AI Enables Industrial Autonomous Operations

According to industry surveys, a growing number of manufacturing organizations are adopting artificial intelligence (AI). AI has grown into a broad category that comprises several technologies such as computer vision, machine learning (ML), and natural language processing (NLP).

Commanding the hype most recently have been generative AI applications such as ChatGPT, which build upon NLP to analyze and generate language-based data such as text and speech. Although this technology appears poised to revolutionize industrial operations, it is far less applicable to industrial process control.

Meanwhile, another AI technology, reinforcement learning, has proven successful for direct control of a plant. In March of 2023, ENEOS Materials Corporation and Yokogawa announced an agreement in which Factorial Kernel Dynamic Policy Programming (FKDPP), a reinforcement learning-based AI algorithm, will be officially adopted for use at an ENEOS Materials chemical plant. The agreement followed a successful field test in which a solution, “autonomous control AI” demonstrated a high level of performance while controlling a distillation column at the plant for nearly a year. This is the world’s first example of the adoption of reinforcement learning AI for direct control of a plant.

Experts in the manufacturing and process industries have estimated that more than 65% of process control loops are underperforming and up to 30% are operating in manual mode. For companies striving toward autonomous operations with little or no human presence at plants, underperforming controls and manual operations have emerged as major issues.

The AI solution controls distillation column operations that were beyond the capabilities of existing technologies such as proportional-integral-derivative (PID) loop control and advanced process control (APC). That had necessitated manual operations based on the judgements of experienced plant personnel.

This session will describe the autonomous control AI development process and how it addressed the challenges presented by difficult and complex process operations. The autonomous control AI eliminated off-spec production and reduced costs in terms of feedstocks, fuel, and labor. The technology also reduced steam consumption and CO2 emissions each by 40%. Ultimately, it enables alignment between management and operations as well as industrial autonomy.