Drilling in the Digital Age: Machine Learning Assisted Bit Selection and Optimization

Abstract Conventionally, a bit is selected from offset well bit run summaries. This method of selection is not always accurate since each bit is run under different conditions which might not be reflected in an offset study analysis. The large quantities of data generated from real time measurements in offset wells makes machine learning the ideal tool for analysis and comparison. Artificial Neural Network (ANN) is a relatively simple machine learning tool that combines inputs and calculation layers to compute a specified output layer. The ANN is fed over thousands of data points from 17-1/2 in hole sections across multiple wells. A specific model is then trained for every bit with weight on bit (WOB), rotary speed (RPM), bit hydraulics, and lithological properties as inputs and rate of penetration (ROP) as output. The model is finalized when a satisfactory statistical set of KPI's are achieved. Using a combination of Monte-Carlo analysis and sensitivity analysis, different bits are compared by varying parameters for the same bit and varying the bit under the same parameters. A bit and its optimized parameters are proposed, resulting in an average instantaneous ROP improvement of 32%. Performance benchmarked with individual drilling parameters shows improved ROP response to WOB, RPM, and bit hydraulics in the optimized run. This project solidifies machine learning as a powerful tool for bit selection and parameter optimization to improve drilling performance. Machine learning will become a significant part of well planning, design, and operations in the future. This study demonstrates how ANN's can be used to learn from previous operations and influence planning decisions to improve bit performance.

Download Full-text

Coal identification using neural networks with real-time coalbed methane drilling data

The APPEA Journal ◽

10.1071/aj18091 ◽

2019 ◽

Vol 59 (1) ◽

pp. 319 ◽

Cited By ~ 5

Author(s):

Ruizhi Zhong ◽

Raymond Johnson Jr ◽

Zhongwei Chen ◽

Nathaniel Chand

Keyword(s):

Real Time ◽

Coalbed Methane ◽

Gamma Ray ◽

Ann Model ◽

Drilling Data ◽

Weight On Bit ◽

Net To Gross ◽

Rotary Speed ◽

Artificial Neural Network Ann ◽

Log Interpretation

Currently, coal is identified using coring data or log interpretation. Coring is the most dependable methodology, but it is costly and its characterisation is expensive and time consuming. Logging methods are convenient, reliable, and reproducible, but can be subject to statistical and shouldering effects and often have operational difficulties in deviated or horizontal wells. Drilling data, which are routinely available, can potentially be used to identify coal sections in a machine learning environment when conventional wireline logs are not available. To achieve this, a four-layer artificial neural network (ANN) was used to identify coals in a well at Walloon Sub-Group, Surat Basin. The ANN model used drilling data and some logging-while-drilling (LWD) data. The inputs for the lithological model from high-frequency drilling data include weight on bit, rotary speed, torque, and rate of penetration. Inputs from LWD data include gamma ray and hole diameter. The criterion for coal identification is based on bulk density cutoff. The simulation results show that the ANN can deliver an overall accuracy of 96%. Due to the low net-to-gross ratio of coals within the Walloon sequence, a lower but reasonable F1 score of 0.78 is achievable for the coal sections. The proposed model can potentially be implemented in real-time to identify coal intervals without additional logs and aid validation of minimal log data.

Download Full-text

Fractional-order controllers for stick-slip vibration mitigation in oil well drill-strings

Journal of Low Frequency Noise Vibration and Active Control ◽

10.1177/1461348420984040 ◽

2021 ◽

pp. 146134842098404

Author(s):

Massinissa Derbal ◽

Mohamed Gharib ◽

Shady S Refaat ◽

Alan Palazzolo ◽

Sadok Sassi

Keyword(s):

Fractional Order ◽

Degrees Of Freedom ◽

Lumped Parameter Model ◽

Oil Well ◽

Proportional Integral Derivative ◽

Stick Slip ◽

Proportional Integral ◽

Drilling Performance ◽

Weight On Bit ◽

Rotary Speed

Drillstring–borehole interaction can produce severely damaging vibrations. An example is stick–slip vibration, which negatively affects drilling performance, tool integrity and completion time, and costs. Attempts to mitigate stick–slip vibration typically use passive means and/or change the operation parameters, such as weight on bit and rotational speed. Automating the latter approach, by means of feedback control, holds the promise of quicker and more effective mitigation. The present work presents three separate fractional-order controllers for mitigating drillstring slip–stick vibrations. For the sake of illustration, the drillstring is represented by a torsional vibration lumped parameter model with four degrees of freedom, including parameter uncertainty. The robustness of these fractional-order controllers is compared with traditional proportional-integral-derivative controllers under variation of the weight on bit and the drill bit’s desired rotary speed. The results confirm the proposed controllers effectiveness and feasibility, with rapid time response and less overshoot than conventional proportional-integral-derivative controllers.

Download Full-text

A Dynamic Model for Rotary Rock Drilling

Journal of Energy Resources Technology ◽

10.1115/1.3230387 ◽

1982 ◽

Vol 104 (2) ◽

pp. 108-120 ◽

Cited By ~ 17

Author(s):

I. E. Eronini ◽

W. H. Somerton ◽

D. M. Auslander

Keyword(s):

Drill Pipe ◽

Tooth Wear ◽

Drill String ◽

Rock Drilling ◽

Large Power ◽

Drilling Rig ◽

Drilling Performance ◽

Weight On Bit ◽

Rotary Speed ◽

Mud Pressure

A rock drilling model is developed as a set of ordinary differential equations describing discrete segments of the drilling rig, including the bit and the rock. The end segment consists of a description of the bit as a “nonideal” transformer and a characterization of the rock behavior. The effects on rock drilling of bottom hole cleaning, drill string-borehole interaction, and tooth wear are represented in the model. Simulated drilling under various conditions, using this model, gave results which are similar to those found in field and laboratory drilling performance data. In particular, the model predicts the expected relationships between drilling rate and the quantities, weight on bit, differential mud pressure, and rotary speed. The results also suggest that the damping of the longitudinal vibrations of the drill string could be predominantly hydrodynamic as opposed to viscous. Pulsations in the mud flow are found to introduce “percussive” effects in the bit forces which seem to improve the penetration rate. However, it is known from field observations that drill pipe movements, if strong enough, may induce mud pressure surges which can cause borehole and circulation problems. Bit forces and torques are shown to be substantially coupled and the influence of certain rock parameters on variables which are measurable either at the bit or on the surface support the expectation that these signals can furnish useful data on the formation being drilled. Other results, though preliminary, show that the effects of the lateral deflections of the drill string may be large for the axial bit forces and significant for the torsional vibrations. For the latter, the unsteady nature of the rotation above the bit increases and the resistance to rotation due to rubbing contact between the drill string and the wellbore accounts for very large power losses between the surface and the bit.

Download Full-text

First Use of Machine Learning for Penetration Rate Optimisation on Elgin Franklin

10.2118/205466-ms ◽

2021 ◽

Author(s):

Daniel O'Leary ◽

Deirdree Polak ◽

Roman Popat ◽

Oliver Eatough ◽

Tom Brian

Keyword(s):

Machine Learning ◽

Learning Algorithm ◽

Guideline Implementation ◽

Integrated Approach ◽

Change Scenario ◽

Directional Drilling ◽

Digital Revolution ◽

Drilling Performance ◽

Scenario Analyses ◽

Weight On Bit

Abstract Optimising the Rate of Penetration (ROP) on Development wells contributes heavily to delivery of projects ahead of schedule and has long been a goal for drilling engineers. Selecting the best parameters to achieve this has often proved difficult due to the extensive quantities of data concerning formation types, bottom-hole assembly (BHA) design and bit specifications. Legacy drilling data can also be vast and not well characterised, making it very difficult to robustly analyse manually. Additionally, multiple stakeholders can each have their own hypotheses on how to improve drilling performance, including bit vendors, directional drilling companies, drilling engineers and offshore supervisors, creating further confusion in this field. Together with its team of data scientists, TotalEnergies E&P UK (TEPUK) has utilised machine learning to analyse field and equipment data and produce guidelines for optimised drilling rate. The machine learning algorithm identifies parameters which have a statistical likelihood of improving ROP performance whilst drilling. The model was developed using offset well data from TotalEnergies' Realtime Support Centre (RTSC) and bit design information. This represented the first use of Machine Learning in the 20+ years of drilling on Elgin Franklin. Adapting to this new data-based method forms part of a wider digital revolution within TEPUK and the Offshore Drilling Industry. In this case, an integrated approach from the data scientists, drilling engineers and supervisors was required to transition to a new way of working. The first trial of using optimised parameters was on a recent Franklin well (F13) in the Cretaceous Chalk formations. The model generated statistically optimised parameter sheets which were strictly executed on site. Within the guideline sheets were suggested ranges of Revolutions per Minute (RPM), Flowrate, Weight on Bit (WOB) and Torque, as well as recommendations for bit blades and cutters. Heatmaps were generated to show what combination of WOB and RPM would likely achieve best ROP in each sub formation. The parameter range defined was specifically narrow to reduce any time spent varying parameters. In practice the new digital approach was successfully adopted offshore and contributed to the delivery of the 12 ½" and 8 ½" sections in record time for the field, resulting in significant savings versus AFE. Following the success of the guideline implementation, steps have been taken to integrate the machine learning model with live incoming data on TotalEnergies' digital drilling online platform. Since the initial trial on Franklin, online ROP optimisation features have been deployed on the Elgin field and currently provide live parameter guidance, a forecast to section TD and data driven bit change scenario analyses whist drilling.

Download Full-text

An Integrated Approach for Drilling Optimization Using Advanced Drilling Optimizer

Journal of Petroleum Engineering ◽

10.1155/2015/281276 ◽

2015 ◽

Vol 2015 ◽

pp. 1-12 ◽

Cited By ~ 6

Author(s):

David Hankins ◽

Saeed Salehi ◽

Fatemeh Karbalaei Saleh

Keyword(s):

Integrated Approach ◽

Economic Benefits ◽

Operational Parameters ◽

Drilling Performance ◽

Drilling Parameters ◽

Weight On Bit ◽

Drilling Operations ◽

Rotary Speed ◽

Well Data ◽

Advanced Drilling

The ability to optimize drilling procedures and economics involves simulation to understand the effects operational parameters and equipment design have on the ROP. An analysis applying drilling performance modeling to optimize drilling operations has been conducted to address this issue. This study shows how optimum operational parameters and equipment can be predicted by simulating drilling operations of preexisting wells in a Northwest Louisiana field. Reference well data was gathered and processed to predict the “drillability” of the formations encountered by inverting bit specific ROP models to solve for rock strength. The output data generated for the reference well was formatted to simulate upcoming wells. A comparative analysis was conducted between the predicted results and the actual results to show the accuracy of the simulation. A significant higher accuracy is shown between the simulated and actual drilling results. Once simulations were validated, optimum drilling parameters and equipment specifications were found by varying different combinations of weight on bit (WOB), rotary speed (RPM), hydraulics, and bit specifications until the highest drilling rate is achieved for each well. A qualitative and quantitative analysis of the optimized results was conducted to assess the potential operational and economic benefits on drilling operations.

Download Full-text

Analysis of Efficiency of Drilling of Large-Diameter Wells With a Profiled Wing Bit / Badania Efektywności Wiercenia Studni Wielkośrednicowych Świdrem Skrawającym z Profilowanymi Skrzydłami

Archives of Mining Sciences ◽

10.2478/v10267-012-0023-4 ◽

2012 ◽

Vol 57 (2) ◽

pp. 363-373

Author(s):

Jan Macuda

Keyword(s):

Average Rate ◽

Large Diameter ◽

Open Pit ◽

Mechanical Parameters ◽

Drilling Rate ◽

Geological Conditions ◽

Rock Drillability ◽

Drilling Technology ◽

Weight On Bit ◽

Rotary Speed

Abstract In Poland all lignite mines are dewatered with the use of large-diameter wells. Drilling of such wells is inefficient owing to the presence of loose Quaternary and Tertiary material and considerable dewatering of rock mass within the open pit area. Difficult geological conditions significantly elongate the time in which large-diameter dewatering wells are drilled, and various drilling complications and break-downs related to the caving may occur. Obtaining higher drilling rates in large-diameter wells can be achieved only when new cutter bits designs are worked out and rock drillability tests performed for optimum mechanical parameters of drilling technology. Those tests were performed for a bit ø 1.16 m in separated macroscopically homogeneous layers of similar drillability. Depending on the designed thickness of the drilled layer, there were determined measurement sections from 0.2 to 1.0 m long, and each of the sections was drilled at constant rotary speed and weight on bit values. Prior to drillability tests, accounting for the technical characteristic of the rig and strength of the string and the cutter bit, there were established limitations for mechanical parameters of drilling technology: P ∈ (Pmin; Pmax) n ∈ (nmin; nmax) where: Pmin; Pmax - lowest and highest values of weight on bit, nmin; nmax - lowest and highest values of rotary speed of bit, For finding the dependence of the rate of penetration on weight on bit and rotary speed of bit various regression models have been analyzed. The most satisfactory results were obtained for the exponential model illustrating the influence of weight on bit and rotary speed of bit on drilling rate. The regression coefficients and statistical parameters prove the good fit of the model to measurement data, presented in tables 4-6. The average drilling rate for a cutter bit with profiled wings has been described with the form: Vśr= Z ·Pa· nb where: Vśr- average drilling rate, Z - drillability coefficient, P - weight on bit, n - rotary speed of bit, a - coefficient of influence of weight on bit on drilling rate, b - coefficient of influence of rotary speed of bit on drilling rate. Industrial tests were performed for assessing the efficiency of drilling of large-diameter wells with a cutter bit having profiled wings ø 1.16 m according to elaborated model of average rate of drilling. The obtained values of average rate of drilling during industrial tests ranged from 8.33×10-4 to 1.94×10-3 m/s and were higher than the ones obtained so far, i.e. from 181.21 to 262.11%.

Download Full-text

Smooth input preparation for quantum and quantum-inspired machine learning

Quantum Machine Intelligence ◽

10.1007/s42484-021-00045-x ◽

2021 ◽

Vol 3 (1) ◽

Author(s):

Zhikuan Zhao ◽

Jack K. Fitzsimons ◽

Patrick Rebentrost ◽

Vedran Dunjko ◽

Joseph F. Fitzsimons

Keyword(s):

Machine Learning ◽

Quantum Algorithms ◽

Machine Learning Algorithms ◽

Low Rank ◽

Smoothed Analysis ◽

Machine Learning Applications ◽

Data Points ◽

Input Model ◽

The Cost ◽

Input Perturbation

AbstractMachine learning has recently emerged as a fruitful area for finding potential quantum computational advantage. Many of the quantum-enhanced machine learning algorithms critically hinge upon the ability to efficiently produce states proportional to high-dimensional data points stored in a quantum accessible memory. Even given query access to exponentially many entries stored in a database, the construction of which is considered a one-off overhead, it has been argued that the cost of preparing such amplitude-encoded states may offset any exponential quantum advantage. Here we prove using smoothed analysis that if the data analysis algorithm is robust against small entry-wise input perturbation, state preparation can always be achieved with constant queries. This criterion is typically satisfied in realistic machine learning applications, where input data is subjective to moderate noise. Our results are equally applicable to the recent seminal progress in quantum-inspired algorithms, where specially constructed databases suffice for polylogarithmic classical algorithm in low-rank cases. The consequence of our finding is that for the purpose of practical machine learning, polylogarithmic processing time is possible under a general and flexible input model with quantum algorithms or quantum-inspired classical algorithms in the low-rank cases.

Download Full-text

Machine Learning-Based Prediction of Air Quality

Applied Sciences ◽

10.3390/app10249151 ◽

2020 ◽

Vol 10 (24) ◽

pp. 9151

Author(s):

Yun-Chia Liang ◽

Yona Maimury ◽

Angela Hsiang-Ling Chen ◽

Josue Rodolfo Cuevas Juarez

Keyword(s):

Machine Learning ◽

Air Quality ◽

Random Forest ◽

Prediction Models ◽

Superior Performance ◽

Support Vector ◽

Economic Activities ◽

Adaptive Boosting ◽

Series Of Experiments ◽

Artificial Neural Network Ann

Air, an essential natural resource, has been compromised in terms of quality by economic activities. Considerable research has been devoted to predicting instances of poor air quality, but most studies are limited by insufficient longitudinal data, making it difficult to account for seasonal and other factors. Several prediction models have been developed using an 11-year dataset collected by Taiwan’s Environmental Protection Administration (EPA). Machine learning methods, including adaptive boosting (AdaBoost), artificial neural network (ANN), random forest, stacking ensemble, and support vector machine (SVM), produce promising results for air quality index (AQI) level predictions. A series of experiments, using datasets for three different regions to obtain the best prediction performance from the stacking ensemble, AdaBoost, and random forest, found the stacking ensemble delivers consistently superior performance for R2 and RMSE, while AdaBoost provides best results for MAE.

Download Full-text

Workflow to build a continuous static elastic moduli profile from the drilling data using artificial intelligence techniques

Journal of Petroleum Exploration and Production Technology ◽

10.1007/s13202-021-01274-3 ◽

2021 ◽

Author(s):

Osama Siddig ◽

Salaheldin Elkatatny

Keyword(s):

Young’S Modulus ◽

Elastic Moduli ◽

Young's Modulus ◽

Correlation Coefficients ◽

Wellbore Stability ◽

Machine Learning Techniques ◽

Support Vector ◽

Ann Model ◽

Weight On Bit ◽

Data Points

AbstractRock mechanical properties play a crucial role in fracturing design, wellbore stability and in situ stresses estimation. Conventionally, there are two ways to estimate Young’s modulus, either by conducting compressional tests on core plug samples or by calculating it from well log parameters. The first method is costly, time-consuming and does not provide a continuous profile. In contrast, the second method provides a continuous profile, however, it requires the availability of acoustic velocities and usually gives estimations that differ from the experimental ones. In this paper, a different approach is proposed based on the drilling operational data such as weight on bit and penetration rate. To investigate this approach, two machine learning techniques were used, artificial neural network (ANN) and support vector machine (SVM). A total of 2288 data points were employed to develop the model, while another 1667 hidden data points were used later to validate the built models. These data cover different types of formations carbonate, sandstone and shale. The two methods used yielded a good match between the measured and predicted Young’s modulus with correlation coefficients above 0.90, and average absolute percentage errors were less than 15%. For instance, the correlation coefficients for ANN ranged between 0.92 and 0.97 for the training and testing data, respectively. A new empirical correlation was developed based on the optimized ANN model that can be used with different datasets. According to these results, the estimation of elastic moduli from drilling parameters is promising and this approach could be investigated for other rock mechanical parameters.

Download Full-text

Prediction of Healing Performance of Autogenous Healing Concrete Using Machine Learning

Materials ◽

10.3390/ma14154068 ◽

2021 ◽

Vol 14 (15) ◽

pp. 4068

Author(s):

Xu Huang ◽

Mirna Wasouf ◽

Jessada Sresakoolchai ◽

Sakdirat Kaewunruen

Keyword(s):

Machine Learning ◽

Search Algorithm ◽

Weather Conditions ◽

Prediction Performance ◽

Machine Learning Algorithms ◽

Coefficient Of Determination ◽

Gradient Boosting ◽

Support Vector ◽

Self Healing ◽

Artificial Neural Network Ann

Cracks typically develop in concrete due to shrinkage, loading actions, and weather conditions; and may occur anytime in its life span. Autogenous healing concrete is a type of self-healing concrete that can automatically heal cracks based on physical or chemical reactions in concrete matrix. It is imperative to investigate the healing performance that autogenous healing concrete possesses, to assess the extent of the cracking and to predict the extent of healing. In the research of self-healing concrete, testing the healing performance of concrete in a laboratory is costly, and a mass of instances may be needed to explore reliable concrete design. This study is thus the world’s first to establish six types of machine learning algorithms, which are capable of predicting the healing performance (HP) of self-healing concrete. These algorithms involve an artificial neural network (ANN), a k-nearest neighbours (kNN), a gradient boosting regression (GBR), a decision tree regression (DTR), a support vector regression (SVR) and a random forest (RF). Parameters of these algorithms are tuned utilising grid search algorithm (GSA) and genetic algorithm (GA). The prediction performance indicated by coefficient of determination (R2) and root mean square error (RMSE) measures of these algorithms are evaluated on the basis of 1417 data sets from the open literature. The results show that GSA-GBR performs higher prediction performance (R2GSA-GBR = 0.958) and stronger robustness (RMSEGSA-GBR = 0.202) than the other five types of algorithms employed to predict the healing performance of autogenous healing concrete. Therefore, reliable prediction accuracy of the healing performance and efficient assistance on the design of autogenous healing concrete can be achieved.

Download Full-text