Improving drilling performance through intelligent and automated operating parameter planning

WEI CHEN and HIEP NGUYEN, SLB 

Choosing the optimum operating parameter values plays a crucial role in facilitating energy access and reducing CO₂ emissions in drilling operation, as it leads to reduced drilling time, improved bottomhole assembly (BHA) integrity, and more accurate well trajectory placement. The critical drilling parameters for well construction process include weight on bit (WOB), top drive rotary speed (RPM), and flowrate.  

For long extended-reach wells, WOB measured at surface might be a poor indicator for actual weight and torque applied at downhole. Either downhole WOB (for BHAs with downhole weight measurement) or motor differential pressure (for motorized BHAs) can be used to replace surface WOB as the controlling parameter.  

Driven by operators’ strong commitment to maximizing reservoir exposure and drilling performance, the number of harsh environment wells being drilled is increasing. These wells present formidable challenges, characterized by high S&V and hard/abrasive formations that result in frequent tool failures, lost-in-hole events, and reduced rate of penetration (ROP). These failures lead to a significant increase in non-productive time, repair cost, and delayed well delivery, Fig. 1. 

Fig. 1. Challenges in well construction drilling operation: (a) severe S&V increase trend over the years; (b) downhole tool damages.

Moreover, these delays impede timely energy access and contribute to an escalation in carbon emissions. To address this challenge, the field domain experts normally go through a very manual and time-consuming process that involves multiple iterations of offset well analysis (OWA) to select the best drilling parameters. Sometimes, particularly when the OWA must be rushed due to tight deadlines, the decisions are based on personal experience. 

With the advancement of oil field digitalization in the last two decades, service providers and rig contractors have captured enormous amounts of drilling data (both surface and downhole sensors measurement). As drilling data are still sporadic in different recording sources and have various storage formats, it is vital to centralize and harmonize the historical drilling data, enrich the data with pre-computed key performance indices (KPI), and merge the data with business context information to enable offset well query and matching automation.  

One of the key influencing factors for drilling performance is lithology. Drilling rocks with different hardness and lithology changes shows different responses. Aligning the historical data by formation top is a critical step, leading to consistent and accurate performance analysis for a specific formation (Shishavan et al. 2020). Analyzing historical drilling data can create many beneficial insights for post-job failure root cause identification and pre-job operating parameters and BHA configuration selection (Shi et al. 2016; Shen et al. 2017; Zhang et al. 2017; Chen et al. 2019).  

By combining data and advanced transient dynamics physics models, we can create more robust and accurate predictive and prescriptive models that can help us make better decisions in the planning phase (Wang et al. 2014; Chen et al. 2015; Chen and Shen et al. 2019; Chen et al. 2021). During the execution phase, a real time surveillance system monitors the log data streamed from the rig and provides the operational recommendations, based on the real-time model-based data reasoning (normally physics or data-driven ROP models). These systems hold great potential for tracking the operation’s compliance with the plan and identifying limiting factors for drilling performance (Barbosa et al. 2019; Ertas et al. 2013; Mazen et al. 2020; Oedegaard et al. 2022; Payette et al. 2015; Singh et al. 2021).  

In this article, a systematic planning framework for drilling operating parameter is introduced. It aims to integrate and automate various parameter planning workflows to pick the correct drilling parameters to maximize ROP with acceptable dysfunction risk. With minimum human interactions, the offset wells can be identified automatically, and the historical data can be pre-processed and aligned with formations and subzones of formations, referred to as drilling zones. By leveraging data analytics, domain knowledge, physics, and machine-learning models, the workflow generates a depth-based, optimum, operating parameter roadmap automatically. The rest of the article is organized as follows: the second section explains the technical framework in detail, the third section briefly discusses the workflow implementation and deployment, and the last section presents the field trial results and conclusions. 

TECHNICAL FRAMEWORK 

The automatic drilling parameter planning includes the following technical workflows: 

• Data cleaning, harmonization and enrichment; 
• Offset well selection; 
• Formation top alignment; 
• Statistical analysis of offset data channels; 
• Sub drilling zone partition inside formations; 
• S&V rating and risk classification per drilling zone; 
• Drilling parameter recommendation, based on domain rule and offset data statistics. 

We explain each workflow in detail, as below. 

DATA CLEANING AND ENRICHMENT 

Data serve as the backbone of any digital workflow. Oilfield service providers have accumulated significant amounts of drilling equipment environmental data. These data hold great value for analyzing the historical time series data, along with business context at large scale. However, the biggest obstacle for the massive data analysis is diverse data sources and various data storage formats. To make the data ready for the parameter planning workflow, an automated data process pipeline is built to extract and link data from different data sources, standardize the file and data channel naming convention, harmonize the data unit, enrich the dataset with performance index computation, and convert the time series log to depth-based data (this is the critical step to enable multiple-well analysis per depth range or formation zone).  

The following performance indices were pre-computed: 

• Inferred operation state and downhole variable: downhole WOB, downhole torque, and motor differential pressure, rig state 
• Bit performance indices: downhole mechanical specific energy (DMSE), bit depth of cut (DOC = ROP Downhole/RPM), formation stiffness (FS = DWOB/DOC), and bit aggressiveness (μ = (36 x DTOR) (DWOB x Dia_bit)).  
• Motor performance indices: motor RPM and torque, motor efficiency and degradation (Zhang et al. 2020) 
• S&V channel standardization and severity level normalization (between 0 and 3). 

MSE is defined as the mechanical energy used to break a unit volume of rock (Teale 2002). It is composed of two energy components: thrust and rotary. It is called DMSE, when using downhole variables to evaluate MSE (Equation 1): 

DMSE = [(4 x DWOB) / (π x Dia²_bit)] + [(480 x DRPM x DTOR) / (Dia²_bit x ROP)] (1)

Here, is the bit diameter (unit: inch), and DWOB (unit: klbf), DRPM (unit: c/min), DTOR (unit: klbf.ft), and ROP (unit: ft/hr) are downhole WOB, downhole RPM, downhole torque, and rate of penetration, respectively. Equation 1 gives a direct description of energy utilization efficiency at the bit. Based on the actual field experience, the DMSE has a good correlation with the rock-confined compressive strength (Dupriest and Koederitz 2005). Therefore, it is normally used as a proxy to the rock strength, if formation evaluation data are not readily available for formation strength computation. In this article, an automatic formation zonation algorithm is developed to detect lithology change (interbeds or hard stringers), based on DMSE.  

OFFSET WELL SELECTION  

For a data-driven parameter roadmap planning workflow, identifying the similar offset wells with high-quality data is very important. Historically, offset well selection was performed manually by local domain experts, based on their experience. To enable the auto offset well selection workflow, the business context information (such as BHA type, trajectory, location, etc.) are assigned when pre-processing the historical data. The job similarity index is computed between the target well and offset wells by considering the following factors: 

• Location: latitude and longitude, country name, and field name 
• Job context: spud time, bit size, BHA type, bit type, mud type, trajectory shape, operator, rig name, depth in, target formation 
• Performance: ROP, S&V, dog-leg severity (DLS), bit dull grading (<=3 dull code is preferred), downhole tool failure. 

Following the auto offset algorithm (Skoff, et al. 2023), the workflow automatically computes similarity index and recommends the offset wells that have the features closet to the target well and the desired drilling performance (such as highest ROP and lowest S&V without tool failures), per the user-defined objectives. 

FORMATION ALIGNMENT AND DRILLING ZONE PARTITION 

Each formation top presents a unique characteristic, not only in terms of geology but also drilling dynamics (ROP, S&V, steering) response. For a reasonable and practical parameter roadmap, it is crucial to conduct parameter optimization at the formation level.  

First, we need to segment the drilling data from multiple offset wells, based on the formation tops that are represented by measured depth (MD), where the formation is first encountered. When choosing offset wells, it is important to pick the wells that drill through the formation layers similar to those in the new target well. By aligning the drilling data from offset wells to the target well, based on the formation top, the information from the historical offset wells can provide insight towards the expected performance of new target wells and facilitate the parameter optimization per formation.  

The main objective of this alignment process is to map drilling data from offset wells to the target well, based on formation. First, formations in offset wells are matched with the one in the target well, using the formation top name. For each formation, depending on the formation layer thickness in the offsets relative to the one in the target well, the drilling data log in each offset well will be shrunk (thicker layer in offset well) or expanded (thinner layer in offset well), so that all the features within that formation are scaled properly before aggregating all offset data into the synthetic data log for target well, Fig. 2. 

Fig. 2. Drilling data depth alignment per formation.

Formation is a rock unit defined by the geologist to distinguish the specific rock layer from the surrounding rock layers. If rock in one formation is relatively homogenous and has little variation in strength, it is natural to treat this formation layer as a separate section in the drilling parameter planning workflow and provide one single parameter recommendation to drill it.  

However, when a formation contains interlayered rock types and possesses greater rock strength variation (more common in thicker formation layers), it is desirable to further divide formation layer into subzones for drilling (referred to as drilling zones). Each drilling zone has distinct patterns in rock strength and drilling responses (ROP and S&V). This provides finer resolution in a formation, such that we can design a separate parameter roadmap for each drilling zone, based on its performance characteristics. DMSE (Eq. 1) is used in this workflow as a formation strength reference for the drilling zone subdivision. To minimize the effect of other factors on DMSE variation, it is recommended to choose offset wells without severe S&V and high bit wear (< 3 dull grading).  

As DMSE is estimated from the sparse and less stable surface log data, the raw DMSE data are generally noisy. Moving average is performed on DMSE data to remove the high-frequency fluctuating noise from the data, Fig. 3a. The changepoint algorithm (Truong et al. 2020) is utilized to detect trend change points in DMSE data that will be treated as potential edges for the drilling zones, Fig. 3b. For each drilling zone, statistics derived from DMSE and S&V data are computed. A decision logic is built to classify the drilling zone into the following types: normal zone (low DMSE average/variance and low S&V), high S&V, interbed (high DMSE variance), and hard stringer (high DMSE average). The automatic drilling zone division and type classification enables an intelligent workflow that picks the strategy matching with the zone type. 

Fig. 3. Drilling zone auto generation: (a) moving averaged DMSE data in one formation; (b) formation subdivision into drilling zones, using the changepoint algorithm; (c) drilling zone tagging, based on drilling performance characteristics.

DATA STATISTICS 

After choosing the offset wells and aligning the formation tops, statistics (P10, P25, median, P75, P90) are determined along the depth for a list of preferred channels: drilling parameters (WOB, RPM, and flowrate), drilling responses (ROP, STOR, DMSE, etc.), and S&V channels (lateral and axial shock RMS, and stick-slip index). The statistics are computed at a 0.5-ft depth interval, based on all of the offset wells within a certain depth window (5 ft in this case). Figure 4 shows an example of drilling data statistical results. 

SHOCK & VIBRATION CONSIDERATIONS 

Per the vibration direction, the drilling dynamics are normally categorized into three modes: axial, lateral and torsional. Various downhole sensors can be deployed at different locations of BHA (measurement while drilling, rotary steerable system, downhole recorders etc.) to detect different vibration modes.

At the rig site, downhole sensor measurements may be transmitted through telemetry and are available in the surface log dataset. We first group the sensor measurements into three vibration mode categories; for example, the stick-slip ratio is assigned to the “torsional” vibration mode group.  

Fig. 4. Drilling data statistics along the depth.

Interpreting the raw absolute measurements to understand the S&V severity is extremely challenging, as different sensors may have different measurement specifications, various data processing procedures, and diverse mounting methods. What we normally do is rate the specific S&V channel, using its own thresholds. S&V rating is a normalization process to reduce the raw S&V sensor measurements (in units of the original physics variables) to the dimensionless severity level (in the scale of 0-3). The rating thresholds can be derived from the operational specification of the tool manufacturer or from the correlation study of S&V and tool failure at large scale. For example, the rating thresholds for stick-slip ratio are given in Table 1. 

After the S&V rating has been done, the interpretation results of different S&V channels (for example, the radial shock RMS and peaks from the rotary steerable system and MWD tool) can be put together for direct comparison. For each S&V mode, all the S&V channels assigned to this mode group are rated, based on their own rating thresholds. The selection of S&V channel and threshold for each channel are open for domain users to adjust, to fit with their application and local best practice.  

The normalized S&V rating results are binned into a 2D binned space, with WOB and RPM as axes. The bin cell is color-coded by the average S&V rating results falling into that cell. The generated 2D histogram is the S&V heat map, Fig. 5. The sum (with weights based on domain knowledge and the detrimental impact of S&V mode) of the heat maps generated from three S&V modes leads to the combined S&V heat map. An auto search algorithm is applied on the combined S&V heat map to find the best possible window (with the lowest S&V rating) for WOB and RPM. The S&V heat map is one of critical analysis tools in the workflow to recommend operating parameters from offset data. 

Fig. 5. Shock and vibration normalization and heatmap generation.

PARAMETER RECOMMENDATION WORKFLOW 

Figure 6 shows the overall process of parameter recommendation workflow. The workflow starts with the formation alignment and drilling zone division and tagging (normal zone, high S&V, interbed, or hard stringer). For each drilling zone or formation layer, all drilling data from offset wells are aggregated, and parameter statistics are computed (P10, Median, and P90) for WOB, RPM, and flowrate. Offset statistics will be used to define the initial optimization region for parameters, Fig. 6a.  

Based on the formation/drilling zone type, the fit-for-purpose parameter recommendation strategy is utilized. In the low-risk homogenous zone, the WOB and RPM are pushed to the top right corner of offset parameter space to maximize the ROP, Fig. 6b. Operators always pursue higher drilling efficiency. If more risk can be taken in the specific zone, flexibility is added in the workflow to allow the user to explore parameters outside the regular parameter range used in offset wells. A small parameter exploration factor (say 10%) is allowed in each parameter planning iteration, Fig. 6c. The user can gradually increase the parameter, iteration by iteration, until another drilling limiter is reached (for instance, bit flounder point or hole cleaning). 

On the other hand, if offset data show certain drilling risks (high S&V, hard stringer, or interbed), the risk mitigation route will be taken. If there is sufficient offset data covering a large operation parameter range in the formation/zone, the S&V heat map can be generated and utilized to find the optimal operating parameter, Fig. 6d. If offset data are limited, or the S&V heat map does not contain any stable parameter window, the domain rule-based approach will be applied to recommend operating parameters, Fig. 6e.  

Fig. 6. Parameter recommendation workflow: (a) offset data statistics to define the parameter optimization boundary; (b) push parameter to the higher end to maximize ROP; (c) explore parameter regime outside the offset parameter range; (d) search stable operating window within S&V heat map; (e) recommend parameter, based on domain rule to mitigate risks.

Domain rules are the best practice or expert knowledge accumulated over the years (Pastusek et al. 2018). With the zone/formation type assigned by the automatic zoning and tagging workflow, we can apply the corresponding domain risk mitigation, accordingly. For instance, in the high S&V zone showing severe torsional vibration risk (stick-slip), we would recommend increasing RPM and reducing WOB to minimize the stick-slip risk. In hard stringer interval, we would recommend maintaining medium-high WOB while reducing RPM to lower the bit’s backward whirl risk. 

WORKFLOW IMPLEMENTATION 

The parameter planning workflow is built into a standalone application that contains all associated steps: data cleaning and enrichment, offset well selection, statistics computation, formation alignment and zoning, S&V heat map generation, and parameter recommendation. The parameter planning application is deployed to fields with a mature data analysis platform that is widely adopted by the drilling community in the service company.

The planning applications can seamlessly utilize the powerful data processing and visualization functions in the mature platform. This makes the application development and deployment cost effective. Figure 7a shows the final presentation of recommended operating parameter roadmap, and Fig. 7b shows the user interface for parameter recommendation strategy customization and S&V heatmap display. 

Fig. 7. Parameter planning workflow implementation: (a) operating parameter window presentation; (b) parameter recommendation strategy customization and S&V heatmap.

FIELD CASE STUDY  

The workflow has been tested in a very challenging drilling environment, with highly interbedded formations, multiple equipment damages, due to destructive shock and vibration, and the operator had to spend several runs to complete the section. After implementing this new parameter planning workflow in seven field test cases in two different platforms, we observed an increase of 66% on rate of penetration (from 56 ft/hr to 93 ft/hr) for the changes in parameter compliance from 25% SWOB compliance to 51% SWOB compliance. Parameter compliance is defined as a measure to track, if the parameters followed the automated operating parameter recommendations. 

Figure 8 shows an interval of one of the planned drilling roadmaps for the 17.5-in sections, with the real-time (RT) parameters (SWOB, RPM, Flow), as well as the RT performance measuring parameters (TORQUE, DMSE, ROP), plotted in dark blue against the corresponding planned parameter windows and the statistical performance of the offset wells, respectively.

Plotting the actual performances against the expected performance from the offset statistics is valuable to monitor the progress and performance of the current job and to understand when, and how, the actual performance starts to deviate from the offset benchmark in a timely manner.  

Fig. 8. Digitalized roadmap for an interval of a challenging 17.5-in. section. Parameters management is key to find the balance with ROP and extreme S&V conditions.

A comparison of real-time S&V measurements, from the current well and the offset S&V statistical median, also provides information related to the current BHA dynamics performance, compared to that of the offset wells. The last track on the right shows the comment on formation zoning type and parameter recommendation method, which brings the visibility of the parameter planning strategy and reasoning to the end-users. This 17.5-in section is drilled through many formations, with varying compressive strengths and interbedding properties that require parameter management to be very specific for each of these formations. Lessons learned from previous offset wells have helped to further optimize the drilling parameters for each of the formations drilled. The digitalized operating roadmap provided not only a tool to aid in the parameter optimization in this case, it also served as a tool to evaluate historical wells for execution compliance (to the plan), compared to performance. 

In Fig. 8, when drilling formation #1 (MD=4,950~5,325ft), SWOB compliance (the measure to show how the actual SWOB follows the recommended parameter window) is low, while RPM and flow have high compliance. The formation is relatively soft, and ROP fluctuates between 200 ft/hr and 400 ft/hr. Mild-to-high stick-slip is expected in this interval. Lateral and axial shock risks are low, based on the offset data. The planned parameter roadmap recommends maintaining high RPM and WOB to push performance. The SWOB was kept lower during the actual operation in this interval.  

In formation #2 (MD=5,324~5,900 ft), SWOB compliance is slightly improved, and RPM and flow compliance are high. The formation is soft at the top and becomes interbedded in the middle. The rock strength in the bottom part gradually stiffens before reaching the next formation (formation #3). The S&V level is similar to that in formation #1. The actual ROP and DMSE are on a par with the offset nominals.  

In formation #3 (MD=5,900~6,200 ft), the rock is considered to be relatively harder, and ROP is about < 50 ft/hour, per the offset performance. A higher WOB is normally required to drill formation #3. High stick-slip was observed throughout the entire formation layer in offset jobs. It is recommended to reduce SWOB and increase RPM to lower stick-slip level. In the actual operation, all parameters (including SWOB) have high compliance. SWOB is lowered (following the parameter roadmap) to manage stick-slip risk. The real time stick-slip measurement shows improvement in this interval. The actual DMSE from the current job is on the lower side of the offset DMSE range, which implies a more efficient drilling condition. 

Figure 9 shows the compliance results for the example case in Fig. 8. The parameter compliance can be evaluated, both for the entire section, as an overall compliance result, and by formation and/or relevant intervals of depth, which then allows detailed analysis of the well-to-well performance in the field. This compliance analysis is done by evaluating the percentage of time the specific parameters were within the boundaries of each of the areas of the recommended parameter windows (lower red, lower yellow, green, upper yellow and upper red). 

Fig. 9. Parameter compliance per formation, for the example shown in Fig. 8. Parameter compliance can be obtained by section, or by formation/interval of interest, and is also shown as overall compliance, and compliance by input parameter (WOB, RPM, flow).

The light blue in the parameter roadmap log (Fig. 8) is the recommended parameter. In order to make rig crews easily follow the drilling roadmap, a large-enough green parameter window (the size is at least 20% of recommended parameter) is proposed in the log. It is also ensured that the upper and lower edges of green window do not violate the equipment and rig operation limits, such as jar neutral zone and pipe buckling thresholds. The effect of green window size on compliance score and drilling performance improvement can be further evaluated, using this automated parameter planning tool. 

The WOB compliance score is relatively low in example formations #1 and #2 in Figs. 8 and 9. Real-time WOB measurements in formations #1 and #2 have significant oscillations, as in drilling in interbedded formation layers and a challenging offshore environment. It is a practical challenge to implement a single recommended value of WOB and hold the WOB steady during execution. Bringing the drilling roadmap to the awareness of the rig team and implementing the planned parameter recommendations through the automated rig control system has the potential to improve parameter compliance.  

The local drilling team may apply the best practices to drill specific formations, to avoid certain risks observed in the field before, such as stuck pipe, loss circulation, etc. It is difficult to learn best practices purely from the historical drilling data. For future work, incorporating the digitalized offset risks and local best practices into the automated workflow can be a practical approach to make a planned roadmap more adaptive to the local drilling application. 

With well-to-well parameters optimization, carefully selecting the most relevant offset wells with similar well profiles and BHA type, every subsequent well drilled is expected to have increased ROP and reduced destructive BHA dynamics. Figure 10 shows a multiple well comparison from this field, with a heat map of ROP, S&V and compliance results for seven wells. With more wells to drill, the heat map will increase in relevance, and a trend toward higher compliance vs higher ROP and lower S&V is expected. Each of the seven wells on Fig. 10 shows parameter compliance (based on the planned parameter roadmap generated from three or more offset wells) and average performances (ROP, lateral shock, and stick-slip index). 

Fig. 10. Heat map, showing ROP on the x axis, shocks on the y-axis, and overall compliance on the z-axis (colored). Several heat maps can be evaluated for different types of drilling performances.

Out of the three input parameters, WOB and RPM are expected to have a bigger impact in the ROP performance and mitigation of destructive BHA dynamics throughout the section, compared with flowrate. For this field case, Fig. 11 shows the ROP performance for all seven wells in the study, compared to the parameter-specific compliance (SWOB compliance). There is correlation between increased SWOB compliance and ROP improvement for the section. 

Fig. 11. Section ROP versus parameter-specific compliance (SWOB) for the seven wells in the study. Plot is sorted by SWOB compliance.

Figure 12 shows the medium and high stick-slip duration metrics for the section, normalized for section footage, versus the SWOB compliance, where there is also a correlation between increasing compliance and lower duration of stick-slip per 1,000 ft drilled in the section. There may be other factors that impact this stick-slip reduction; further implementation of this workflow and study would provide more insight into this correlation. 

Fig. 12. Section stick-slip duration metric versus parameter-specific compliance (SWOB) for seven wells in the study. Plot is sorted by SWOB compliance.

To demonstrate the effectiveness of new parameter planning workflow, we have collected 11 offset wells drilled previously as the comparison benchmark. BHA run counts and average ROP for 11 benchmark offset wells (drilled before the new workflow) and seven test wells (drilled after the new workflow) are summarized in Fig. 13. After implementing the new parameter planning workflow, the average rate of penetration of 17.5-in. section has increased considerably from 58 ft/hr to 93 ft/hr, and the average bit counts used to finish the 17.5-in. section drops from 2.5 to 1.1. Not only the drilling efficiency is improved, but also the obvious reduction in drilling incidents is observed in the field test cases with the new planning workflow. As shown in Table 2, the average incidents per well are reduced from 3.5 to 1.3 after the introduction of new planning workflow.   

Fig. 13. BHA run counts and average ROP comparison for 17.5-in. section between jobs, executed before and after implementing new parameter planning workflow.

CONCLUSIONS 

For decades, the drilling community has used manual techniques to analyze offset wells to pick the correct drilling parameters to maximize ROP with stable drilling dynamics. It lacks the systematic methodology for the decisions that drilling engineers must make on parameter selections. In this article, we demonstrated an intelligent parameter planning workflow that aims to address the challenges and inefficiency in drilling roadmap design. It automates these lengthy manual tasks in offset well analysis and allows the user to be more accurate in defining the parameter roadmap by combining data analytics, domain knowledge, physics, and machine-learning models. 

In the seven field test cases in this article, an obvious 66% increase in ROP was reported after SWOB drilling roadmap compliance was improved from 25% to 51%. This demonstrates the effectiveness of this ROP-centric approach on drilling performance improvement. On the other hand, in the challenging intervals identified from offset data (such as hard stringer, interbeds, and high S&V prone zone), precautious measures are recommended to mitigate the S&V risk and to improve the drilling tool longevity.  

Deploying this workflow to multiple locations for field testing helps expose it to different drilling applications and brings opportunity to evaluate it using large-scale, high-quality, cleaned and enriched offset data. After executing the drilling roadmap in real time, it is critical to collect post-run data and feed into planning workflow to capture lesson learnings and to continuously improve planning algorithm. For future work, we plan to link the planning parameter roadmap to rig control system (RCS) setting (such as control set points and control modes), so that it can be seamlessly implemented in the rig.    

ACKNOWLEDGMENT 

The authors would like to thank SLB management for permission to publish the information and data contained in this article, which is derived from SPE paper 218613-MS, presented at the SPE Conference at Oman Petroleum & Energy Show, Muscat, Oman, April 22-24, 2024.  

_{ABOUT THE AUTHORS}

_{WEI CHEN is a principal engineer at SLB, who leads a team working a unified well construction solution. He has been with SLB since 2010, and his expertise includes drilling optimization and efficiency improvement, shock and vibration simulation, and drill system modeling and data analysis. Mr. Chen holds a doctorate in mechanical engineering from Northwestern University, awarded in 2009.}  

_{HIEP NGUYEN is a data scientist at SLB’s Cambridge Research Center in the UK. He has 16 years of diverse drilling and measurement experience as a field engineer, drilling engineer, and data research scientist. He holds a bachelor’s degree in instrument and industrial informatics engineering and is currently pursuing a Master of Science in Computer Science with Artificial Intelligence.} 

_REFERENCES 

• _{Barbosa, L. F., A. Nascimento, M. H. Mathias, et al. “Machine learning methods applied to drilling rate of penetration prediction and optimization - A review,” Journal of Petroleum Science and Engineering 183., 2019. https://doi.org/10.1016/j.petrol.2019.106332} 
• _{Chen, W., Y. Shen, R. Harmer, et al., “Defining design and optimization method: Dynamic simulation model produces integrated BHA solutions for efficient wellbore delivery,” paper SPE-173008-MS, presented at the SPE/IADC Drilling Conference and Exhibition, London, England, UK, March 17–19, 2015. https://doi.org/10.2118/173008-MS} 
• _{Chen, W., Y. Shen, Z. Zhang, et al., 2019. “Understand drilling system energy beyond MSE,“ paper SPE-196050-MS, presented at the SPE Annual Technical Conference and Exhibition, Calgary, Alberta, Canada, Sept. 30 -Oct. 2, 2019. https://doi.org/10.2118/196050-MS} 
• _{Chen, W., Y. Yu, Y. Shen, et al., “Automatic drilling dynamics interpretation using deep learning,” paper SPE-195919-MS, presented at the SPE Annual Technical Conference and Exhibition, Calgary, Alberta, Canada, Sept. 30–Oct. 2, 2019. https://doi.org/10.2118/195919-MS} 
• _{Chen, W., Y. Shen, R. Chen, et al, “Simulating drillstring dynamics motion and post-buckling state with advanced transient dynamics model,” SPE Drill & Compl 36 (03): 613–627, 2021. https://doi.org/10.2118/199557-PA.} 
• _{Dupriest, F. and W. Koederitz, 2005. “Maximizing drilling rates with real-time surveillance of mechanical-specific energy,” paper SPE-92194-MS, presented at the SPE/IADC Drilling Conference, Amsterdam, The Netherlands, Feb. 23, 2005. https://doi.org/10.2118/92194-MS.} 
• _{Ertas, D., J. R. Bailey, L. Wang, et al., “Drillstring mechanics model for surveillance, root cause analysis, and mitigation of torsional and axial vibrations,” SPE Drill & Compl 29 (04): 405–417, 2013. https://doi.org/10.2118/163420-MS.} 
• _{Skoff, G., D. Fink, A. Poor, et al., “Automated offsets for drill bit performance evaluation, analysis, and monitoring at-scale,” paper SPE-212464-MS, presented at the SPE/IADC International Drilling Conference and Exhibition, Stavanger, Norway, March 2023. https://doi.org/10.2118/212464-MS} 
• _{Mazen, A., N. Rahmanian, et al., 2020. “Prediction of penetration rate for PDC Bits using indices of rock drillability, cuttings removal, and bit wear,” SPE Drill & Compl 36 (02): 320–337, 2020. https://doi.org/10.2118/204231-PA.} 
• _{Oedegaard, S., S. Helgeland, M. G. Mayani, et al., “Drilling parameter optimization in real time,” paper SPE-208734-MS, presented at the SPE/IADC International Drilling Conference and Exhibition, Stavanger, Norway, March 7–9, 2022. https://doi.org/10.2118/212484-MS} 
• _{Pastusek, P., D. Sanderson, A. Minkevicius, et al., 2018. “Drilling interbedded and hard formations with PDC bits considering structural integrity limits,” paper SPE-189608-MS, presented at the IADC/SPE Drilling Conference and Exhibition, Fort Worth Texas, USA, March 6–8, 2018. https://doi.org/10.2118/92194-MS} 
• _{Payette, G. S., D. Pais, B. Spivey, et al., 2015. “Mitigating drilling dysfunction using a drilling advisory system: Results from recent field applications,” paper IPTC-18333-MS, presented at the International Petroleum Technology Conference, Doha, Qatar, Dec. 6–9, 2915. https://doi.org/10.2523/IPTC-18333-MS.} 
• _{Shen, Y., Z. Zhang, J. Zhao, et al., “The origin and mechanism of severe stick-slip,” paper SPE-187457-MS, presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, Oct. 9–11, 2017. https://doi.org/10.2118/187457-MS.} 
• _{Shi, J., B. Durairajan, R. Harmer, et al., “Integrated efforts to understand and solve challenges in 26-in. salt drilling, Gulf of Mexico,” paper SPE-180349-MS, 2020, presented at the SPE Deepwater Drilling and Completions Conference, Galveston, Texas, USA, Sept. 14–15, 2016. https://doi.org/10.2118/180349-MS.} 
• _{Shishavan, R. A., D. Adam, and R. Banirazi, 2020. “Data-driven optimization of drilling parameters,” paper SPE-199568-MS, presented at the IADC/SPE International Drilling Conference and Exhibition, Galveston, Texas, USA, March 3–5, 2020. https://doi.org/10.2118/199568-MS.} 
• _{Singh, K., S. Yalamarty, C. Cheatham, et al., 2021. “From science to practice: Improving ROP by utilizing a cloud-based machine-learning solution in real-time drilling operations,” paper SPE-204043-MS, presented at the SPE/IADC International Drilling Conference and Exhibition, Virtual, March 8–12, 2021. https://doi.org/10.2118/204043-MS.} 
• _{Teale, R., “The concept of specific energy in rock drilling,” Int. J. Rock Mech. Mining Sci. 2 (1): 57–73, 1965. https://doi.org/10.1016/0148-9062(65)90022-7.} 
• _{Truong, C., L. Oudre, L., and N. Vayatis, “Selective review of offline change point detection methods,” Signal Processing 167: 107299, 2020. https://doi.org/10.1016/j.sigpro.2019.107299.} 
• _{Wang, Y., Y. Shen, Y., M. Charter, et al., 2014. “High-frequency vibration measurement coupled with time-based dynamic simulations: New system to predict/solve instability issues,” paper SPE-170708-MS, presented at the SPE Annual Technical Conference and Exhibition, Amsterdam, The Netherlands, Oct. 27–29, 2014. https://doi.org/10.2118/170708-MS.} 
• _{Zhang, Z., Y. Shen, W. Chen, et al., 2017. “Continuous high frequency measurement improves understanding of high frequency torsional oscillation in North American land drilling,” paper SPE-18173-MS, presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, Oct. 9–11, 2017.  https://doi.org/10.2118/187173-MS.} 
• _{Zhang, Z., Y. Shen, W. Chen, et al., 2020. “Analyzing energy and efficiency of drilling system with mud motor through big data,” paper SPE-201502-MS, presented at the SPE Annual Technical Conference and Exhibition, Virtual, Oct. 26–29, 2020. https://doi.org/10.2118/201502-MS}