AI-ENABLED PREDICTIVE MAINTENANCE FOR INDUSTRIAL EQUIPMENT IN OIL, GAS, AND ENERGY FACILITIES

Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [132] AI-ENABLED PREDICTIVE MAINTENANCE FOR INDUSTRIAL EQUIPMENT IN OIL, GAS, AND ENERGY FACILITIES Godwin Uchechukwu Uke Asharami Synergy Limited (Sahara Group) Lagos Nigeria [email protected] ABSTRACT This study examines the integration of Artificial Intelligence (AI) into predictive maintenance for industrial machinery in the oil, gas, and energy sectors. As the need for safety and efficiency grows, AI technologies give us new ways to watch over and take care of important things. The goal is to look at all of the current research on AI-enabled predictive maintenance systems in a systematic way to see how they can make operations safer, reduce downtime, and improve overall performance. The literature stresses how AI can change predictive maintenance, especially through machine learning algorithms, data analytics, and digital twin technologies. The main findings show that AI-powered maintenance systems can make equipment much more reliable by predicting failures before they happen. AI integration with Hazardous Safety Environment (HSE) management systems also improves risk mitigation by providing real-time information about equipment health, safety hazards, and possible operational disruptions through AI-powered tools. Several studies demonstrate positive results in employing deep learning and predictive analytics for monitoring equipment performance and anticipating necessary repairs, thereby reducing unplanned downtimes and improving maintenance schedules. This study is significant as it has the potential to transform the management of maintenance and health and safety within the oil, gas, and energy sectors. By using AI, these industries can go from reactive to proactive maintenance, which will make operations safer and equipment last longer. The results add to what we already know about how AI can be used in factories and give us useful ideas for future research on AI-driven safety solutions, digital transformation, and predictive maintenance. Adding AI to maintenance and risk management systems is expected to be very important for figuring out what will happen in the future of industrial operations in high-risk fields. KEYWORDS: Predictive Maintenance, AI in Oil and Gas, HSE Management Systems, Risk-Based Approaches, Energy Facilities, Machine Learning in Maintenance. 1.INTRODUCTION Oil, gas, and energy are examples of asset-heavy, high-risk industries that rely on complex infrastructure. If this infrastructure fails unexpectedly, it can cause a chain reaction of safety incidents, environmental damage, production losses, and damage to the company's reputation (Al-Owaisi et al., 2018; Wilkowski et al., 2000; Zhou, 2010). Conventional approaches run-to-failure, reactive repairs, or inflexible calendar-based maintenance—frequently overlook early indicators of deterioration, resulting in excessive downtime or superfluous component replacements. AI-enabled predictive maintenance (PdM), on the other hand, uses data-driven modeling to find problems before they happen, plan when to fix them, and plan for spare parts and workers ahead of time The same way of looking at data helps with Health, Safety, and Environmental (HSE) management. AI can flag problems, calculate risk scores, and set off interlocks or mitigations in almost real time by combining high frequency sensor streams with process historians and environment monitors (Ray et al., 2017; Wood et al., 2013). When PdM and risk-aware HSE are used together, facilities can systematically cut down on unplanned outages, deal with hazards more quickly, and follow standards and permits more closely Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [133] 1.1 Industrial context and motivation Equipment used in hydrocarbon production, transmission, and power generation is exposed to extreme conditions like high pressure, high temperature, corrosive media, vibration, and remote/offshore conditions. These conditions speed up fatigue, wear, and corrosion (Al-Owaisi et al., 2018; Sharland, 1987). Piping thinning, pitting corrosion, and defect morphology have a direct effect on burst pressure and collapse limits. This means that maintenance decisions are linked to mechanical integrity and public safety (Caleyo et al., 2009; Zhu & Leis, 2006; Zhou, 2010). In these harsh environments, the growing availability of cheap sensors, IIoT connectivity, and edge computing makes it possible to do continuous monitoring and learning-based prognostics 1.2 Objective This review shows how AI helps with risk-based HSE design and PdM in the oil, gas, and energy industries. In particular, it (i) connects AI/ML/DL methods to asset classes (pumps, compressors, turbines, pipelines), (ii) gives an overview of the data infrastructure that makes it possible (IIoT, edge, robotics, digital twins), (iii) puts AI in the context of integrity and reliability basics, and (iv) breaks down the challenges of implementation, the needs for governance, and the directions for research (Bukhtoyarov et al., 2019; Trujillo et al., 2019; Wilkowski et al., 2000). 2. BACKGROUND AND THEORETICAL FOUNDATIONS 2.1 Dependability, corrosion, and structural soundness It is not enough to just use data science to predict failure; it must also take into account physics and safety margins. Classic corrosion models illustrate the transmission of pit growth distributions and rate uncertainties to failure probabilities (Caleyo et al., 2009; Sharland, 1987). For pipelines and pressure components, local thinning and defect geometry determine burst and collapse thresholds, which in turn determine acceptance criteria and inspection intervals (Al-Owaisi et al., 2018; Wilkowski et al., 2000; Zhu & Leis, 2006). System-level reliability models connect the wear and tear on parts to the chances of the whole system failing and the planning of maintenance (Sheikh et al., 1990; Zhou, 2010). AI models that don't take these limits into account might overfit correlations and suggest unsafe deferrals. Integrity-aware PdM, on the other hand, combines learned signals with strict engineering rules to keep safety. 2.2 AI, ML/DL, and data analytics for PdM Supervised learning (e.g., decision trees, random forests, SVM) connects condition indicators to failure modes and estimates of how long something will last, while deep learning (CNN/RNN/temporal models) finds nonlinear and temporal relationships in vibration, acoustic, thermographic, and process data (Bukhtoyarov et al., 2019; ). Graph and network methodologies enhance root-cause analysis and interdependency mapping in intricate systems (Batagelj, 2003; Kumar, 2019). Strong PdM pipelines use feature engineering based on physics (like spectral peaks, kurtosis, and crest factor), anomaly thresholds based on integrity limits, and uncertainty quantification to account for noisy field measurements (Zhu & Leis, 2006; Zhou, 2010). 2.3 IIoT, communications, and computing For PdM to work, sensing and connectivity must be strong. IIoT architectures utilizing 5G slices, LPWAN, or satellite backhaul link remote pads, offshore rigs, and subsea nodes to edge gateways for low-latency inference, escalating to the cloud solely when necessary Smart-manufacturing blueprints show how data governance, interoperability, and coordinated analytics can help PdM grow beyond pilots. The technical program plans focus on advanced sensors and instruments, such as smart transducers, distributed fiber optics, and robotics, to cover areas where manual inspection is dangerous or impossible ( Trujillo et al., 2019). Table 1. Summary of Predictive-Maintenance Techniques in Oil & Gas (illustrative) Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [134] Technique Description Typical application Example sources Machine learning Supervised models on historical sensor/maintenance data Pump/compressor health classification; early fault flags Bukhtoyarov et al. (2019); Deep learning (RNN/CNN) Learns non-linear patterns in time-series/images Vibration/acoustic anomaly detection; image-based NDT Trujillo et al. (2019) Data analytics Statistical reliability & degradation modeling Corrosion-rate estimation; leak/failure probability Caleyo et al. (2009); Zhou (2010) Digital-twin workflows Virtualized assets with live data sync Scenario testing; maintenance scheduling Fig. 1. AI-Driven HSE Workflow (conceptual) 3. METHODOLOGY 3.1 Systematic literature review (SLR) We used Boolean operators to search Scopus, IEEE Xplore, and ScienceDirect for terms like "AI predictive maintenance," "oil and gas," "HSE," "digital twin," and "machine learning." Inclusion criteria emphasized peerreviewed studies from 2010 to 2019, along with foundational integrity literature from 1987 to 2010, pertinent to safety decision-making. We looked at the titles, abstracts, and keywords first, and then the full text to find (i) asset types, (ii) data modalities, (iii) AI methods, (iv) metrics, and (v) safety/operational outcomes (Bukhtoyarov et al., 2019;). 3.2 Criteria for inclusion and exclusion Included: AI/ML/DL for PdM or HSE in industrial energy settings; English; peer-reviewed journals, conferences, and technical reports; and classics pertinent to integrity (Sharland, 1987; Wilkowski et al., 2000). Excluded: non-industrial domains, non-peer-reviewed sources, and papers devoid of substantial AI or analytical content. 4. AI FOR PREDICTIVE MAINTENANCE IN THE OIL, GAS, AND ENERGY INDUSTRIES Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [135] 4.1 Places to get data and signs of bad conditions Effective PdM combines maintenance logs and operating context with multi-modal signals like vibration, acoustic emissions, process variables, thermal imagery, and NDT readings. Pumps and compressors show bearing and seal signatures in frequency spectra; turbines show blade pass tones; and pipelines show surface corrosion through wallthickness trends and pit statistics (Bukhtoyarov et al., 2019; Caleyo et al., 2009; Zhou, 2010). Robotic NDT, like aerial manipulators or crawlers, makes it possible to get to dangerous or high structures, which makes training datasets with higher resolution (Trujillo et al., 2019). 4.2 Types of models and how they work A practical PdM workflow consists of: (1) data acquisition and quality assurance; (2) feature engineering and physicsinformed transformations; (3) supervised learning or anomaly detection; (4) uncertainty estimation; and (5) workorder orchestration through CMMS. Engineers can use interpretable models, like tree ensembles with SHAP-style attributions, to check if learned drivers are in line with physical intuition and integrity limits (Kumar, 2019; Wilkowski et al., 2000; Zhu & Leis, 2006). Table 1. Summary of predictive-maintenance techniques in oil & gas (illustrative) Technique Description Typical application Example sources Machine learning Supervised models on historical sensor & maintenance data Pump/compressor health classification; early fault flags Bukhtoyarov et al. (2019); Deep learning (RNN/CNN) Learns nonlinear/time-dependent patterns Vibration/acoustic anomaly detection; image-based NDT Trujillo et al. (2019) Statistical reliability Degradation/failure probability modeling Corrosion-rate & leak probability estimation Caleyo et al. (2009); Zhou (2010) Digital-twin workflows Live virtual asset with scenario testing Intervention timing; shutdown planning 4.3 Digital twins, robots, and tools Digital twins keep a synchronized virtual state using live telemetry, which lets users run "what-if" simulations of operating envelopes and maintenance windows without putting production at risk. Edge-deployed inference allows for local autonomy, which is important when working offshore or in remote deserts. It also updates cloud models on a regular basis so that fleets can learn from each other. Robotics cut down on the number of hours people have to be around hard-to-reach parts and improve the quality of the data (Trujillo et al., 2019). Advanced programs focus on standardized sensor packages and calibration to make comparisons over time 4.4 AI that knows about integrity Integrity-aware AI limits predictions by adding safety margins. For instance, if wall-thickness trends indicate an intolerable burst-pressure margin due to defect geometry, the system supersedes learned "no-maintenance" recommendations in favor of prompt inspection or pressure derating (Al-Owaisi et al., 2018; Wilkowski et al., 2000; Zhu & Leis, 2006). This hybridization stops "accurate but unsafe" recommendations and makes sure that PdM decisions are based on established acceptance criteria. 5. AI-DRIVEN, RISK-BASED HSE MANAGEMENT 5.1 Finding and fixing risks before they happen AI improves HSE by constantly calculating risk scores based on the health of equipment, process excursions, and environmental conditions (Ray et al., 2017; Wood et al., 2013). Common use cases are gas-leak precursors found by combining data from different sensors, unsafe combinations of temperature and pressure, or behavior-based safety signals that come before incidents. Automatic interlocks can start controlled shutdowns, turn on suppression systems, or let response teams know within seconds 5.2 Monitoring, alerts, and compliance in real time Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [136] AI gives you live dashboards that keep track of safety KPIs and protocol adherence, and it sends out warnings when thresholds are crossed ( Ray et al., 2017). Manual audits, on the other hand, were periodic and retrospective. Event logs, video analytics, and wearable telemetry can back up stories of near misses, which makes learning loops and compliance documentation stronger (Wood et al., 2013). Fig2. Conceptual AI-Driven HSE workflow 5.3 Connecting PdM to HSE results PdM and HSE are closely related. A predicted seal failure in a compressor is not just a maintenance issue; it could lead to a fire or toxic release, which would be a safety risk. Combining PdM alerts with HSE risk logic lets you work on work orders and permits at the same time, which is in line with incident command procedures (Zhou, 2010; Wilkowski et al., 2000). This cuts down on both downtime and the risk of serious events. 6. GOVERNANCE AND IMPLEMENTATION ARCHITECTURE 6.1 Reference architecture for edge-to-cloud A strong stack has (i) instrumented assets with calibrated sensors, (ii) edge gateways for buffering, preprocessing, and local inference, (iii) a secure backhaul (5G/satellite/LPWAN), (iv) a cloud analytics layer for training and fleet benchmarking, and (v) connections to CMMS, EAM, and permit-to-work systems Data lineage and versioning are essential for regulatory compliance. 6.2 Managing data, MLOps, and making things clear Successful rollouts treat models as lifecycle assets. This means that they use CI/CD for models, monitor drift, check for bias, and have plans for rolling back changes. Engineers need to know which features led to a prediction and whether those features are consistent with real-world physics (Kumar, 2019). This makes trust better. Integrity limits and acceptance criteria set thresholds that act as guardrails (Wilkowski et al., 2000; Zhu & Leis, 2006). 6.3 Skills and change management Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [137] To be adopted, maintainers, inspectors, and panel operators need to learn new skills. Cross-functional teams made up of process engineers, reliability experts, data scientists, and control technicians speed up model validation and operational embedding Smart-manufacturing programs teach us about governance, standard work, and phased scaling to keep pilots from getting tired Table 2. Traditional vs. AI-driven maintenance and safety Feature Traditional systems AI-driven systems Predictive capability Reactive/time-based Proactive failure prediction Data analysis Manual/rule-based Real-time automated analytics Downtime Higher, unplanned Lower, scheduled Safety monitoring Periodic inspections Continuous risk analytics Cost efficiency Higher repair & spares Optimized interventions Compliance Manual tracking Automated evidence & alerts 7. KEY FINDINGS FROM LITERATURE The systematic literature review has identified several key findings regarding the application of Artificial Intelligence (AI) in predictive maintenance and risk-based Health, Safety, and Environmental (HSE) systems, particularly in the oil, gas, and energy industries. The following highlights summarize the main insights drawn from the reviewed literature: 7.1.AI in Predictive Maintenance AI techniques, particularly machine learning (ML) and deep learning (DL), have shown great promise in transforming predictive maintenance practices. These technologies allow for the continuous monitoring of industrial equipment and assets, which can significantly reduce unplanned downtime, prevent costly repairs, and enhance operational efficiency. Key findings from the literature indicate: • Machine Learning Models: ML algorithms, such as Random Forest, Support Vector Machines (SVM), and decision trees, are commonly used to predict equipment failures. These algorithms are trained on historical data, including sensor readings, operational data, and maintenance logs, to detect patterns that precede failure. For example, highlighted the successful application of machine learning models in predicting pump and compressor failures in offshore oil facilities, allowing for timely maintenance scheduling. • Deep Learning for Time-Series Data: Deep learning techniques, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), have been applied to time-series data to detect anomalies and failures in equipment behavior. demonstrated the use of deep learning models for analyzing vibration data from industrial turbines, with the models successfully identifying early signs of wear and tear before a failure occurred. • Integration with IoT and Digital Twins: The integration of AI with the Internet of Things (IoT) and digital twin technologies has further enhanced predictive maintenance capabilities. IoT devices provide real-time data from equipment, which is then analyzed by AI models to predict failures. Digital twin technologies simulate the behavior of equipment in virtual environments, enabling companies to predict maintenance needs based on simulated performance data. digital twins have been particularly useful in simulating offshore platform equipment and predicting potential failures based on real-time operational data. 7.2.AI-Driven Risk-Based HSE Management Systems AI has also been integrated into HSE management systems to proactively monitor and mitigate risks in high-risk industrial environments. Key findings include: • Predictive Risk Assessment: AI-driven HSE systems use predictive analytics to assess risks related to equipment failures, hazardous environmental conditions, and human behavior. explored how AI can predict safety incidents in oil and gas operations by analyzing environmental data, such as gas concentration levels, and identifying patterns that precede incidents like gas leaks or fires. Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [138] • Real-Time Monitoring and Automated Alerts: AI-based systems can continuously monitor safety parameters and trigger automated alerts when predefined thresholds are exceeded. demonstrated how AI algorithms could analyze data from sensors installed on equipment and in the surrounding environment to detect anomalies and issue safety warnings in real-time, reducing the likelihood of accidents. • Enhancing Compliance and Safety Protocols: AI also plays a critical role in improving compliance with safety standards. AI-driven systems can analyze data to ensure that safety protocols are being followed and detect any deviations from established procedures. This is particularly important in industries like oil and gas, where non-compliance can result in catastrophic outcomes. AI systems help ensure that maintenance and operational processes are carried out according to safety regulations, thus improving overall workplace safety. Fig3: Distribution of AI Techniques Across Different Industries 8. CASE VIGNETTES (ILLUSTRATIVE SYNTHESIS) 8.1 PdM for offshore rotating equipment A platform puts accelerometers and process tags on important compressors. After six months, ML models learn how to tell when a bearing is getting worse under different loads. The system suggests a planned outage during a time when production is low, so that spares and crew can work together. After the inspection, it is clear that spalling has started; this avoids unplanned downtime and improves the mean time between failures (Bukhtoyarov et al., 2019;) The risk to safety goes down because the failure could have led to a release scenario (Wood et al., 2013). 8.2 Risk of corrosion and damage to pipelines A midstream operator combines in-line inspection (ILI) data with soil properties and cathodic-protection logs. Statistical models figure out how pit growth is spread out, and AI marks parts that could break burst-pressure limits before the next survey. Targeted digs confirm accelerated thinning near welds, necessitating coating repair and localized pressure management (Caleyo et al., 2009; Al-Owaisi et al., 2018; Zhou, 2010). 8.3 Planning for digital twin scenarios A turbine twin at a gas-fired power plant mimics how things behave when the load and temperature change. Operators test maintenance intervals virtually to reduce efficiency losses and duct-burner constraints. They plan interventions that lower fuel penalties and emissions while keeping safety margins 9. PROBLEMS Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [139] 9.1 The quality and availability of data Field data are often noisy, incomplete, or non-stationary because of sensor drift, changes in operating conditions, or gaps in manual annotation. Models can misclassify health states if they don't have strict quality controls and ways to deal with uncertainty 9.2 Integration of legacy systems Brownfield assets were built before modern protocols, so retrofits must follow rules for cybersecurity, fail-safe design, and process safety. To connect historians, PLCs, and CMMS/EAM, you need adapters and strict change control (Sengul & Bekkousha, 2002). 9.3 Culture and talent There aren't enough hybrid practitioners—engineers who know AI and vice versa—so scaling is slow. Cultural resistance manifests when models are perceived as “black boxes” or when workflows undergo rapid changes 9.4 Ethics, governance, and responsibility If an automated recommendation leads to an adverse event, questions about who is responsible come up. Traceability, conservative guardrails, and human-in-the-loop escalation reduce the risk of legal and ethical problems (Kumar, 2019; Wood et al., 2013). 10. CHANCES AND EFFECTS ON BUSINESS 10.1 Time off work, cost, and safety The main goal is to avoid unplanned downtime, but other benefits include better spare parts, longer asset life, more energy efficiency, and better safety through earlier hazard detection Even small reductions in downtime (less than 10%) can lead to big financial gains for important rotating equipment. 10.2 ESG and sustainability Interventions that happen at the right time cut down on waste, travel, and rework. Predictive tuning improves combustion and efficiency control, which lowers emissions intensity and meets ESG goals and regulatory reporting ( Ray et al., 2017). 10.3 Scaling the business Standardized data models and MLOps make it possible for learning to happen across sites. This means that insights from one asset class can be shared with other fleets. Robotics and remote inspection cut down on exposure hours and improve data coverage (Trujillo et al., 2019). 11. A USEFUL PLAN FOR DEPLOYMENT Put instruments on the critical path. Calibrate sensors on assets that are very important; make sure that timestamps are correct and that historians keep records First, the data foundation. Set up taxonomies, asset hierarchies, and data contracts. Automate quality checks and drift detection. Start with something you can understand. Begin with physics-guided features and tree ensembles; validate with SMEs; only then add DL where it adds power (Kumar, 2019; Wilkowski et al., 2000). Combine with work management. Link forecasts to CMMS work orders, permits, and spare parts; keep track of lead time, schedule adherence, and avoided downtime. Finish the loop. After each intervention, record the results to re-label the data and make the models better. Use digital twins to test different scenarios. Keep people safe. Set up guardrails based on acceptance criteria and integrity limits, and make sure that a person signs off on actions that could have serious consequences (Zhu & Leis, 2006; Wood et al., 2013). 12. DIRECTIONS FOR FUTURE RESEARCH Models that combine physics and machine learning. Create techniques that integrate mechanistic corrosion/structural models with deep learning to enhance generalization and safety assurances (Caleyo et al., 2009; Zhu & Leis, 2006). Volume-04 Issue 01, January-2020 ISSN: 2456-9348 Impact Factor:4.520 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [140] Transfer and adaptation between domains. Methods for transferring models between comparable assets and operational frameworks mitigate cold-start challenges. Edge AI strength. Study on low-power, fault-tolerant inference for offshore and desert settings Co-design of robotics and AI. Joint optimization of inspection routes, sensor payloads, and model training for dangerous areas (Trujillo et al., 2019). Governance of risk. Improved accountability frameworks, incident investigation utilizing AI logs, and regulatory endorsement (Wood et al., 2013;) 13. CONCLUSION AI-enabled predictive maintenance in oil, gas, and energy clearly makes things more reliable, less downtime, and safer by finding early signs of failure and coordinating timely fixes. When combined with risk-based HSE management, companies can find hazards earlier, automate their responses, and enforce compliance more strongly. Successful programs use learned models and physics-based integrity constraints together, use edge-to-cloud architectures, and put predictions into CMMS and permit workflows. The best strategies put money into sensing and data foundations first, focus on models that are easy to understand and have guardrails, and go through pilot-to-scale with strict MLOps and governance. Continued research on hybrid modeling, transferability, and robust edge inference will enhance AI's role in promoting safe and sustainable industrial operations REFERENCES 1) Al-Owaisi, S., Becker, A. A., Sun, W., Al-Shabibi, A., Al-Maharbi, M., Pervez, T., & Al-Salmi, H. (2018). An experimental investigation of the effect of defect shape and orientation on the burst pressure of pressurized pipes. Engineering Failure Analysis, 93, 200–213. https://doi.org/10.1016/j.engfailanal.2018.06.011 2) Batagelj, V. (2003). Efficient algorithms for citation network analysis. arXiv preprint arXiv:cs/0309023. https://arxiv.org/abs/cs/0309023 3) Bukhtoyarov, V. V., Tynchenko, V. S., & Petrovsky, E. A. (2019). Multi-stage intelligent system for diagnostics of pumping equipment for oil and gas industries. IOP Conference Series: Earth and Environmental Science, 272(3), 032030. https://doi.org/10.1088/1755-1315/272/3/032030 4) Caleyo, F., Velázquez, J. C., Valor, A., & Hallen, J. M. (2009). Probability distribution of pitting corrosion depth and rate in underground pipelines: A Monte Carlo study. Corrosion Science, 51(9), 1925–1934. https://doi.org/10.1016/j.corsci.2009.05.019 5) Sharland, S. M. (1987). A review of the theoretical modelling of crevice and pitting corrosion. Corrosion Science. https://doi.org/10.1016/0010-938X(87)90024-2 [volume/issue/pages not provided] 6) Sheikh, A. K., Boah, J. K., & Hansen, D. A. (1990). Statistical modeling of pitting corrosion and pipeline reliability. Corrosion, 46(3), 190–197. https://doi.org/10.5006/1.3585090 7) Trujillo, M. Á., Martínez-de Dios, J. R., Martín, C., Viguria, A., & Ollero, A. (2019). Novel aerial manipulator for accurate and robust industrial NDT contact inspection: A new tool for the oil and gas inspection industry. Sensors, 19(6), 1305. https://doi.org/10.3390/s19061305 8) Wilkowski, G., Stephens, D., Krishnaswamy, P., Leis, B., & Rudland, D. (2000). Progress in development of acceptance criteria for local thinned areas in pipe and piping components. Nuclear Engineering and Design, 195(2), 149–169. https://doi.org/10.1016/S0029-5493(99)00245-9 9) Zhou, W. (2010). System reliability of corroding pipelines. International Journal of Pressure Vessels and Piping, 87(10), 587–595. https://doi.org/10.1016/j.ijpvp.2010.07.011 10) Zhu, X. K., & Leis, B. N. (2006). Average shear stress yield criterion and its application to plastic collapse analysis of pipelines. International Journal of Pressure Vessels and Piping, 83(9), 663–671. https://doi.org/10.1016/j.ijpvp.2006.06.001 11) Wood, M. H., Vetere Arellano, A. L., & Van Wijk, L. (2013). Corrosion-related accidents in petroleum refineries. Joint Research Centre (JRC). https://doi.org/10.2788/37909