Abstract: To address the disconnect between engineering mechanics teaching and engineering practice, a three-stage problem-chain teaching mode of “loading scenario → mechanical model → calculation verification” covering the core knowledge points of engineering mechanics was constructed based on a real deep foundation pit support project, and its effectiveness was verified through a quasi-experimental design with parallel classes. Results indicated that the experimental class scored significantly higher on the mechanics modeling test (78.56 ± 8.23) than the control class (64.31 ± 12.50), with a 22.0% improvement in engineering problem identification scores and the accuracy in drawing internal force diagrams increased by 17.8 percentage points, corresponding to a relative increase of 30.4%. The main effect of the teaching method remained significant after controlling for entrance scores via ANCOVA. Learning engagement in the experimental class was significantly higher than that in the control class, and the mediating effect of the teacher’s “practicing engineer identity” accounted for 38% of the total effect, serving as one of the important transmission mechanisms. The design logic of this mode can be transferred to bridge, tunnel and other engineering scenarios, providing a transferable implementation path for contextualized teaching reform in basic engineering courses.