Fixed parameters in metaheuristics like Differential evolution (DE) limit truss optimization efficiency. This study proposes a PS-QDE(Phased strategy Q-Learning DE) algorithm that uses Q-learning to dynamically adapt parameters. A novel "Strategy switching factor" is also introduced to adjust the exploration-exploitation balance based on convergence. The PS-QDE algorithm was validated on four truss optimization problems (10-bar to 200-bar) with frequency constraints. Results show PS-QDE provides more stable convergence and superior or competitive optimal solutions compared to standard DE.

Abstract
1. 서론
2. 선행연구
    2.1 DE
    2.2 Q-러닝
3. PS-QDE
    3.1 전략 전환 계수
    3.2 Q-러닝 에이전트 설계 및 하이퍼파라미터
    3.3 민감도 분석
4. 수행 예제
5. 수행 결과
    5.1 10바 트러스
    5.2 72바 트러스
    5.3 120바 트러스
    5.4 200바 트러스
6. 결론
감사의 글
References

• 노승현(한국기술교육대학교 건축공학과, 석사과정) | Noh Seung-Hyeon (Dept. of Architectural Engineering, Korea University of Technology and Education)
• 이돈우(한국기술교육대학교 디자인건축공학 부 박사후 연구원, 공학박사) | Lee Don-Woo (School of Design and Architectural Engineering, Korea University of Technology and Education) Corresponding author
• 이승재(한국기술교육대학교 디자인건축공학부 교수, 공학박사) | Lee Seung-Jae (School of Design and Architectural Engineering, Korea University of Technology and Education)