Newsroom: Highlights

Bookmark and Share Email Print

07.26.18 - Got Data


Should you find yourself inside a cereal-processing plant in the not-too-distant future, you may notice − along with conveyor belts moving heaps of grain to be steamed, dried, shredded, baked or toasted − that a revolution is underway. 

Motors running the conveyor belts are deciding when production lines will speed up, slow down or stop. The machinery gets real-time data on environmental factors, schedules, staffing, incoming orders and other variables that affect productivity and energy use. Without any human input, the motor processes this information and decides what action to take.

This is all possible due to a microprocessor chip embedded in the conveyor-belt motor. About the size of a postage stamp, the chip serves up a wondrous bit of programming based on collected data and a machine-learning algorithm. The algorithm provides the motor with a “model of the world,” which is packed with every contingency it might encounter in its environment and a correct response to each.

 Is it humid today? What is the weight difference for each grain due to humidity? How much more or less energy is needed?

The motor also monitors its own performance specifications. Is this vibration excessive? Is the bearing lubricant too hot? These are all questions the motor can answer and act on.

It's not only motors that are smart; sensors, actuators and countless other pieces of machinery are "thinking" and acting on their own. All of this activity is generating massive amounts of data as every movement and observation is recorded and stored. This is the Industrial Internet of Things (IIoT).  

Smart devices and systems are operating in thousands of factories, processing plants and warehouses. As industries become more adept at working with data and connecting to IIoT, manufacturers will increasingly rely on machine learning and smart applications to help deliver maximum productivity and energy efficiency.  

CALIT2 Director G.P. Li is prepared to advance these types of energy-saving, smart manufacturing applications when CALIT2’s new Data Engineering Infrastructure Lab opens this summer.

Housed on the second floor of the CALIT2 Building, the 3,000-sq.-ft., state-of-the-art high-performance computer lab will be equipped with neural network processors, collaboration tools, visualization software and a brand new class of hardware designed expressly for machine learning, AI (artificial intelligence) and deep learning exploration. Faculty, students and industry scientists will have a playground for experimenting, Li said.  

The lab was funded by UC Irvine, CALIT2 and CESMII (the Clean Energy Smart Manufacturing Innovation Institute). In 2016, The U.S. Department of Energy established CESMII to work with industry to reduce the energy cost of manufacturing through the application of smart manufacturing technologies. CALIT2 is the CESMII Southern California Regional Demonstration Center.

"CESMII's purpose is to develop a program to help manufacturers across all sectors take advantage of the explosion of data from IIoT that's available to them," said Richard Donovan, director of research development for UCI’s information and computer science, and engineering schools. "There is all this data being generated in these plants so now the question is, ‘How can I use this data to make manufacturing more energy efficient?'"

Donovan, who worked with Li to design the lab, noted a common obstacle for small and medium-sized companies. "The technology is so new, many of these companies don't fully understand what to do with the data," he said.   

To meet the needs of CESMII companies, the lab will facilitate heavy calculations required to create reduced-order models, Donovan said.  [Reduced-order models are scaled-down versions of physics-based simulation results and real sensor data that can provide the system with a model of the world from which it can make decisions (remember the conveyor belt motor in the cereal plant?)]. These simple programs use little energy. They don’t take a lot of memory or need a complex processor, Donovan said. "Any time the device wants to know something, it will say, ‘Here's the current state, tell me what will happen next,’" he added.

CESMI will be the first project to utilize the data lab, but the facility will also take on challenging data projects across many domains. Configured with advanced networking infrastructure, and computer hardware and software specifically designed for large-dataset machine learning, it will offer clients fast processors, lots of memory, and sophisticated computer programs for molecular dynamics, computational fluid dynamics, or sophisticated heat-transfer simulations, to name a few.

Li calls this explosion of data the fourth generation of the industrial revolution. "We are creating an ecosystem for these efforts," he said. “This latest revolution is marked by breakthroughs in emerging fields such as analytics, artificial intelligence and IoT, and builds on the digital revolution era started in the 1980s,” he said.

“The Data Engineering Infrastructure Lab will position UCI and CALIT2 to provide essential services as we embark on this new era.”

– Sharon Henry