 (more conservative) and Z. Liu et al. 2014) used in the industry. The results were also verified with the Daily Drilling reports from the field to confirm the predictions of the algorithm presented. The results predicted were synchronous. The paper presents field examples to predict and estimate the bit wear using actual data. Estimation of rate of penetration (ROP) using the new bit wear model was also carried out using actual field drilling parameters. The calculated ROP profile closely matched with the actual data within reasonable accuracy of less than 5%. Specific procedures are proposed for drilling parameter estimation and effective prediction of ROP and bit life.