'Secondary flow' - learning fluid mechanics in a holiday version

You may have heard or have heard the phrase "secondary flow" before. If you want to know what it is, let me explain the mentioned phenomenon in a holiday setting. This article is a summary of a project carried out as part of the Miniatura programme, funded by the National Science Centre.

What are secondary flows?

We're in the middle of the summer holidays, so some of us are probably out canoeing right now. A mainstream flow is the current of a river flowing along its course - from source to mouth. But anyone who has paddled a river knows that the water doesn't actually flow in a regulated stream along the riverbed. In addition to the main current, it often moves in a lateral direction. The cause of this movement can be an obstacle - a stone, a stump or a fault. Their occurrence affects the river current even many metres beyond the obstacle itself. In extreme cases, secondary flows can cause whirlpools and backflows, so-called "backwaters". Secondary flows associated with centrifugal force are also responsible for so-called river meandering.

Similarly, gas behaves in the channel of a flow compressor. "Mainstream" flow has an inlet to outlet direction and is the most desirable regulated flow. Outside of it, many types of secondary flows are observed. The gas under the influence of numerous physical phenomena (centrifugal force, Coriolis force, over-blade flows) can move in a transverse direction. Secondary flows are mostly undesirable - they cause energy losses and consequently reduce compressor efficiency.

Secondary flows can also arise due to small irregularities in the shape of the compressor channel; small ones are caused by surface changes or minor nicks. The aim of our project is to refine how to detect them. We do this by using mathematical algorithms that analyse pressure signals.

Let's go back to the river analogy. In this case, our project would mean that we monitor the pressure on the left bank of the river. From this measurement, we draw conclusions about whether any new objects have appeared near the right bank, disrupting the flow. Sounds like a big challenge - the algorithm has to be very sensitive, but also able to isolate a specific phenomenon. In other words, it must be able to distinguish between a change created by the appearance of a fixed obstacle and a change caused by other factors.

Why is this important?

Basic research is motivated by curiosity and the desire to understand a phenomenon more deeply. But in the longer term, future areas of  application resulting from this project can be imagined. A change in the geometry of the channel could be due to wear and tear on the machine or chipping of the blade. If we can identify small changes in pressure readings caused by such a situation, we can design systems to monitor the operating condition of such a machine.

Many condition monitoring and failure prediction systems based on artificial intelligence (AI) algorithms are currently being developed. This is a promising direction towards lower machine failure rates. The problem with AI is that it does not provide a physical explanation. An algorithm can tell us that a machine is about to fail, but it will not explain why it 'thinks' so. For this reason, it is important to continue basic research into the physical origins of phenomena, which together with AI algorithms can provide a complete system for monitoring and predicting compressor performance.