Representative End User Clients
Representative Automation Clients
Representative Software Clients
Mitsubishi Electric Corporation announced that together with the National Institute of Physical and Chemical Research (RIKEN) it has developed an AI technology that clarifies the rationale underpinning each AI-based control system to enable such systems to be deployed in infrastructure and various equipment with confidence. Mitsubishi Electric will provide the new technology in its Maisart lineup.
Features of Development
1) Step-by-step use of new AI realizes explicit control rationale
By estimating the characteristics of the equipment's operating environment, the new AI quickly identifies and quantifies physical parameters using simulation rather than taking measurements with numerous sensors. Past working data can be used to learn equipment sensor values and physical quantities in order to predict future changes in the operating environment.
Using predicted values and specified parameters, the simulation accurately predicts changes in the operating environment and then creates an optimal control plan in the scheduler.
By visualizing predicted values, work environment changes and a control plan, the new AI clarifies the control rationale to eliminate black boxes. By increasing the confidence level for equipment control and operational confirmation, the new AI technology will enable AI-equipped infrastructure, air-conditioning equipment, etc., to be used confidently and allow managers to clarify the basis of their AI control, such as when responding to complaints.
2) New AI clarify causes of equipment malfunction
The new AI technology determines physical parameters and compares actual past values with predicted or planned values to visualize deviations. If a malfunction occurs, it identifies any deviation from predicted values and then uses physical parameters to identify the cause of malfunction due to controls not being implemented as planned.
In addition, if the control system operates as planned but produces unexpected results, equipment abnormalities or changes in the operating environment can be quickly recognized, thereby allowing maintenance and recovery operations to be performed before an actual failure occurs.