ESS: Scheduling, Inventory Optimization, and Coordination of Maintenance Networks

许多行业依赖于其关键资产的正常运作，例如航空公司使用喷气发动机，电力公用事业公司使用发电机，军队使用直升机。运营效率的一个主要障碍是由于强制停机和定期维护操作而导致的资产不可用。后者虽然有计划，但通常具有持续时间长和不确定的特点。现在的趋势是监控资产状况，并根据状况执行维护操作，以提高资产可用性和可靠性，而不是开始实施更密集的预防性维护计划。这给包括大修车间、维修车间、零部件分销商和备件制造商在内的维修服务提供商网络带来了重大挑战，即在传统上具有巨大不确定性的系统中，如何在减少库存的同时获得短且可预测的周转时间。研究动态和不确定环境下维修服务网络的调度、库存优化和协调，以应对状态维修的挑战。通过与我们的行业合作伙伴联合技术研究中心和i2技术公司的密切合作，将执行三项任务。首先，基于我们的工业伙伴提供的现实离散事件仿真模型，建立了大规模维修服务网络的随机调度模型。将开发一种协同结合拉格朗日松弛、随机优化和模拟的求解方法，以提供具有可量化质量和可信度的接近最优的解。第二项任务是针对任务1中开发的调度方法优化部件库存，平衡库存成本和部件可用性。将检查持续和定期审查策略，并将使用顺序优化来优化策略参数。在第三项任务中，将调查维修服务网络中各种组织的自治性质。任务1和任务2的模型将是分散的。将根据与任务1中制定的方法一致的定价概念，建立一个分布式和异步协调机制。然后，将设计和实施一个移动多代理系统，以便在因特网上部署这些方法，为下一代电子维修服务网络铺平道路。我们的目标是通过缩短和可预测的周转时间，同时降低库存水平，最大限度地利用资产状况信息和信息技术基础设施，充分利用新的维护服务模式。本研究还将对调度、库存管理和供应链管理的基础理论和实践做出重要贡献，提高服务企业的竞争力和可靠性。

Many industries rely on the proper functioning of their key assets, such as airlines on jet engines, electric utilities on generators, and army on helicopters. A major obstacle on operation efficiency is asset unavailability due to forced outages and regular maintenance operations. The latter, although planned, are generally characterized by long and uncertain durations. Instead of embarking on more intensive preventative maintenance programs, the trend is to monitor asset conditions, and perform maintenance operations based on conditions to increase asset availability and reliability. This imposes major challenges on the networks of maintenance service providers, including overhaul shops, repair shops, part distributors, and spare part manufacturers, to have short and predictable turn-around-times while reducing inventory in systems traditionally characterized by massive uncertainties.The research is on scheduling, inventory optimization, and coordination of maintenance service networks under dynamic and uncertain environments to meet the challenges of condition-based maintenance. By closely collaborating with our industrial partners United Technologies Research Center and i2 Technologies, Inc., three tasks will be performed. The first is to develop stochastic scheduling models for large-scale maintenance service networks based on a realistic discrete event simulation model provided by our industrial partners. A solution methodology that synergistically combines Lagrangian relaxation, stochastic optimization, and simulation will be developed to provide near-optimal solutions with quantifiable quality and confidence. The second task is to optimize part inventory, balancing inventory costs and part availability for the scheduling method developed in Task 1. Continuous and periodic review policies will be examined, and ordinal optimization will be used to optimize policy parameters. In the third task, the autonomous nature of various organizations within a maintenance service network will be investigated. The models of Tasks 1 and 2 will be decentralized. A distributed and asynchronous coordination mechanism will be established, building on the pricing concept that is consistent with the methods developed in Task 1. A mobile multi-agent system will then be designed and implemented for Internet-based deployment of the methods, paving the way for the next generation e-maintenance service networks. Our goal is to reap the full benefit of the new maintenance service paradigm by having short and predictable turn-around-times while reducing inventory levels, making the best use of asset condition information and the information technology infrastructure. The research shall also significantly contribute to the fundamental theory and practice of scheduling, inventory management, and supply chain management, improving the competitiveness and reliability of service enterprises.