HomeNanotechnologyNew machine-learning simulations cut back power want for masks materials, different supplies

New machine-learning simulations cut back power want for masks materials, different supplies


Nov 01, 2022 (Nanowerk Information) Making the numerous numbers of N95 masks which have protected hundreds of thousands of People from COVID requires a course of that not solely calls for consideration to element but additionally requires a number of power. Most of the supplies in these masks are produced by a way known as soften blowing, during which tiny plastic fibers are spun at excessive temperatures that necessitate the usage of numerous power. The method can be used for different merchandise like furnace filters, espresso filters and diapers. Due to a brand new computational effort being pioneered by the U.S. Division of Vitality’s (DOE) Argonne Nationwide Laboratory together with 3M and supported by the DOE’S Excessive Efficiency Computing for Vitality Innovation (HPC4EI) program, researchers are discovering new methods to dramatically cut back the quantity of power required for soften blowing the supplies wanted in N95 masks and different functions. Presently, the method used to create a nozzle to spin nonwoven supplies produces a really high-quality product, however it’s fairly power intensive. Roughly 300,000 tons of melt-blown supplies are produced yearly worldwide, requiring roughly 245 gigawatt-hours per yr of power, roughly the quantity generated by a big photo voltaic farm. By utilizing Argonne supercomputing assets to pair computational fluid dynamics simulations and machine-learning methods, the Argonne and 3M collaboration sought to scale back power consumption by 20% with out compromising materials high quality. The soften blowing course of makes use of a die to extrude plastic at excessive temperatures. Discovering a solution to create an identical plastic elements at decrease temperatures and pressures motivated the machine-learning search, mentioned Argonne computational scientist Benjamin Blaiszik, an creator of the research. “It’s type of like we try to make a pizza in an oven — we’re looking for the correct dimensions, supplies for our pizza stone, and cooking temperature utilizing an algorithm to reduce the quantity of power used whereas conserving the style the identical,” he mentioned. By utilizing simulations and machine studying, Argonne researchers can run a whole lot and even hundreds of use circumstances, an exponential enchancment on prior work. “We’ve got the power to tweak issues just like the parameters for the die geometry,” Blaiszik mentioned. “Our simulations will make it doable for somebody to make an merchandise at an precise industrial facility, and our pc can let you know about its potential for real-world functions.” The simulations present key insights into the method, a way to evaluate a mix of parameters which are used to generate knowledge for the machine-learning algorithm. The machine-learning mannequin can then be leveraged to finally converge on a design that may ship the required power financial savings. As a result of the method of constructing a brand new nozzle could be very costly, the data gained from the machine-learning mannequin can equip materials producers with a solution to slender right down to a set of optimum designs. “Machine-learning-enhanced simulation is one of the best ways of cheaply getting on the proper mixture of parameters like temperatures, materials composition, and pressures for creating these supplies at top quality with much less power,” Blaiszik mentioned. The preliminary mannequin for the melt-blowing course of was developed via a sequence of simulation runs carried out on the Theta supercomputer on the Argonne Management Computing Facility (ALCF) with the computational fluid dynamics (CFD) software program OpenFOAM and CONVERGE. The ALCF is a DOE Workplace of Science person facility positioned at Argonne.



RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular

Recent Comments