Quality control in automated manufacturing increasingly relies on predictive modelling to ensure product reliability. This study introduces a Hybrid Stacked Ensemble Model (HSEM) combining five base learners – GLM, GBM, RF, XGBoost, and Deep Learning – with GBM as the meta-learner. Using the Yeh dataset for concrete compressive strength (CCS), a standard benchmark in construction quality assessment, the HSEM achieved outstanding performance with R2 = 0.99 and RMSE = 0.0139 on the test set, surpassing both individual and state-of-the-art models. Robustness analyses through cross-validation, bootstrap sampling, and noise perturbation confirmed its strong generalization ability. SHAP interpretation identified total material content, specimen age, cement, and water-cement ratio as dominant factors influencing CCS. The results demonstrate HSEM’s potential as a reliable and interpretable approach for predictive quality control in manufacturing.