SBIR PHASE I: MultiFACET - An Approach for Solving Large Scale Planning and Scheduling Problems

米勒 这个小企业创新研究（SBIR）第一阶段开发了新的软件，用于解决灵活的制造设施中的计划和调度问题。 软件设计是针对大规模工业规划和调度的解决方案。 预计将提高生产能力、减少浪费、降低运营成本、减少库存和缩短对客户订单的响应时间。 该方法是一个严格的数学公式的规划和调度问题，并依赖于一个新的问题分解方法。 分解方案使用包含详细模型的滑动时间窗口，再加上剩余时间范围的粗粒度规划模型。 该方法的核心是一种自动机制，用于以迭代方式沿时间轴沿着传播详细的窗口，很像数值积分。 时间轴被分解为细粒度和粗粒度部分。 在细粒度部分中，调度问题将被建模和求解为混合整数线性规划。 选择细粒度时间窗口的长度，使得调度子问题可以使用定制的算法来解决。 剩余的时间范围（细粒度窗口之外）将使用聚合约束进行建模，并通过放松完整性条件来解决。 这使得这样的宏观功能，工厂的能力和预期的产品需求，传播到每个详细的调度子问题。 如果成功的话，这项研究将提供一个新的引擎调度工具有用的过程工业。 现有的软件安装和维护困难且昂贵，并且不提供高质量的解决方案。 新软件有望减少定制和安装时间，同时提供高质量的结果。

DMI-9561853 Miller This Small Business Innovation Research (SBIR) Phase I develops new software for solving planning, and scheduling problems in flexible manufacturing facilities. The software design is directed to solutions for large-scale industrial planning and scheduling. @enefits are expected in increased production capacity, reduced waste, lower operating costs, reduced inventories, and shorter response time to customer orders. The approach is a rigorous mathematical formulation of planning and scheduling problems and relies on a novel problem decomposition methodology. The decomposition scheme uses a sliding time window containing a detailed model, coupled to a coarse-grained planning model for the remainder of the time horizon. The heart of the method is an automated mechanism for propagating a detailed window along the time axis in an iterative fashion, much like numerical integration. The time axis is decomposed into fine and coarse grained sections. In the fine grained section, the scheduling problem will be modeled and solved as a mixed-integer linear program. The length of the fine-grained time window is chosen so that the scheduling subproblem may be solved using a customized algorythm. The remainder of the time horizon (outside the fine grained window) will be modeled using aggregate constraints and solved by relaxing integrality conditions. This allows such macroscopic features as plant capacity and anticipated product demand to be propagated into each detailed scheduling subproblem. If successful, this research will provide a new engine for scheduling tools useful in process industries. Available software is difficult and expensive to install and maintain and does not provide high quality solutions. The new software is expected to reduce customization and installation time and yet provide high quality results.