Machine Learning-based Classification Model for Noise Assessment of Ammunition Types in Combined Military Firing Ranges

Current noise-evaluation practices for military firing ranges are based on weapon caliber: small firearms (less than 20 mm) apply A-weighting with a +12 dB impulse correction, whereas large firearms (20 mm or more) apply C-weighting with a +18 dB impulse correction. However, in combined firing ranges where both small and large firearms are used, this caliber-based approach does not adequately reflect the distinct acoustic characteristics of each weapon type. In particular, practice rounds from large firearms typically exhibit acoustic patterns similar to those of small firearms, thus rendering accurate classification challenging under the existing system. Hence, this study developed a machine learning-based classification model and verified its ability to classify noise more accurately by ammunition type. The Random Forest model achieved a high accuracy of 98.7 %, and when practice-round data were reclassified, 90.9 % were categorized as having small-firearm acoustic characteristics, thus confirming that large-firearm practice rounds resemble small firearms in terms of sound. These findings enhance the objectivity of noise analysis and noise-impact assessment for military firing ranges and are beneficial for improving noise-evaluation criteria in combined firing ranges in the future.