Yokogawa Electric Corporation has released a new version of CENTUM with built-in AI functionality. This product belongs to the OpreX Control and Safety System product range and provides AI control functionality. It also provides AI estimation functionality. The product name has also changed from CENTUM VP to CENTUM.
Autonomous control AI:
Yokogawa’s autonomous control AI is used for complex processes that are difficult to automate, which were previously manually adjusted. It is also used for control processes requiring balancing multiple conflicting requirements. Even with high level disturbances, appropriate control actions are automatically decided and executed by the AI. Since it was introduced at a chemical plant in 2022, the control has been fully automated for over 35 days. Stable operation has since been achieved for over 4 years, achieving improved energy efficiency and reduced environmental impact. The reinforcement learning technology behind this solution, called Factorial Kernel Dynamic Policy Programming (FKDPP), was developed jointly with the Nara Institute of Science and Technology (NAIST).
AI control models:
Control models created using the autonomous control AI can be implemented in the CENTUM system. Once the models are downloaded from the CENTUM engineering environment, they can be monitored in the familiar CENTUM monitoring/operation environment in the form of graphics, trends, alarming and faceplates. Before implementation, the AI has been repeatedly trained using a plant simulation, to create an optimal control model. Once the customer confirms that the model is reliable, it is implemented in plants with safety precautions in place.
Soft Sensors for quality-related values:
AI Estimation provides soft sensor functionality to estimate values, such as constituent concentrations, based on process variables such as temperatures, flow rates and pressures. These values are valuable for confirming product quality and conditions within the plant, but can be difficult to measure in real time and with high frequency. Some of these values need to be measured in laboratories. Using machine learning, these values can be estimated. The contribution of each process variable to the final result can be seen. Estimated values can be viewed along with other process information in the CENTUM monitoring/operation environment.
Main application industries include Refining/Petrochemicals, Chemicals, Power, Gas, Pulp & Paper, Pharmaceuticals, Food, Iron & Steel, Metals, Water/Wastewater and Cement.








