Why this article is worth reading:
- Albermarle deployed AI across legacy infrastructure, including plants built 70 years ago, with strategies for standardizing data and scaling analytics across global manufacturing operations.
- Aveva’s PI System transformed weeks of manual data analysis into fast AI-powered insights, freeing up engineering time for innovation instead of firefighting.
- Albermarle’s focus on people and training delivered tens of millions of dollars in ROI.
For manufacturers faced with increasing production demands, Albemarle's approach to AI offers valuable lessons in scaling data-driven optimization. As one of the world's largest lithium producers, Albemarle transformed how it extracts value from manufacturing data — achieving more than $50 million in annual savings through systematic deployment of AI analytics.
Driving Albermarle’s approach to AI is the huge growth expectation for lithium. In 2023, electric vehicles (EV) sales reached 15.7 million. By 2030, this figure is expected to jump to 46.9 million. Lithium is a primary ingredient in EV batteries and these batteries need a lot of it. For comparison, consider that the battery in your cellphone contains about a gram of lithium — an EV battery contains 10,000 grams.
Jonathan Alexander, manufacturing AI and analytics manager at Albemarle, explained, "We’re focused on meeting the growing demand for lithium. As a lithium producer, this means you can either mine more or make your processes more efficient. The latter is what we've focused on." This efficiency-first approach led Albemarle to develop what it calls "Albemarle Intelligence," an AI-powered system built on the Aveva PI System that serves as the data infrastructure and analytics engine across the company's global manufacturing network.
Albemarle's system centers on event frames, a core feature of Aveva PI System that allows users to sample time-series data and convert it into datasets for advanced analytics. Event frames use start and end triggers to isolate specific manufacturing events — such as startup, shutdown, running or idle — across process phases. This enables comparison of current performance against historical data to identify optimization opportunities.
Scaling AI across legacy infrastructure
The company has created more than 3,800 reusable asset framework templates for heat exchangers, reactors, pumps, distillation columns and other equipment. These templates are visualized through more than 1,000 Aveva PI Vision displays and supported by more than 9,000 custom formulas and analytics.
From over 20 million event frames, Albemarle has derived more than 500 datasets, 20,000 statistical process control charts and 500 machine learning models deployed across its operations sites.
Alexander's personal experience as an engineer at Albermarle illustrates the transformation AI analytics presents to manufacturers. "I started out as maintenance engineer and advanced to become a process engineer. In this role, we would pull up process trends every morning looking for insights. It took weeks to interpret this data. All the engineers would do this and collaborate on what patterns and signals we saw," he said. "If AI can quickly find those insights for you, that's a major change. But you still have to determine what actions you'll take. The difference is that now you spend time looking to add innovations around these findings rather than firefighting."
Alexander explained the company’s 2019 strategic AI commitment: "We made a global AI program commitment with the Aveva PI system as a core component. We trained our AI on equipment using our manuals and procedures. Some of our plants were built 70 years ago so we had to deal with legacy technologies."
The Aveva PI Asset Framework system enabled Albermarle to do data contextualization and curation at scale. "Once we standardized our data infrastructure, we could then add standardized analytics and algorithms at scale using machine learning for predictive maintenance. We relied on machine learning to find outliers in production factors such as temperature and pressure."
The criticality of workforce buy-in Despite the clear advances the Aveva PI technology provided to Albermarle, an important lesson emerged.
"Not long after implementation, we started to ask about the ROI (return on investment) we expected to find. We discovered, however, that workers weren't using the technology," Alexander said. "So, we spent time with them to make sure they knew how to work with the new technology and understood how it would affect their jobs."
The results were immediate and dramatic. "After this training, we didn't make any changes to the technology, but we immediately began getting ROI — in the tens of millions in annual savings without changing the tech."
Emphasizing Albermarle’s human-centered approach here, he said, "We first focused on turning data into insights, but those are really just opportunities because they don't impact business until they get implemented. That's why we centered the project on people to get the insights implemented."
Albemarle has now trained more than 1,000 operators and 300 engineers on the new system, enabling delivery of more than 200 improvement projects to date.
Albermarle’s quantified results
The scale of improvements experienced at Albermarle demonstrates the potential for similar manufacturing operations:
- Equipment failure prevention — A single statistical process control chart analyzing three months of data identified equipment failure patterns, saving $500,000 annually.
- Capacity — Process optimizations deliver $450,000 in annual savings.
- Quality — Advanced analytics enable $850,000 in annual savings.
- OEE (overall equipment effectiveness) — A unit operation plagued by issues for 20 years was stabilized using real-time machine learning monitoring 300 variables, resulting in $1 million annual savings and 75% reduction in environmental incidents.
Alexander noted many more improvements have been made with the help of Aveva PI, leading Albemarle to estimate savings of at least $50 million every year.
Implementation lessons
Albemarle's experience offers several key insights for manufacturing engineers and plant managers:
- Standardize first: Establish consistent data infrastructure before deploying analytics.
- Focus on people: Technology adoption requires comprehensive training and change management.
- Scale systematically: Develop reusable templates and frameworks for consistent deployment.
- Measure impact: Track both operational and financial metrics to demonstrate value.
- Start with known problems: Apply AI to well-understood processes where an expert can validate results.
