All Issue

2026 Vol.3, Issue 1 Preview Page

Research Article

Development of Machine Learning-based Depth of Penetration Prediction Model for Unmanned Combat Vehicle Protection against Small Caliber Threats

무인전투체계 방호를 위한 소화기 위협 대비 관입깊이 예측 기계학습 모델 개발

Jong-Hwan Kim1
,
Hyeonjun Kim2
,
Jang-Woon Baek3*
김 종환1
,
김 현준2
,
백 장운3*
1Professor, Department of Civil Engineering and Environmental Sciences, Korea Military Academy, Seoul, 01805, Rep. of Korea
2Lecturer, Department of Civil Engineering and Environmental Sciences, Korea Military Academy, Seoul, 01805, Rep. of Korea
3Associate Professor, Department of Architectural Engineering, Kyung Hee University, Yongin, 17104, Rep. of Korea
1육군사관학교 기계·시스템 공학과 교수
2육군사관학교 기계·시스템 공학과 강사
3경희대학교 건축공학과 교수
*Corresponding Author
31 January 2026. pp. 15-24
PDF XML Citation
Abstract

With the advancement of cutting-edge weapon systems based on the 4th Industrial Revolution, the importance of unmanned combat vehicles is increasing. Unmanned combat systems such as unmanned reconnaissance vehicles are exposed to various threats including small arms, explosives, and electronic warfare, and securing protection performance against enemy small arms threats is critical. Previous studies have mainly focused on determining penetration through ballistic limit velocity (V50), but sensors and critical equipment in unmanned combat systems can be damaged by the depth of penetration (DOP) of projectiles even before complete penetration. This study aims to develop machine learning models that can accurately predict the depth of penetration of armor plates against enemy small arms threats using data collected through ballistic experiments for optimal protection design of weapon systems. For this purpose, live-fire ballistic experiments were conducted on two types of armor materials, steel alloy (Type A) and aluminum alloy (Type B), against the threat of 5.45 mm 7N10 ordinary bullets. A total of 44 experiments were conducted to collect 32 valid data points, and multivariate regression analysis and five AI models (linear regression, Bayesian regression, SVM, XGBoost, deep neural network) were trained using strike velocity, bullet mass, and material properties as input variables. The experimental results showed that steel alloy (Type A) exhibited an average DOP of 14.6 mm (range 10.1~21.6 mm) and aluminum alloy (Type B) exhibited an average DOP of 30.3 mm (range 8.2~42.1 mm). The XGBoost model achieved the highest performance with RMSE of 1.7 (Type A) and 0.6 (Type B) mm and R2 of 0.6 (Type A) and 0.9 (Type B), confirming that machine learning models can accurately predict DOP. The results of this study provide a practical tool for optimal protection design of unmanned combat systems and establish a foundation for development into a versatile protection performance prediction system by expanding to various bullet types and materials in the future.

Keywords
unmanned combat vehicle
depth of penetration
protection performance
artificial intelligence
machine learning

4차 산업혁명 기반의 첨단무기체계 발전과 함께 무인전투체계의 중요성이 증대되고 있다. 무인수색차량 등 무인전투체계는 소화기, 폭발물, 전자전 등 다양한 위협에 노출되어 있으며, 특히 적성 소화기 위협으로부터의 방호성능 확보가 핵심적이다. 기존 연구는 주로 방호한계속도(V50)를 통해 관통 여부를 판정하는데 집중하였으나, 무인전투체계의 센서 및 핵심 장비는 완전 관통 이전에도 탄자의 관입깊이(Depth of Penetration, DOP)에 의해 손상될 수 있다. 본 연구는 탄도 실험을 통해 수집된 데이터를 활용하여, 무기체계의 최적 방호 구축을 목표로 적성 소화기 위협에 대응 가능한 장갑의 관입 깊이를 정밀하게 예측할 수 있는 머신러닝 모델을 개발하는 것을 목적으로 한다. 이를 위해, 5.45 mm 7N10 보통탄을 위협으로 설정하고, 강철 합금(Type A)과 알루미늄 합금(Type B) 2종의 장갑 재질에 대한 실사격 탄도실험을 수행하였다. 총 44발의 실험을 통해 32개의 유효 데이터를 수집하였으며, 충돌속도, 탄자중량, 재질 물성치 등을 입력변수로 하는 다변량 회귀분석 및 5가지 AI 모델(선형 회귀, 베이시안 회귀, SVM, XGBoost, 심층신경망)을 학습하였다. 실험 결과, 강철 합금(Type A)은 평균 14.6 mm(범위 10.1~21.6 mm), 알루미늄 합금(Type B)은 평균 30.3mm(범위 8.2~42.1 mm)의 관입깊이를 나타냈으며, XGBoost 모델이 RMSE는 1.7(Type A)와 0.6(Type B) mm, R2 는 0.6(Type A)과 0.9(Type B)로 최고 성능을 달성하여 머신러닝 모델이 관입깊이를 정밀하게 예측할 수 있음을 확인하였다. 본 연구 결과는 무인전투체계의 최적 방호 설계를 위한 실용적 도구를 제공하며, 향후 다양한 탄종과 재질로 확장하여 범용적인 방호성능 예측 시스템으로 발전시킬 수 있는 기반을 마련하였다.

키워드
무인전투체계
관입깊이
방호성능
인공지능
기계학습
References
1

Choi, H. S., Shin, H. H., Yoo, Y. H., Park, J. H., and Kim, J. B. (2016) Investigation Into Protection Performance of Projectile Using Flying Plate. Journal of the Korean Society for Precision Engineering 33(12), 1039-1045. (In Korean)

10.7736/KSPE.2016.33.12.1039
2

Cortes, C. and Vapnik, V. (1995) Support-vector Networks. Machine Learning 20, 273-297.

10.1023/A:1022627411411
3

Hwang, H. G., Kim, B. S., Kang, J. W., and Lee, J. S. (2016) A Method and Application of Vulnerability Analysis for Combat Systems Considering Threats and Defense Ability: Focused on PKM Model. Journal of the Korea Institute of Information and Communication Engineering 20(8), 1623-1631. (In Korean)

10.6109/jkiice.2016.20.8.1623
4

Kim, J. H., Baek, S. W., Yun, B. J., and Jo, S. S. (2016) Support Vector Machine based Ballistic Limit Velocity Measurement for Small Caliber Projectile. Journal of the Korea Institute of Military Science and Technology 19(5), 629-637. (In Korean)

10.9766/KIMST.2016.19.5.629
5

Kim, J. H., Jeong, J. H., and Jung, Y. Y. (2022) Machine Learning Model Analysis for Armor Penetration Depth Interpretation Against Small Calibers. Journal of the Military Operations Research Society of Korea 48(2), 14-24. (In Korean)

6

Lee, G. W., Baek, J. W., Lee, B. H., Kim, J. Y., and Kim, J. H. (2021) Ballistic Protection Effectiveness Analysis of Armor Plates with Various Incident Angles using Small Caliber Live Fire Test. Journal of the Korea Society for Simulation 30(1), 55-63. (In Korean)

7

Ministry of National Defense (2018) Defense Reform 2.0 Basic Plan. Seoul: Ministry of National Defense, Republic of Korea. (In Korean)

8

Park, Y. J., Park, S. J., Yu, Y. J., Kim, T. H., and Son, K. Y. (2016) Reinforcing Method for the Protective Capacities of Dispersal and Combat Facilities using Logistic Regression. Journal of the Korea Institute of Building Construction 16(1), 77-85. (In Korean)

10.5345/JKIBC.2016.16.1.077
9

Republic of Korea Army Headquarters (2020) Army Vision 2050: The Super-Class Army Leading Time and Space. Gyeryong: Republic of Korea Army.

10

U.S. Department of Defense (USA DoD) (1997) MIL-STD-662: Department of Defense Test Method Standard: V50 Ballistic Test for Armor. Washington, D.C.: U.S. Department of Defense.

11

U.S. National Institute of Justice (USA NIJ) (2008) NIJ Standard-0101.06: Ballistic Resistance of Body Armor. Washington, D.C.: U.S. Department of Justice, Office of Justice Programs.

Information
  • Publisher :Korea Protective Facility Institute
  • Publisher(Ko) :한국방호시설학회
  • Journal Title :Protective Facility
  • Journal Title(Ko) :방호시설
  • Volume : 3
  • No :1
  • Pages :15-24
  • Received Date : 2026-01-11
  • Revised Date : 2026-01-20
  • Accepted Date : 2026-01-20
  • DOI :https://doi.org/10.23310/PF.2026.3.1.015
Journal Informaiton Protective Facility Protective Facility
Online Submission & Review System
  • KISTI Current Status
  • KISTI Cited-by
  • KCI 문헌 유사도 검사
  • orcid
  • open access
  • ccl
Journal Informaiton Journal Informaiton - close