Drives and Controls Magazine
‘First’ AI-based diagnostics will cut downtime and boost productivity
Published:  15 August, 2019

Mitsubishi Electric claims to have developed the world's first diagnostic technology that uses machine learning to analyse sensor data, thus detecting machine abnormalities rapidly and accurately, and helping to improve productivity. The development is based on the company's Maisart  AI (artificial intelligence) technology.

Unlike conventional diagnostic technologies, which use the same conditions to detect abnormalities, regardless of a machine's current operating state, the new technology analyses sensor data to generate a model of the machine’s transition between various states. It then uses this model to set optimal conditions for detecting abnormalities. Mitsubishi says that the technology will detect signs of machinery failure before breakdowns occur, helping to reduce the costs associated with performing maintenance and dealing with downtime.

The new technology also eliminates the need for complicated manual settings, thus making it easier to implement abnormality-detection systems.

The conventional approach to detecting machine abnormalities (top) compared to the new AI-based technique (bottom)

Mitsubishi says that with maintenance regimes shifting from time-based periodic measurements to condition-based techniques, the need for accurate abnormality-detecting technologies is growing. The way that machines are used is also changing, in turn affecting advance signs of machine abnormalities which depend on a machine’s usage and state of operation.