基于多代理强化学习的结构拆解资源化评估及决策优化

Evaluation and Decision Optimization for Structural Deconstruction Using Multi-agent Reinforcement Learning

摘要:

针对老旧结构构件拆解施工过程，通过资源流动约束、柔性工作分解和选择性拆解策略建立了拆解工序计划数学模型，并提出了一个以回收效率为目标的多代理强化学习拆解方案优化方法。案例应用结果表明，与遗传算法相比，该方法具有更高的碳排放回报/消耗比，验证了基于智能算法优化的拆解方案对节能减碳的有效性。

Abstract:

For the deconstruction of components in aging structures， a mathematical model for process planning is developed based on resource flow constraints， flexible work breakdown， and selective dismantling strategies. A multi-agent reinforcement learning method is then proposed to optimize the deconstruction planning with recovery efficiency as the core objective. Case study results indicate that， compared to the genetic algorithm， the method achieves a higher carbon emission return-to-consumption ratio， demonstrating the effectiveness of intelligent algorithm-based optimization in promoting the energy efficiency and carbon reduction.