Abstract: Aiming at the high volatility and time-varying arrival characteristic of orders in servitization of manufacturing,a queueing network model oriented to multi-type production services was proposed to address resource allocation optimization in distributed manufacturing under system performance constraints,which model ensured the rational utilization of manufacturing resources and stability of the service level.Considering the heterogeneity of multi-type production in terms of pricing,service rates,abandonment costs and abandonment rates,the population was initialized using the Tent chaotic mapping method.An adaptive adjustment mechanism based on the queueing system state was introduced for inertia weights and learning factors,while the Metropolis criterion from the simulated annealing algorithm was incorporated.A Multi-strategy Improved Particle Swarm Optimization algorithm (MIPSO) was developed to achieve rational resource allocation and maximize the distributed manufacturing platform's profit.Furthermore,the model emphasized the necessity of considering enterprise or user budget constraints during resource allocation and set appropriate resource upper limits.Numerical experiments validated the effectiveness of the proposed algorithm,providing theoretical support and management insights for resource allocation in distributed manufacturing service networks.Aiming at the high volatility and time-varying arrival characteristic of orders in servitization of manufacturing,a queueing network model oriented to multi-type production services was proposed to address resource allocation optimization in distributed manufacturing under system performance constraints,which model ensured the rational utilization of manufacturing resources and stability of the service level.Considering the heterogeneity of multi-type production in terms of pricing,service rates,abandonment costs and abandonment rates,the population was initialized using the Tent chaotic mapping method.An adaptive adjustment mechanism based on the queueing system state was introduced for inertia weights and learning factors,while the Metropolis criterion from the simulated annealing algorithm was incorporated.A Multi-strategy Improved Particle Swarm Optimization algorithm (MIPSO) was developed to achieve rational resource allocation and maximize the distributed manufacturing platform's profit.Furthermore,the model emphasized the necessity of considering enterprise or user budget constraints during resource allocation and set appropriate resource upper limits.Numerical experiments validated the effectiveness of the proposed algorithm,providing theoretical support and management insights for resource allocation in distributed manufacturing service networks.

Key words: distributed manufacturing, queueing network model, dynamic resource allocation, particle swarm optimization algorithm, simulated annealing algorithm

摘要： 在制造业服务化模式下,针对制造订单的高波动和时变特性,构建了一种面向多类型生产服务的排队网络模型,用以解决分布式制造场景下具有系统性能约束的资源配置优化问题,以保证制造资源的合理使用及制造服务水平的稳定可控。由于多类型生产的价格、服务速率、放弃成本和放弃速率具有异构性,采用Tent混沌映射初始化种群,引入基于排队系统状态自适应调整的惯性权重和学习因子,并融入模拟退火算法的Metropolis准则,提出了一种多策略改进的粒子群算法(MIPSO),以实现制造资源的合理配置并最大化制造平台利润。此外,研究发现分布式制造平台在资源配置时须考虑企业和用户的预算限制并设定合适的资源上限。最后,通过数值实验证明了所提模型与算法的有效性,为分布式制造服务网络的资源配置提供了理论支持与管理洞见。

关键词: 分布式制造, 排队网络模型, 资源动态配置, 粒子群算法, 模拟退火算法