Fatigue reliability estimation framework for blade-disk based on improved constrained boundary sampling and dual-point enrichment active learning strategy

The fatigue reliability of blade-disk significantly impacts the performance and safety of heavy-duty gas turbines. To achieve higher computing precision and efficiency for blade-disk reliability estimation, an improved constrained boundary sampling and dual-point enrichment active learning strategy (DP-ACBS) is proposed. Two numerical examples are employed to verify the feasibility of the proposed strategy. Considering multi-uncertainty comprehensively, a framework of blade-disk fatigue reliability assessment is developed based on the DP-ACBS method. Low-cycle fatigue (LCF) reliability and sensitivity analyses on a typical compressor blade-disk are conducted. Variations in both environmental and design factors are considered. The results indicate that the proposed DP-ACBS algorithm achieves a superior balance between computational accuracy and efficiency on numerical examples. The dovetail is the crucial failure zone of the compressor blade-disk. Mortise and tenon fillet radius are the primary factors affecting LCF lifespan with an overall sensitivity index of 0.731. The failure probability of the compressor blade-disk is 2.11% at a safety lifetime of 15,500 cycles.