IIC’s learnings from its AI testbed at deep-learning facility

Dec. 3, 2018
IoT sensors improve efficiency of smart buildings.

The Industrial Internet Consortium (IIC) announced results for the Deep Learning Facility Testbed. IIC-member participants Dell EMC and Toshiba developed an application to explore deep learning through neural networks within an IoT platform to optimize asset utilization in an office building.

The Deep Learning Facility Testbed, located in a Toshiba facility in Kawasaki, Japan, analyzes 35,000 measured data points per minute, working to optimize the maintenance of monitored assets of the building. With this amount of data collected, the testbed relies on artificial intelligence to detect anomalies in order to improve the visitor experience with things like prioritized elevator scheduling and automated temperature and lighting controls.

The structure of buildings varies and it is costly to design an anomaly-detection systems. The testbed application learns what a normal condition would be using data aggregated from many sensors installed in the facility. The team has been able to use the data to determine an unusual condition, locate the suspected device and let staff check if the inference is correct. For example, the testbed detected the unusual state of the air-conditioning equipment in the kitchen, and building facility management staff found out that the air-intake ducts in the kitchen had been closed to avoid odor by kitchen staff.

“IoT is an important enabler for energy efficiency, but a large building complex requires thousands of sensors to track ambient conditions, traffic flow and occupancy,” said Ken Hatano, chief specialist in the Deep Learning Technology Department of Software and AI Technology Center, Toshiba Digital Solutions Corporation. “Buildings also have HVACs, fans, lights and elevators that consume energy. Monitoring all of these sensors and assets is no small task.”

“The deployment of an IoT system for a smart building will maximize the value of big data collection through deep learning analytics,” said Dr. Said Tabet, IIC Deep Learning Testbed lead, and lead technologist for IoT Strategy, Dell EMC. “A smart building will improve operational efficiency, reduce maintenance costs and maximize the use of assets.”