An axial compressor for a domestic appliance can be designed to be smaller than an equivalent centrifugal compressor. However, the performance of such a compact axial compression system is limited by increased viscous losses and reduced flow turning at low Reynolds numbers ($Re$).

In domestic appliance compressors, $Re$ is typically in the range $104$ - $105$. Although the aerodynamics of isolated aerofoils operating at these $Re$ have been studied extensively, the flow fields within low $Re$ axial compressors have not been investigated in detail. This dissertation aims to develop an improved understanding of loss variation at low $Re$ and to explore how the losses can be reduced through design changes.

Experiments on a 5 times scaled-up single stage axial compressor have been conducted across a range of $Re$ of $104$ - $105$. The flow field has been characterised using detailed area traverses with a miniaturised five-hole probe at the rotor inlet, rotor exit and stator exit and a miniature hot-wire at the rotor exit. The probe was specifically designed and calibrated for the scale of the experiments and methods to improve the accuracy of the measurements have been applied including a probe geometry correction. The traverse experiments were performed at the design operating condition ($\varphi =0.55$ and $Re=6\times 104$) and at a condition close to stall for a datum stage design, a stage with an improved stator design and two stators with compound lean.

It was found that losses in the rotor were greater than the stator losses across the whole range of $Re$. As expected, the loss decreased with increasing $Re$ for both the stator and rotor. The losses were also increased by three-dimensional flow, with typical loss coefficients at the hub and tip of the blade rows in the range of $20-30\%$.

A major contributor to the rotor loss was an unexpected hub separation that increased in size as $Re$ was reduced. At higher $Re$, the major loss sources were the rotor tip leakage, the stator wake and the stator hub separation. The results indicate that an improved stator design that accounts for the actual, measured, rotor exit flow field at low $Re$ could reduce the $Re$ at which blade row losses start to rise dramatically as well as reduce the loss across all $Re$.

The improved stator design was better matched to the radial distribution of rotor exit flow angle, which led to a decrease in stator loss across all $Re$. For all stator designs, however, the measured stage stall margin was identical at all $Re$. This, along with the increase in velocity deficit in the rotor tip region at off-design indicates that stall occurred in the rotor and was neither $Re$ nor stator design dependent.

The introduction of compound lean to the the stator design had the expected result of decreasing the endwall corner separation loss and increasing midspan losses. The experiments have shown that there was a loss increase in both the midspan and casing region much greater than the corresponding decrease in the stator hub. Also the mass flow redistribution in the experiments was larger that the redistribution predicted by the CFD.

Three-dimensional RANS computations at low $Re$ of the same designs as experimentally studied were also conducted in order to investigate the predictive accuracy of industry standard CFD. The simulation results predicted the overall loss distribution but overestimated the end-wall losses and failed to capture the drop in stage performance at low $Re$. The differences with the experiments were caused by the inherent limitations of a fully turbulent solver that cannot reproduce transitional flow-features. Similarly to the experiments, there was no stall margin dependency on $Re$ in the simulations.

This thesis has shown that with axial compressors designed specifically for low $Re$, the $Re$ at which the losses start increasing exponentially can be shifted from $10\times 104$ to $ 4\times10^4$. The loss increase is predominantly caused by the rotor hub corner separation.