AspenTech recommits to innovation and service

Fifteen months after its acquisition by Emerson, AspenTech’s biannual Optimize 26 event on May 11-14 in Houston had a definite Emersonian flavor. Its keynotes and presentations, panel discussions, technical sessions and exhibits retained AspenTech’s cornerstones of innovation, software development, talented personnel and customer focus, but this time they also paralleled the structure and style of Emerson Exchange events.

This year‘s edition of Optimize attracted more than 1,500 attendees, including more than 600 process-industry customers to more than 150 sessions divided into two main divisions—process and power—and nine mostly concurred tracks.

In his opening keynote address, Vincent Servello, president of AspenTech, reported that Emerson is accelerating its investment in AspenTech to achieve a more agile technology approach for faster innovations; more services that can accelerate time to value; and a programmatic approach to deployment and sustainment. These efforts are performed by the 1,200 R&D and product management personnel working in AspenTech’s process division.

“There are lots of ways to capture and sustain value when you work at AspenTech,” said Servello. “We’re transforming how we innovate for customers play streamlining our R&D efforts from hundreds of small programs to 10 critical technology programs that are well-resource to deliver industry-leading innovations. We’re deploying an agile framework consisting of a disciplined approach to align priorities, accelerate execution, and increase the productivity of our technology initiatives. And, we’re accelerating investment in AI, so it can move users’ operations from data to decisions at enterprise scale.”

AI platform and data fabric

Most notably, Servello announced two major releases during Optimize 26:

• AspenTech AVA AI platform helps users accelerate agentic, domain-aware AI adoption. It’s reported to deliver the agility, efficiency and autonomy users need to respond more quickly to changes in operating conditions, continuously improve performance with trusted-domain context, and act with greater confidence using AI-assisted recommendations embedded in users’ operations. AVA also integrates decades of Emerson‘s expertise and first-principles models into its skills and work flows, while using large language models (LLM).
• AspenTech InMation OT Data Fabric has been enhanced to create a modern, always-on, industrial-data backbone that scales with users as their digital needs evolve, supports analytics and AI capabilities, and develops an operations platform that connects data, context and organization-wide decisions. This is reported to standardize how data is managed and shared, making it easier to connect systems, apply consistent context and governance, and deliver trusted real-time information wherever it’s needed. Inmation OT Data Fabric is a cornerstone of an AI-ready, integrated and scalable AspenTech Inmation Data Platform that supports modules for visualization, workflow engines, applications, private clouds and other functions.

These two releases organizationally bookend and enable AspenTech’s traditional modeling and optimization responsibilities of performance engineering, manufacturing, and supply chain, asset, performance management (APM), and subsurface science and engineering. In addition, AspenTech has also hired 150 new staffers beef up its service offerings, R&D and expert program management.

“We’re delivering these innovations so individual users and their teams can work better,” added Servello. “As part of that effort, we’re also working on combining our Hysys and Aspen Plus engineering suites, so they’ll also be easier to use, and can take advantage of AI capabilities.”

ExxonMobil’s experience

In a subsequent, two-person chat, Dylan Pugh, engineering VP at ExxonMobil, reported it’s using digitalization and AI in its operations in the Permian basin, where it’s benefited from its longtime partnership with AspenTech, and expects to rely on it further as it seeks to achieve net-zero emissions operations by 2030-35.

“We’re pursuing lower-carbon solutions in our upstream and downstream operations and global projects, and we need to identify the best options for us to deploy,” said Pugh. “We need solutions we can take upstream and downstream, and implement across product information management systems (PIMS) multiple sites.”

Pugh reports that ExxonMobil started its low-carbon business five years ago to reduce emissions in hard-to-abate processes that it and its partners operate. However, it needs help from partners like AspenTech for its sustainability efforts to be safe, reliable, and produce good financial results.

“ExxonMobil and AspenTech have been working together for 30 years, so we’ve evolved together, and can leverage each other’s technologies to new challenges like sustainability and apply them at scale,” explained Pugh. “We’ve all worked with oil, gas and chemical processes for many years, but now we need help deploying and running unconventional applications safely, reliably and profitably,, and we know AspenTech will be able to help us achieve these ambitious goals with centralized engineering.”

For instance, ExxonMobil recently implemented an Aspen DMC3 unified architecture with adaptive AI functions on 20 PLCs that are optimizing production at 100 wells.

“With digitalization and AI, many things are happening fast, but that doesn’t mean they’ll necessarily deliver value,” adds Pugh. “We have to make sure we understand the risk implications of new technologies, and not lose sight of historical solutions that were reliable and benefitted us previously. For example, our OT data must be in good shape if we’re going to try and allow more autonomous operations, and we may also need training for behavioral or organizational changes for increased autonomy to be successful.“

User community coalesces

To demonstrate the strength of its user community, Claudio Fayad, CTO at AspenTech, reported that it’s collectively deployed $500 billion worth of process-industry investments over the past 12 years, and helped save more than 150 million tons of CO₂ during the same period. However, he added that many challenges remain, including supply and demand variations, cyber security, increasing regulations, larger, and more complex operations, aging assets and changing workforces.

“A few years ago, clouds were just water vapor in the atmosphere, and now we have cloud computing, so it’s not surprising that everything can seem like it’s upside down,“ said Fayad. “However, I choose to be excited because our community has so many technologies they’re bringing to solve these challenges.”

For instance, beyond ExxonMobile, several major users at Optimize 26 presented case studies about their recent upgrades, migrations and other projects, which they usually implemented with AspenTech’s assistance.

Versalis reinvigorates

For instance, Adriano Alfani, CEO of Versalis (https://www.versalis.eni.com), ENI’s chemical company, reported that it’s been trying to transform the 26 facilities that it operates worldwide, and shift from basic chemicals to more varied and sustainable processes and products.

“We decided we had to change Versalis’ DNA, and develop bio-based products and circular economies,” said Alfani. “We knew this would only work if we put people at the center of transforming the economic, environmental and social practices that successful sustainability requires. We also knew we need to transform many of our legacy technologies, and invest in more digitalization and AI that could enable predictive analytics, machine learning, sustainability, efficiency, maintenance, and safety and environmental issues.”

Consequently, Versalis adopted:

• AspenTech’s engineering software for process design, asset modeling and parameters tuning;
• AspenTech’s manufacturing and supply chain software for planning its production cycles, and optimizing productivity for a higher-value product yields; and
• Aspen Mtell predictive maintenance software with AI and ML functions for monitoring process parameters, and avoiding business continuity impacts.

“Previous optimizations weren’t enough anymore. We needed new models for enterprise-wide optimization, and that began with a data fabric that brought in more data from the OT side, which has more different data streams, which are often created at the millisecond level, and we’re often not shared before because they were so complex,“ explained Alfani. “AspenTech Inmation was ideal for bridging OT and IT data, and beginning to work with AI and large language models (LLM). However, operations must be based on first-principle constraints, and this is when AspenTech AVA can help. It adds first-principle-based guardrails that ground LLMs in operations methods, and makes AI more expert in process applications, so we can compress decision cycles from weeks to hours. Validations that used to take months can now be done in hours, and we can identify red flags and get reports in real time.”

Saudi Aramco

To improve on its linear programming (LP) accuracy and blending decisions, Saudi Aramco’s India-based partner, Mangalore Refinery and Petrochemicals Ltd., recently migrated to non-linear modeling. Its existing Hysys simulations were perfectly tuned for specific operations ranges, and this linear solution was good at predicting dependent variables. However, once outside its typical tuning range, deviations, and other problems could quickly arise, according to Sanjerao Chopade and Rafael Soto, lead asset optimizers at the Mangalore facility.

Consequently, their team integrated AspenTech’s Hysys, Aspen Process Industry Modeling - Advanced Optimization System (PMS-AO), Multicase Analysis and AI Model Builder software on Mangalore‘s continuous catalytic regeneration (CCR) unit to handle non-linear tasks, which included employing more rigorous Hysys models, and using AI-assisted model workflows for 206 independent variables, and 22 dependent variables. These updates allowed the CCR unit to improve its accuracy by almost 99%.

“Non-linear modeling with Hysys and Aspen PIMS-AO also let us do data conditioning in AI-based model-builder software, and select and edit independent and independent variables,” said Soto. “The advantages of integrating an AI model included model applicability for a wide range of operating conditions and feed qualities; greater prediction accuracy due to non-linear equations between independent and dependent variables leading to better optimization and GRM improvement; and also being able to use the AI model for refinery-wide simulation. Shortcomings include the fact that AI model can be a black box for non-expert PIMS users; troubleshooting simulations can be difficult; and non-linear equations are required for transferring feed-pool properties to external AI Model software.”

Takeda Pharmaceuticals

Similarly, W. E. Sanderson, global lead for reliability engineering at Takeda Pharmaceuticals, reported that four of its 25 plants have implemented Aspen Inmation as their data management platform and Aspen Mtell is their asset performance management system (APMS), as well as a computerized maintenance management system (CMMS) for proactively planned in scheduled maintenance interventions, all in conjunction with starting the company’s predictive maintenance program in 2022.

“We started with a lighthouse proof of concept (PoC), and that went well, so we also conducted a four-day Mtell training program, and then an AI agent design workshop, where subject manor experts shared their knowledge of processes and equipment,” said Sanderson. “Predictive maintenance typically uses two types of AI agents. The first are rules agents that operate on single sensor thresholds, specific conditions, and runtime or counter values. They employ engineering calculations that use multiple sensor inputs and first-principles logic.

“The second are machine learning (ML) agents, which include anomaly agents that detect deviations from normal behavior, and are trained on multivariant data representing normal behavior. The second type of ML agents are failure agents that detect specific failure patterns, and are trained on multivariate data that represents known failure examples. The lesson we learned is that predictive maintenance and AI functions may seem like a quick solution, but they aren’t. Designing AI agents requires deep domain knowledge and a lot of work to be done properly.“

Sanderson reported that InMation, Mtell and the CMMS are live at four sites. Each has a local version of InMation, while the company also runs a global InMation version.

“Most failures occur randomly, making calendar-based maintenance and routine inspections inefficient and ineffective,” explained Sanderson. “Agents help us predict where we are on the failure curve, making maintenance more effective and efficient. Agents are also much better than humans at detecting patterns trends, and at connecting the dots. However, this approach inherently involves some uncertainties.”

To maintain autoclave uptime that’s critical to its filling operations, Sanderson reported that Takeda recently deployed, anomaly agents trained on normal autoclave behavior, and they detected early issues, and sent alerts with a rapid checklist for faster, proactive responses.

“The agent flagged an anomaly, despite completing a compliance cycle, and the rapid checklist showed that the vacuum pump and cooling water supply were inspected. This identified a root cause, which turned out to be a malfunctioning, external cooling water-supply pump for the vacuum pump,” explained Sanderson. “The agent’s alert avoided a delay in the fill schedule, prevented having to redo an autoclave cycle, which can cost €80,000-€90,000, prevented a deviation for a wet load, avoided damaging the vacuum pump, and saved water and energy.”

Awards presented in four categories

In all, Fayad added that 250 submitted papers were evaluated before Optimize 26 based on innovation, value delivery and repeatability. And, of the 150 sessions conducted, 12 were picked as finalists for four customer-excellence awards in the modeling and optimization, AI, data strategy, and sustainability categories.

The winners were:

• Phillips 66 for detailing critical challenges and value delivery in modeling and optimization. Its presentation was titled, “From advisory to autonomous: capturing value in midstream through an accelerated APC program.“
• Envision for customer excellence in sustainability with its paper “AspenTech’s asset optimization suite empowers world’s largest off grid green hydrogen ammonia plant.”
• Total Energies in the data strategy category for its presentation “unlocking industrial data value at scale: total energies selected AspenTech in Inmation as the foundation for real time operational excellence”
• Saudi Aramco in the AI category for its paper, “Industrial AI driven, non-linear, modeling improves, refinery, LP accuracy, and blending decisions.”