Building Industrial Intelligent Engine, Reshaping a New Paradigm of Industrial Applications: SUPCON Unveils Groundbreaking Time-series Pre-trained Transformer (TPT)

HANGZHOU, China, June 7, 2024 /PRNewswire/ -- On June 5th, SUPCON unveiled its groundbreaking TPT (Time-series Pre-trained Transformer), the first AI time-series model, in Singapore, introducing it to global clients and initiating its full industrial application. Dr. Wenyue Ding, SUPCON's Senior Researcher Scientist, presented the product under the theme of "Pioneering Industrial AI New Frontiers".

TPT is based on a generative AI algorithm framework and integrates vast data on production, operations, processes, equipment, and quality across industries during its training. It significantly improves modeling efficiency, unifies industrial applications, and demonstrates adaptive capabilities across different units and conditions, enabling highly accurate and reliable closed-loop applications. TPT stands ready to lead the acceleration of comprehensive intelligentization in the process industry.

Deeply Dive into Core Industrial Scenarios

In recent years, AI has garnered global attention, ushering in an AI era with the advent of Artificial Intelligence Generated Content (AIGC). We are now exploring the application of AI in industrial scenarios to improve production efficiency and address industry challenges.

However, the industrial operation process is predominantly recorded and characterized by time-series data, while existing AI models fall short in addressing the core challenges like abnormal warnings, operation and control optimization, and bottleneck analysis. SUPCON's groundbreaking TPT, the first time-series pre-trained transformer in the process industry, will revolutionize data application and drive sustainable industrial development as a new generation of industrial intelligent engine. Furthermore, this marks a remarkable milestone in SUPCON's pursuit of establishing its "AI+data" advantage and advancing AI innovations and implementations.

Multi-Ability, Generalized, and Reliable

SUPCON's TPT features three characteristics: multi-ability, generalized, and reliable.

Multi-Ability: TPT integrates simulation and prediction for various industrial tasks by training with diverse cycle predictions and dynamic steady-state simulations, effectively handling complex scenarios by unifying analysis, optimization, and control.

Generalized: TPT gathers massive data for pre-training and learns industrial operation laws. It can be reused across different units and conditions, demonstrating strong adaptation capabilities.

Reliable: TPT offers higher accuracy and inclusiveness than traditional AI models, enabling precise control for optimized, closed-loop production in industrial units.

Reshaping a New Paradigm of Industrial Applications

In the industrial sector, TPT will totally revolutionize traditional fragmented industrial applications and data, transforming from the traditional mode of "N models + N Applications" to a new paradigm of "TPT + 1 software" in the plant.

SUPCON's TPT applications have made breakthroughs in various units, spanning chlor-alkali, thermal power, and petrochemical plants. TPT equips these production units with an "intelligent brain" that enables them to autonomously "think" and "communicate" like experts, while monitoring and optimizing unit operations. This achieves the goals of improving efficiency, stable operation, and increasing profits, ultimately solving the unsolved challenges.

SUPCON is currently accelerating its efforts in AI technology, aiming to leading industrial AI growth with its new mission and vision. The company is committed to promoting sustainable industrial development with AI and embarking on a new era of the process industry towards AGI. Undoubtedly, the global launch of TPT and its subsequent large-scale industrial applications mark a significant milestone in this great transformation journey.