Indian Institute of Technology Develops Applications for AI and ML to Solve Engineering Problems

2019-10-29

IIT Madras researchers have developed an algorithm that can address engineering problems in new technologies such as artificial intelligence, machine learning, and deep learning. The researchers will deploy their AI software called "AISoft" in a company to solve engineering problems in areas such as thermal management, semiconductors, automotive, aerospace, and electronic cooling applications.


Artificial intelligence, machine learning, and deep learning have been around for over a decade, but have only been used in traditional fields such as signal processing, speech recognition, image reconstruction, and prediction. There have been limited attempts globally to use these algorithms to solve engineering problems, such as thermal management, electronic cooling industry, automotive issues (such as fluid dynamics prediction in engine hood or internal engine), and aerospace industry (such as fluid dynamics problems in aerodynamics and the entire aerospace field).A group of researchers led by Professor Vishal Nandigana from the Fluid Mechanics Laboratory in the Department of Mechanical Engineering at IIT Madras has developed AI and deep learning algorithms to solve engineering problems. Dr. Vishal Nandigana emphasizes the uniqueness of these algorithms, stating, "We tested AISoft and used it to solve such thermal management problems. We found that it is millions of times faster compared to the existing solutions currently used in this field. Our AI is applicable to any generalized linear and curved geometric input. Our research saves computational time in engineering, which is also a bottleneck for most engineering problems."


The researchers have used data-driven AI and deep learning models to derive solutions for engineering problems after training the AI with datasets. These previous datasets can come from existing big data in relevant engineering industries, where there is a wealth of experimental data. Similarly, if there is no data available to train the AI, commercially available CFD (Computational Fluid Dynamics) software can be used to generate this data on comprehensive small-scale problems.Nandigana adds that their AI is independent of training units and can solve target problems without the need for information from other training units. It can handle sparse datasets and solve engineering problems, making it stand out from commercially available software in this aspect.


IIT Madras researchers have also developed hardware products that utilize GPUs and multi-threading for addressing thermal management problems in the electronic cooling industry. Both the software and hardware products are several times faster than commercial numerical methods software and open-source software available in the market. These algorithms will address many industry-critical problems and can also be used for educational purposes.

RECENT AI NEWS

RECENT AI TOOLS