In January, specialty materials maker Celanese Corp. (Irving, Tex.; www.celanese.com) beta-launched a remote-operations control room (ROCR) that allows hands-free queries of plant data for faster data gathering and decision-making. The ROCR leverages recent advances in generative-artificial intelligence (gen-AI)-powered natural language processing.
In the ROCR, large-language models (LLM) overlay the plant’s data, such that the AI-driven language models enable users to request and manipulate plant data and information using only conversational language. Inside the ROCR, a 10-ft x 6-ft multi-screen display shows the current plant situation, and, using verbal requests and commands, plant operators can ask the system to retrieve asset and process information to address problems or optimize operations.
Located at the Celanese facility in Clear Lake, Tex. and developed in partnership with industrial software maker Cognite AS (Oslo, Norway; www.cognite.com), the ROCR provides a “centralized view of Celanese’s contextualized industrial data, alongside a deterministic generative AI copilot,” Cognite says. “This allows Celanese to find cross-data source insights to understand and solve safety, reliability and quality risks across the operation in real time,” according to the companies.
Initial impressions of the ROCR beta launch have been positive. Celanese digital manufacturing director Ibrahim Al-Syed says the biggest benefit has been from increased productivity. “The system has cut down on the time required for making decisions and solving problems that arise,” Al-Syed says.
The ROCR beta launch represents the culmination of two years of work on data contextualization that allows the LLM to identify the correct data and manipulate it in a way that communicates the information that is needed.
Following the successful rollout, the partners are working to scale the capabilities to multiple locations and refine its capabilities.