OCTOBER 26, 2023

AI-Driven Enhancements for Optimal Asset

Performance

By Peter Reynolds

Keywords
Asset Performance Management, Process Optimization, Scaled Agile
Framework, IT-OT Convergence

Overview
Many process manufacturing enterprises need help to apply artificial intelli-
gence (AI) and machine learning tools to improve asset performance. ARC
Advisory Group’s research identifies that while condition-based mainte-
nance is most prevalent, leaders have successfully built the AI capability to
improve asset performance practices through in-house platforms, commer-
cial solutions, and a combination thereof. Today, a small percentage of users
leverage AI to identify maintenance and process is-
AI functionality in asset performance sues and expand the scope of their traditional
management needs to provide early maintenance practices.
warning to minimize process interruptions,
provide information to adjust the process, AI functionality in asset performance strategy pro-
prevent a failure, and ensure the right vides early warning to minimize process
assets are available to match the interruptions, information to adjust the process,
performance goals of the plant. and ensures the right assets are available to match
the performance goals of the plant. Most compa-
nies use internal data science teams and in-house platforms to develop AI
capabilities. This approach, however, comes with the price of higher tech-
nical debt.

Asset performance strategy focuses on maximizing reliability, process effi-

ciency, and achieving the best possible return on investment from the assets.
In the past, users supplemented the strategy with rudimentary real-time data
collection and seldom involved IT. A broader skill set is required to link a
proactive asset performance strategy with AI/ML, achieved through Scaled
Agile Framework (SAFe). SAFe addresses the challenges of procuring multi-
ple commercial AI/ML solutions and integrates these into internal
platforms, reducing technical debt and improving interoperability.

VISION, EXPERIENCE, ANSWERS FOR INDUSTRY

 ARC Insights, Page 2

The Current State of AI and Machine Learning

ARC Advisory Group recently conducted research to gain a better under-
standing of how process manufacturing industry leaders are using Artificial
Intelligence (AI) and machine learning tools to improve uptime, process
health, and return on assets. A global web survey of 155 experts and in-depth
discussions with several
subject matter experts
(SMEs) across different in-
dustries identified
compelling best practices
and reasons for industrial
organizations to rethink
their current approaches to
AI.

While condition monitoring

and rules-based tools were
most prevalent in the in-
dustry, we found 63 percent
of industrial process com-
Which statement most closely aligns with your asset
pany respondents have a
management strategy? program in place to im-
prove asset performance
using advanced analytics and AI. Of those surveyed, 25 percent claim they
can use AI to detect failures and generate maintenance alerts. However, 16
percent use AI and machine learning models to identify and address mainte-
nance or process issues to avoid disruptions, thereby expanding the scope of
asset performance strategy from traditional maintenance practices.

We also surveyed users to determine which functional organizations use the

output of current asset management tools, and the results were alarming. 36
percent of users claim the Internal IT or Data Science organizations use the
output of advanced tools like AI and machine learning and only 13 percent
of maintenance and operations organizations benefit from current AI tools.

Much Needed Functionality in APM Software

The survey results show a high maturity level in understanding the capabil-
ity of AI tools. Users understand the importance of modern tools for

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com

 ARC Insights, Page 3

leveraging data insights across various functional roles, spanning from

maintenance and reliability to operations to improve asset performance. In
order of importance, the top three software functionality needs were:

• Receive early warnings to minimize process interruptions.

• Provide information to adjust the process and prevent failures.
• Ensure the right assets are available to match the performance goals of
the plant.

Which asset management software decision support functionality is important?

Expanding the Scope of Traditional Maintenance Practice

Oleoducto de Crudos Pesados (OCP) Ecuador is a midstream oil and gas
company that transports, stores, and ships crude oil. The company prides
itself on providing reliable, safe, efficient, and environmentally committed
crude oil transportation operations, with a capacity to transport 450-517K
barrels/day from the Amazon to the marine terminal in coastal province of
Esmeraldas. As the country’s production fell, OCP Ecuador was forced to
operate its facilities at reduced capacity (just 30- 50 percent of maximum ca-
pacity) while ensuring reliability of spare assets. However, transportation
fees and tariffs are typically fixed for pipeline companies regardless of utili-
zation.

The company performed calendar-based and condition-based maintenance

activities across its assets installed in remote locations along the 485-km

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com

 ARC Insights, Page 4

pipeline. This strategy had inherent issues, and could not predict and pre-
vent equipment failures from happening. Faced with the pressures of
tightening regulatory requirements, past historical failures, and a goal to ex-
ercise control on operating costs, OCP Ecuador decided to enhance its
existing maintenance strategy. This fast-tracked project would succeed only
if the solution could be implemented in six weeks across 31 crucial assets.

OCP Ecuador selected Aspen Mtell® from several solutions for its ability to
work with both equipment and process data, provide early and accurate
warning of potential issues, and deploy quickly at scale. Aspen Mtell’s ability
to predict and prevent potential operational disturbances made a great im-
pact at OCP Ecuador by reducing maintenance spend, managing spares
effectively, allocating limited resources efficiently, contributing to a healthier
planet, and protecting maintenance personnel.

For example, Aspen Mtell, leveraging its AI/ML based failure and anomaly
detection capabilities, detected an increase in the charged air temperature
and stalling at four of the company’s main engines. Because increases in air
temperature can lead to poor combustion and higher carbon monoxide (CO)
levels, this was a critical “find” by Aspen Mtell. A field inspection based on
the alert resulted in removing one of the engine’s motors from service due to
water leakage. Aspen Mtell prevented excessive damage of the engine, im-
proved combustion, and reduced carbon monoxide emissions.

Delivering AI and Machine Learning as Organizational

Capability
Organizational capability refers to an organization's capacity and ability to
effectively and efficiently perform tasks, activities, and functions required to
achieve its strategic goals and objectives. It encompasses
Most companies still develop AI a wide range of elements, including the organization's
capability using internal data science human resources, processes, technologies, culture,
teams and in-house platforms. This knowledge, and overall capacity to adapt and respond
approach, however, comes with the to changing operational and business conditions. When
price of higher technical debt.
considering AI and machine learning as a capability,
there are some key differences in how organizations ad-
dress both smaller-scope and larger-scope projects involving both similar
and multiple asset classes. Most companies still develop AI capability using
internal data science teams and in-house platforms. This approach, however,
comes with the price of higher technical debt.

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com

 ARC Insights, Page 5

Technical debt refers to the concept in software development where

shortcuts or suboptimal solutions are knowingly or unknowingly taken dur-
ing the development process. It represents the accumulated consequences of
choosing quick and easy solutions in the short term, which may lead to ad-
ditional work or problems in the future.

Technical debt arises when developers or teams make tradeoffs regarding

code quality, architecture, and interoperability, or other aspects of software
development to meet immediate deadlines or deliver software quickly. Lead-
ers prefer to procure commercial AI/ML solutions, integrate them into
internal platforms, and maintain a robust process to integrate operations,
maintenance, and reliability teams with IT and data science groups.

How does your company deliver AI and machine learning capability to improve asset
performance?

Asset Performance Strategy Development and AI/ML Development

Developing an asset performance strategy involves creating a comprehen-
sive plan to ensure the optimal performance and longevity of equipment and
systems within an organization. This strategy focuses on maximizing relia-
bility & process efficiency and achieving the best possible return on
investment from the assets. Major activities include asset criticality assess-
ment, defining maintenance objectives and performing failure mode effect
analysis (FMEA), RAM analysis, and reliability centered maintenance
(RCM).

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com

 ARC Insights, Page 6

In the past, at best, users supplemented a strategy with rudimentary real-

time data collection and condition-based maintenance often without involv-
ing IT or other specialists. With the application of advanced AI, machine
learning algorithms, modern frameworks and methodologies are needed. A
broader skill set is required to link proactive asset performance strategy with
AI/ML, and this can be achieved through Scaled Agile Framework (SAFe).

The Scaled Agile Framework (SAFe) is a widely used set of principles, prac-
tices, and guidelines designed to help IT-OT organizations scale Agile
methodologies to larger and more complex AI projects or across entire enter-
prises. SAFe provides a structured approach for implementing Agile
practices at scale while maintaining alignment with business goals, improv-
ing IT-OT collaboration, and delivering value to customers more efficiently.

For IT-OT organizations, SAFe addresses the challenges of procuring multi-

ple commercial AI/ML solutions and integrating these into internal
platforms, thus reducing technical debt and improving interoperability. Ag-
ile methods bring together large organizations with various competencies,
each with their dependencies with coordinated planning and execution,
which is different from traditional waterfall projects.

To build AI capability and align AI/ML with a performance strategy, opera-

tions, reliability and SAF must be linked through effective IT-OT teams.
Adjustments to most IT-OT organizations are required. Building a more cen-
tralized capability for agile transformation, centralized enterprise
architecture, and portfolio management will help better screen and deploy
new AI technologies for asset performance.

Recommendations
Improving asset performance requires a strategy far greater than improving
asset utilization or uptime. Leading practices bring the promise of proactive
management to ensure optimal performance of all aspects of asset perfor-
mance through work process improvements and a collaborative, skilled
workforce. Where asset decisions must be made, AI can help analyze the cost
and risk of such choices, not just on a single asset basis but its effects on other
assets, production performance costs, and its interactions with logistical
events.

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com

 ARC Insights, Page 7

Asset management is evolving from its roots in reliability and maintenance

to provide an assurance of uptime or enhanced asset utilization to ensure the
process is fully optimized. AI predictive alerts and prescriptive analytics en-
able early warnings of process and mechanical issues with sufficient time to
change the process operation to correct issues and avoid poor asset yields
and product quality, and evade process-induced degradation of mechanical
equipment. Where maintenance service is inevitable, predictive alerts give
enough time to plan safe and environmentally conscious interventions,
avoiding sudden and sometimes catastrophic breakdowns.

We recommend the following considerations for owner-operators and other

technology users to consider when building AI capability into asset perfor-
mance management. ARC Advisory Group has written about Scaling Asset
Performance Management Across the Enterprise, addressing the resource-
intensiveness of many AI solutions by minimizing the customization of AI
implementations and creating better harmony between project teams. The
report can be found here in the ARC client portal.

Based on ARC research and analysis, we recommend the following actions

for owner-operators and other technology users:

• Organizational frameworks and workflows such as the Scaled Agile

Framework should be adopted by IT-OT organizations.
• A robust process to incorporate SAFe into asset performance strategy
and programs is required.
• Users should seek vendors and solution providers with the ability to
work with both equipment and process data, provide early and accurate
warning of potential issues and deploy quickly at scale.

For further information or to provide feedback on this Insight, please contact your
account manager or the author at [email protected]. ARC Insights are pub-
lished and copyrighted by ARC Advisory Group. The information is proprietary to
ARC and no part may be reproduced without prior permission from ARC.

©2023 • ARC • 3 Allied Drive • Dedham, MA 02026 USA • 781-471-1000 • arcweb.com