Integration of acoustic resonance testing for accelerated and cost-effective quality control of pins in screw coupling manufacturing

Modern railway manufacturing is under increasing pressure to reduce quality-control (QC) costs without compromising safety. This study investigates Acoustic Resonance Testing (ART) as a fast, non-destructive pre-screening method for detecting structural and material non-conformities in screw coupling pins, which are safety–critical components of 850 kN railway couplers. Eighteen heat-treated steel pins were evaluated using standard QC results and frequency response functions (FRFs), with rule-based and Random Forest classification. The rule-based classifier achieved 88.9% accuracy and detected six of eight NOK pins (NOK recall = 0.75). In a single six-pin test split, the Random Forest correctly classified five pins (83.3% accuracy), detecting both NOK pins and generating one false-positive NOK decision. These results provide preliminary pilot-scale evidence; 90 features and only 12 training pins pose a substantial overfitting risk and preclude reliable assessment of model stability. Economic modelling of Standard, ART, Hybrid and Full Skip strategies showed operational cost reductions of approximately 96.7% and processing-time reductions of 49.5% for Full Skip. Including initial investment, expenditure fell by 81.5% and 87.1% over 3 and 5 years. The scenarios assume that validated ART screening permits omission of routine hardness and laboratory testing; batch-level sensitivity analysis quantifies the effect of classification-error consequences. ART shows promise for pre-screening when combined with controlled excitation, independent validation, and risk-informed quality management.