Under the high-load design requirements of modern aviation compressors

the internal flow field faces severe issues of flow separation and enhanced secondary flow

which can easily trigger compressor instability. To broaden the stable operating range of centrifugal compressors

this study establishes an optimization design platform for non-axisymmetric endwall contouring. A surrogate model is constructed using a back propagation (BP) neural network

and multi-objective optimization is carried out with the non-dominated sorting genetic algorithm. Targeting improvements in stability margin and efficiency maximization

an optimization design study is conducted on the disk-side endwall contouring of the diffuser. The results show that the optimal endwall contouring improves the stability margin of the centrifugal compressor by 6.89% and increases the isentropic efficiency under design conditions by 0.63%. Further flow field analysis reveals that the stability enhancement mechanism is mainly reflected in two aspects. First

the local convexity of the optimal endwall contouring increases the local flow velocity

thereby enhancing the resistance to adverse pressure gradients and weakening flow reversal. Second

the local convexity suppresses circumferential flow

increases the flow angle

and improves throughflow capacity

thus achieving stability enhancement. The conclusions of this study can provide theoretical support for the design of centrifugal compressors with wide stability margins.