No-Code ML for Citizen Data Science

PTT Global Chemical (PTTGC) is Southeast Asia’s leading integrated petrochemical and refining company covering manufacturing and distribution of upstream, intermediate, and downstream petrochemical products.

PTTGC wants to enable both front-line engineers and operators, as well as its central advanced analytics team, to rapidly deliver AI-driven solutions to increase profits and efficiency across its operations. PTTGC turned to BHC3 Ex Machina and BHC3 support teams for help.

• Get analysts and engineers productive quickly
• Rapidly scope and deliver use cases
• Generate business value across the refinery
• Create executive awareness for the value of advanced analytics

Prior to engaging BakerHughesC3.ai, engineers and operators at PTTGC relied on a mixture of analytics tools and spreadsheets to analyze data for optimized refinery operations.

PTTGC’s approach to process giveaway is an example. Refineries aim to produce an optimal mix of products at the lowest cost. “Giveaway” occurs when a refinery produces a higher-grade product than is specified for downstream processing. Refineries aim to minimize giveaway to maximize profit.

An optimal giveaway solution would aggregate operational, sensor, and lab test data, the key variables that determine output grade, and predicts and tunes operating conditions to optimize these variables.

Engineers in the refinery did not have a solution sophisticated enough to handle the diversity and volume of data required or intuitive enough to be used by non-coders.

As a result, operators at PTTGC’s refineries addressed the problem by compiling data in Excel and manually reporting on giveaway associated with a few key variables—a manual process that engineers could only complete monthly or even quarterly.

After a brief introductory training, the process engineer in charge of reporting built a BHC3 Ex Machina workflow that ingests operational data from OSI Pi, calculates giveaway, and reports on the results. This reporting is more accurate than the previous manual calculation and can be run with the click of a button.

The engineer then configured an ML model to predict process giveaway three to seven days in advance, enabling operational adjustments to be made in order to manage giveaway.

“Before BHC3 Ex Machina, it took hours of work to keep up with minimum reporting,” said the engineer responsible for process giveaway. “Now, I can complete reports with one click. But more than that, BHC3 Ex Machina lets me run predictive analyses that help us operate more efficiently. I’m delivering better outputs, using a data-driven approach that wasn’t available to me before.”

PTTGC has since trained an additional 18 individuals on BHC3 Ex Machina who will use the product to deliver more accurate reporting, faster analytic development timelines, and predictive recommendations to improve process operations across its refineries.

• Projects save analytic time and effort, deliver process optimization recommendations 3–7 days in advance
• 4 use cases delivered and 1 in development
• Projects expected to deliver at least $0.5M in recurring annual economic benefit
• 18 PTTGC project owners trained on BHC3 Ex Machina

For PTTGC, BHC3 Ex Machina provides a clear path to productizing all workflows by leveraging the model management, application development, and security, operations, and governance capabilities of the underlying BHC3 AI Application Platform.

Over the next few years, PTTGC will leverage BHC3 technology and domain expertise to protoype and productionize hundreds of robust AI applications to address their highest value use cases.

“This is a new way of working, combing both predictive models and business workflows for driving operational improvements. With BHC3 Ex Machina our business units can learn quickly, and drive rapid improvements supported by BakerHughesC3.ai. In addition, the central advanced analytics team can reuse successful approaches, and scale them to deliver benefits across the company.” —Pitak Lausangngam | Head of Digital Transformation, PTTGC