EPFL scientists have devised a new technique that involves using artificial intelligence to develop next-generation heat-pump compressors. The technique can reduce the power requirement of the pumps by nearly 25%.
Heat pumps are equipped in around 50%–60% of new homes in Switzerland. These systems suck in thermal energy from the ambient environment — like air, the ground, or an adjacent river or lake — and convert it into heat for buildings.
Although the existing heat pumps usually perform well and are eco-friendly, they still have considerable scope for improvement. For instance, engineers can use microturbocompressors in the place of traditional compression systems to minimize the power requirement of the heat pumps by 20%–25% together with their influence on the environment.
This is due to the fact that turbocompressors are highly efficient and 10 times smaller when compared to piston devices. However, integrating these mini components into the designs of heat pumps is not straightforward; problems emerge due to their very small diameters of less than 20 mm and fast rotation speeds of greater than 200,000 rpm.
A research team at EPFL’s Laboratory for Applied Mechanical Design on the Microcity campus was led by Jürg Schiffmann to develop a technique that enables turbocompressors to be incorporated into heat pumps in a simple and fast manner.
The researchers used a machine-learning method known as symbolic regression to develop simple equations for rapid calculation of the ideal dimensions of a turbocompressor for a specified heat pump. Recently, their study won the Best Paper Award at the 2019 Turbo Expo Conference conducted by the American Society of Mechanical Engineers.
1,500 Times Faster
The technique devised by the research team considerably simplifies the first step in developing turbochargers. In this step, the optimal size and rotation speed for the intended heat pump is roughly calculated. This step is highly crucial because a good initial estimate can significantly reduce the overall design time.
To date, design charts were being used by engineers to size the turbocompressors; however, these charts turn out to be highly inaccurate if the equipment size is smaller. Moreover, the charts have not been updated according to the latest technology.
Hence, Violette Mounier and Cyril Picard, EPFL PhD students, worked to create an alternative. They fed machine-learning algorithms with the results of 500,000 simulations and created equations replicating the charts but with various benefits: they are dependable even for turbocompressors of smaller sizes; they are just as detailed as more intricate simulations and are 1,500 times faster.
The technique devised by the team also allows engineers to skip a few steps in the traditional design methods. It opens the door to easier execution and more extensive application of microturbochargers in heat pumps.
The Advantages of Microturbocompressors
In traditional heat pumps, pistons are used to compress a fluid known as a refrigerant and advance a vapor-compression cycle. It is necessary for the pistons to be lubricated well to function properly; however, the oil can get adhered to the walls of the heat exchanger and hinder the process of heat transfer.
However, microturbocompressors with diameters of only a few dozen millimeters can function even without oil; they rotate on gas bearings at speeds of hundreds of thousands of rotations per minute. The gas layers and rotating movement between the components indicate that the friction is nearly zero. Consequently, these miniature systems have the ability to improve the heat transfer coefficients of the heat pumps by 20%–30%.
This microturbocharger technology has been under development for many years and is now fully developed.
We have already been contacted by several companies that are interested in using our method.
Jürg Schiffmann, Laboratory for Applied Mechanical Design on the Microcity, EPFL
The researchers’ study will enable companies to easily integrate microturbocharger technology into the heat pumps.