Role of AI AI is the component that facilitates the leap from industrial automation to industrial autonomy. There are many differ- ent forms of AI that meet the needs of the downstream oil and gas industry. Different AI approaches include: • ML analyses data to recognise patterns, predict out- comes, and optimise tasks, often for fault detection. • Deep learning, an advanced subset of ML, employs sophisticated neural networks for processing data to iden- tify intricate details with greater accuracy. • Generative AI learns from existing datasets to produce new data samples, emergency instructions, or smart summaries. • Natural language processing (NLP) and computer vision interpret language and visual data to generate reports and perform quality control. • Edge AI involves processing data locally on IoT devices, eliminating the need for cloud storage or internet connec- tivity. This approach significantly enhances the ability to adjust machinery settings, track sensor readings, and mon- itor safety conditions remotely. • Agentic AI is a subset of AI that can act autonomously to achieve specific goals. Examples of semi-autonomous operations Industrial AI is being applied to just about every aspect of the refining and petrochemical industry, including pre - dictive maintenance, process optimisation, quality control, demand forecasting, inventory management, logistics and distribution, emissions monitoring and compliance, safety, and energy management. The prototypical example of semi-autonomous opera- tions is predictive and prescriptive maintenance. Various parts of compressors, pumps, and other rotating equip- ment tend to deteriorate and wear out over time due to fluctuations in the quality of fluid components and envi - ronmental factors. The resulting performance decline and unexpected failures can have detrimental financial, safety, and environmental effects. Consequently, it is important to detect abnormalities in equipment at an early stage so that quick planning and implementation of a new or additional maintenance plan can minimise production losses. Anomalies can be detected using vibration and ammeters or monitoring process variables (see Figure 3 ). However, anomalies can be caused by many factors, such as cor- rosion, contamination, and polymerisation. Detecting the cause is critical. If enough abnormal data is available, ML can create a classifier to identify the most likely culprit. If not, then cluster analysis can be used. AI can be used to detect cavitation in pumps. Cavitation is the formation and abrupt collapse of vapour-filled bubbles in a flowing liquid, usually on the low-pressure side of a pump or valve. Symptoms of cavitation include decreased flow, unexpected vibration, noise, erosion, and, eventually, structural damage to equipment. Suppressing cavitation is the most effective way to prevent cavitation. Early pre- diction using ML is crucial to give operators time to make mitigating changes to the operating conditions. Another example is neural networks, which can be used
Normally unattended facilities
Integrated operations centre Industrial autonomy
Augment automation
IA2IA (AI optimisation) Process, material and energy Equipment utilisation Equipment reliability
Increase worker productivity
Worker enablement
Figure 2 Operating companies focus areas for autonomy
Pathway to autonomous operations The journey towards autonomous operations involves sev- eral stages, each requiring careful planning, investment, and collaboration between technology providers and stakehold- ers. Not all companies start at the same point or have the same end goals. Some will be focused on digital transforma- tion, while others will focus on decarbonisation, improving efficiencies, implementing AI solutions, and so on. Some of the stages towards autonomous operations may include: • Defining strategic focus to identify areas where auton - omy can provide the most significant benefits. • Benchmarking/assessment to analyse workflows, data sources, and existing automation levels. • Data integration and management – this is particularly important, as high-quality data is the basis for autonomous systems. Managing data from different sources helps AI and ML models learn and decide effectively. This includes upgrading infrastructure, setting data governance policies, and using data integration platforms. • Developing a roadmap and selecting the right technolo- gies is critical for successful implementation. This includes selecting AI and ML platforms, robotics, IoT devices, and analytics tools that align with the organisation’s goals and existing systems. • Skill development and change management to upskill employees to work with advanced technologies and allow them to perform higher value-added tasks. When deploying autonomous systems, companies typi- cally focus on three areas: Remote or integrated operations centre to remove peo- ple from hazardous environments or inconvenient locations. Augment automation to improve production, asset relia - bility, and safety. Increase worker productivity by automating tasks and providing them with AI-driven operational advisory deci - sion support systems.
66
PTQ Q2 2025
www.digitalrefining.com
Powered by FlippingBook