Engineering news
Known as ‘federated learning’, the approach will be used in a project led by the University of Michigan (U-M), which aims to speed up adoption of a new breed of semiconductors in solar energy. Backed by $3m from the US National Science Foundation (NSF), the work includes partners at the University of California San Diego (UCSD).
The project will include hands-on work to improve upon the process of layer-by-layer deposition of semiconductor materials during production.
“Halide perovskites, a class of materials that has been largely developed over the past decade, represent a promising new semiconductor material that can, among other things, boost solar cell efficiency,” a Michigan announcement said. In less than 15 years of study, solar cells using perovskites have increased in efficiency from 10% to 26%, the researchers added.
“What’s amazing is the rapid rate of how perovskites have caught up to silicon,” said U-M associate professor Neil Dasgupta, principal investigator on the project. “From a manufacturing standpoint, they can be less energy intensive to process. You can print them almost like an ink onto materials. They're also very tuneable and customisable.”
This means perovskites can be optimised to capture different parts of the spectrum, the researchers said. It also means that they may ultimately be cheaper to produce.
“Newer technologies like perovskite semiconductors inevitably pit companies against each other in a race to improve performance, streamline manufacturing and bring products to market. But pure competition slows progress down as companies perform similar experiments, covering the same ground,” the announcement said.
The project aims to use an information-sharing approach that boosts cooperation between companies while protecting proprietary information and worker interests. Multiple organisations will feed test results into a predictive model, which could help all parties improve their manufacturing process while protecting trade secrets.
"With something like perovskite manufacturing, you have different sources of data on factors such as the optimal processing parameters," said Raed Al Kontar, U-M assistant professor of industrial and operations engineering. "The question becomes how these different companies that are doing their own research can optimally collaborate and distribute the data they’re collecting through trial-and-error testing."
Engineers at U-M, and their partners at UCSD, will conduct isolated experiments with perovskite semiconductors. Al Kontar will take data collected from each to build predictive models for forecasting product quality and performance – helping to hone key parameters such as optimal pressure and temperature during manufacturing.
Pooling information in this way should reduce costs and allow for faster progress in development, the researchers said. The new process will be a form of ‘cyber manufacturing’, according to the NSF, which "exploits opportunities at the intersection of computing and manufacturing with the potential to radically transform concepts of manufacturing."
U-M's team includes Wei Lu, professor of mechanical engineering, who will lead efforts to model the mechanical and material aspects of the process. Partners at UCSD include David Fenning, assistant professor of nanoengineering, who will lead the solar cell testing and design aspects of the project.
The four-year, $3m grant for the work is part of NSF's Future Manufacturing programme supporting "fundamental research and education of a future workforce to overcome scientific, technological, educational, economic and social barriers in order to catalyse new manufacturing capabilities that do not exist today."
Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.
Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.