The
researchers utilised a data-driven AI and a deep learning model to arrive at
solutions for engineering problems after training the AI with data sets .
The Indian Institute of Technology,
IIT-Madras on Monday said its researchers have developed algorithms that enable
novel applications for artificial intelligence (AI), machine learning and deep
learning to solve engineering problems.
The researchers are going to establish
a start-up to deploy their AI Software called ‘AISoft’ to develop solutions to
engineering problems in varied fields such as in thermal management,
semiconductors, automobile, aerospace and electronic cooling applications. “We
tested AI soft and used it to solve such thermal management problems. We found
it to be nearly million-fold faster compared to existing solutions currently
used in the field,” said Vishal Nandigana, Assistant Professor, Fluid Systems
Laboratory, Department of Mechanical Engineering. “Our AI works on any
generalised rectilinear and curvilinear input geometry. Our research saves the
computational time, which is the bottleneck to solve most engineering problems,
Nandigana added.
The researchers utilised a
data-driven AI and a deep learning model to arrive at solutions for engineering
problems after training the AI with data sets. These prior data sets can be
from existing big data in the relevant engineering industry where there are
lots of experimental data available..
Also, if data is not available for
training the AI, it can be generated using commercially-available CFD
(Computational Fluid Dynamics) software on small independent pieces of the
full-blown problem. This idea is new and is only now being looked at by a few
research groups across the world. Most of these research groups use
Convolutional Neural Networks (CNN) or C-GAN (conditional generative
adversarial network) to solve engineering problems. They have also developed
hardware products using graphics processing unit (GPU) and multi-threading
processing to solve thermal management problems in thermal and electronic
cooling industries..
Both the software and hardware
products are several times faster than commercial numerical method software and
open source software in the market. These algorithms will solve a lot of
pressing problems for industries and can also be used for educational purposes
