High-pressure ratio centrifugal compressors’ diffusers face challenges from high-velocity, non-uniform flow at the impeller outlet, decreasing efficiency and stall margin. To address this, this paper presents a novel vaned diffuser passage design method that successfully improved the compressor’s performance. An optimization method using axisymmetric hub contours and NURBS curves was applied to modify the diffuser’s design. After optimization, centrifugal compressor peak efficiency increased by 0.78%, and stall margin expanded from 12.8% to 20.4%. Analysis at the peak efficiency point showed loss reduction mainly from decreased recirculation and mixing losses in the diffuser’s vaneless and semi-vaneless spaces. Furthermore, correlation analysis and Mach number distribution revealed that flow behavior at the diffuser’s leading edge significantly influences efficiency. Consequently, design principles emphasize satisfying specific Mach number distribution rules at the diffuser’s leading edge under certain inflow conditions for optimal performance.

To explore the internal flow mechanism and improve the performance of a supersonic compressor, an efficient global optimization design method was developed for an axial flow compressor and applied in the optimization design of a prototype supersonic compressor. Based on the multiple circular arc (MCA) blade parameters, the method can be used to parameterize the elementary stage of the blade. The optimized solution is obtained by changing the elementary stage and stacking lines of the blade during the optimization process. It has the advantages of fewer optimization variables, strong physical intuition, and a smooth surface. The optimization results show that a change in the rotor blade shape parameters has a significant effect on the compressor efficiency under design conditions, while a change in the skewed-swept parameters of the stator is the main factor that improves the compressor’s performance under near-stall conditions. Further numerical results show that the optimized rotor changes the form of the shock, weakens the degree of shock boundary layer interference, inhibits the radial migration flow of the supersonic rotor, reduces the loss of the rotor blade top, and improves the performance of the compressor under design conditions. The stator’s optimization restrains the generation of a concentrated shedding vortex at the root of the blades and greatly improves the stall margin of the compressor. Finally, the total pressure ratio and flow rate are less than 1% of the values based on the prototype operating conditions, the design mass flow of the optimized high-load supersonic compressor is increased by 0.25%, the isentropic efficiency is increased by 1.05%, and the stall margin is enhanced by 3.5%, thus verifying the effectiveness of the optimization method.

The research is based on a full 3D model of a medium variable bypass ratio compression system. The effects of core-driven fan stage (CDFS) VIGV adjustment and forward variable area bypass injector (FVABI) adjustment on the internal flow and component matching characteristics of the compression system during the mode transition process are investigated. The findings of the study reveal that during the mode transition period, the adjustment of variable-geometry components has a negligible impact on the aerodynamic performance of the fan. The closure of the CDFS VIGV has a significant effect on reducing the pressure ratio of the CDFS but only a minor effect on its efficiency. Additionally, the efficiency and pressure ratio of the high-pressure compressor (HPC) increase with the closure of the CDFS VIGV, particularly the second stage of the HPC. When the CDFS VIGV is closed by 30°, the efficiency of the second stage of the HPC increases by 3.4%, while that of the first stage only increases by 3.1%. Under the high bypass mode, increasing the FVABI opening will move the matching point of the CDFS towards the choke boundary. When the FVABI opening increases from 0.30 to 0.58, the efficiency of the CDFS decreases by 3.6%, while the efficiency of the HPC only increases by 0.4%. In summary, closing the CDFS VIGV and increasing the opening of the FVABI will increase the bypass ratio of the compression system, and the former has a more significant impact. When the compression system transitions from the low bypass mode to the high bypass mode, opening of the FVABI should be appropriately reduced to prevent CDFS matching in the operating condition too closely to the choke boundary, thereby maintaining the efficiency of each component of the compression system at a relatively high level.

The phenomenon of data distribution evolving over time has been observed in a range of applications, calling the needs of adaptive learning algorithms. We thus study the problem of supervised gradual domain adaptation, where labeled data from shifting distributions are available to the learner along the trajectory, and we aim to learn a classifier on a target data distribution of interest. Under this setting, we provide the first generalization upper bound on the learning error under mild assumptions. Our results are algorithm agnostic, general for a range of loss functions, and only depend linearly on the averaged learning error across the trajectory. This shows significant improvement compared to the previous upper bound for unsupervised gradual domain adaptation, where the learning error on the target domain depends exponentially on the initial error on the source domain. Compared with the offline setting of learning from multiple domains, our results also suggest the potential benefits of the temporal structure among different domains in adapting to the target one. Empirically, our theoretical results imply that learning proper representations across the domains will effectively mitigate the learning errors. Motivated by these theoretical insights, we propose a min-max learning objective to learn the representation and classifier simultaneously. Experimental results on both semi-synthetic and large-scale real datasets corroborate our findings and demonstrate the effectiveness of our objectives.