Enhanced models for large scale supply chain tactical planning problems

供应链就是为了顺畅地管理产品需求而进行的协作。随着产品(以及合作伙伴)数量的增加，供应链规划变得复杂起来。规划的困难大多源于需求、供应、时间等方面的不确定性，不确定参数越多，规划的难度就越大。森林/农产品供应链等以自然资源为基础的供应链是这类供应链的一种，各级的不确定性都很高，产品数量巨大。为了管理这些不确定性，管理者要么选择高水平的冗余，也就是说，要么在整个供应链的各个层面缓冲库存，要么进行多重采购，等等。所有的选择都有优缺点。多种策略的可获得性使得决策和供应链规划变得更加复杂。为了顺利管理供应链，其中一个选择是限制产品数量。然而，企业提供了一系列产品来满足大量客户的需求。在这种情况下，为了保持供应链的精益，产品的设计方式是大多数最终产品共享许多公共组件(原材料)。换句话说，一个组件可以在许多不同的产品中使用。有两个极端：为每个产品设计特定于产品的组件；以及设计可以适合每个产品的单个通用组件。这两种极端都有优缺点，为了平衡与这两种极端相关的优缺点，最终产品被分成不同的类别，每个类别共享组件。例如，一个特定宽度的卷筒可以用来生产几种最终产品。确定类别的数量和类别中的产品分类是一个非常复杂的问题。在数学术语中，该问题称为集合划分问题，在实践中称为分类问题。在供应链中，我们在不同的层面上遇到了许多分类类型的问题。例如，为有能力限制的多个产品选择供应商，将零售商分配到仓库，为有先例限制的多个任务选择承包商，资源分配等。由于与能力、成本和随机性相关的非线性，解决这些问题并不容易。高浪费和有限的替代供应(在最后一刻)在以天然产品为基础的供应链中非常常见。合理的品种规划是保持供应链畅通和运营成本最小化的唯一途径。在这项研究中，我们的目标是对相关文献进行系统的回顾，并通过提供新的工具来解决这些复杂的问题，丰富现有的知识。我们将确定关键的战术层面规划问题，并开发基于数学建模的方法。根据问题的复杂性、问题的大小和使用情况，我们将开发和测试方法的组合。将通过出版物和研讨会传播最有效和最适当的办法。具体地说，我们的主要目标是：1)开发和测试在这种情况下更适合供应链的策略。2)发展数学规划求解方法，有效地解决这类复杂模型的求解问题。3)开发/探索作为性能基准的测试用例。这项研究的结果将是一套规划工具，可以帮助该行业改进规划，从而更具成本效益和竞争力。这项研究的其他成果包括为研究和该领域的进一步发展培训几名高素质的人员。所产生的成果将通过出版物、会议演示文稿分享。

Supply chain is all about collaboration for managing the product demand smoothly. The supply chain planning becomes complex as the number of products (and thus partners) grows. Most of the difficulty in planning arises from the uncertainties in demand, supply, time, etc. The more the number of uncertain parameters, the more difficult the planning is. Natural resources based supply chains such as forest/agro product supply chain are one of such supply chains where uncertainties at all levels are high with huge number of products. To manage these uncertainties, either managers opt for high level of redundancy, that is, either buffer stocks at all levels across the supply chain or do multiple sourcing, etc. All the options come with advantages and disadvantages. Availability of multiple strategies makes decision making and supply chain planning more complex. To manage the supply chain smoothly, one of the options is to limit the number of products. However, businesses offer a portfolio of products to cater a large pool of customers. In this situation, to keep the supply chain lean, products are designed in such a way that most of the final products share lot of common components (raw material). In other words, a component can be used in many different products. There are two extremes: Designing the product specific component for each product; and designing a single universal component that can fit in every product. Both extremes have advantages and disadvantages and to balance the advantages and disadvantages associated with the both extremes, final products are segregated into different categories and each category shares the component. For instance, a roll of a specific width can be used to produce several final products. Determination of number of category and classification of products in categories is a very complex problem. In mathematical terms, the problem is known as set partition problem and in practice called the assortment problem. In supply chains we encounter numerous assortment type problems at various levels. For instance, selecting suppliers for multiple products with capacity restrictions, assigning retailers to warehouses, selecting contractors for multiple jobs with precedent constraints, resource assignments, etc. Resolving these problems are not easy due to non-linearity associated with capacity, cost and randomness. High waste and limited alternative supply (at the last moment) are very common in natural products based supply chains. Proper assortment planning is the only way to keep the supply chain smooth and operating cost minimal. In this research, our goal is to do systematic review of the relevant literature and enrich the existing knowledge by contributing novel tools to solve such complex problems. We will identify key tactical level planning problems and develop mathematical modeling based approaches. Based on the problem complexity, problem size and usage we will develop and test combination of approaches. Most effective and suitable approaches will be disseminated through publications and seminars. In particular, our main objectives are: 1) Develop and test strategies which are more suitable for supply chains under these circumstances. 2) Develop mathematical programming solution approaches which are effective and efficient in solving such complex models. 3) To develop /explore test cases for bench-marking the performance. The outcome of this research will be a set of planning tools that can help the industry in improved planning and thus more cost effective and competitive. Other outcomes of the research include training of several highly qualified personnel for research and further advancement of the field. The generated results will be shared through publications, conference presentations.