Simulation-based dynamic heuristic for the distributed optimisation of complex multi-objective multi-project multi-resource production processes

调度和运力规划中的决策问题广泛存在于经济和社会的各个领域。它们尤其可以在工业生产中找到，也可以在建筑和医疗保健中找到。这些问题的典型特征是其复杂性和多样性。高质量机械、设备和车辆的开发和生产具有客户专门化和产品多样化的趋势，同时产品生命周期也变得越来越短。这导致了极其复杂的、不确定的和动态的多资源生产过程，就高效生产而言，很难对其进行最优计划和控制。这个问题在客户专用重型机械的装配过程中变得更加尖锐，因为由于几个因素(例如，优先关系、工艺选择、资源之间的权衡等)的重叠，可能的序列变体的数量呈指数级增加。此外，目标函数(例如，成本、时间、资源利用率等)在这样的多目标问题中，工作具体不同，甚至在一个订单内也可能随着时间的推移而变化。研究和实践经验表明，有效解决这些问题具有巨大的潜力。到目前为止，还没有足够强大和应用灵活的模型、程序和解决方法。目前，无论是精确解方法、启发式方法还是元启发式方法都不能很好地解决这些问题，通过这一基础研究项目，将探索合适的参考过程模型和求解方法，并可推广到其他具有类似问题特征的应用领域。核心目标是开发概念、体系结构和整合方法，这些概念、体系结构和整合方法能够高效和有效地解决该结构内所描述的复杂问题。概念上的基本组成部分应是：-由生成器、优化器、模拟器和分析器组成的模块化优化框架--使用类似代理和元启发式函数的迭代、启发式初始化、优化和过程概念。主要方法目标是：-开发有效处理大搜索空间的方法-探索价值系统，作为评估和比较在考虑不同目标和约束下产生的解决方案的方法。

Decision problems in scheduling and capacity planning exist in many sectors of economy and society. They can be found especially in industrial production but also in construction and healthcare. Typical properties of these problems are their complexity and diversity. The development and production of high-quality machinery, equipment and vehicles has the tendency towards a permanent increase of customer specificity and product diversification whilst product life cycles become shorter at the same time. This leads to extremely complex, non-deterministic and dynamic multi-resource production processes which can hardly be planned and controlled optimally in terms of an efficient production. The issue grows more acute particularly at assembly processes for customer-specific heavy machinery since the number of possible sequence variants increases exponentially due to the overlapping of several factors (e.g., precedence relations, process alternatives, trade-offs between resources etc.). In addition, the objective functions (e.g., cost, time, resource utilisation, etc.) vary job-specifically and may change over time even within an order in such multi-objective problems. Experiences from research and practice show that the effective solution of such problems opens an enormous potential. To date, there are no sufficiently powerful and application flexible models, procedures, and solution methods. Neither exact solution methods, heuristics nor meta-heuristics are currently able to solve these problems satisfactorily.With this basic research project, suitable reference process models and solution methods will be explored which are to be transferable to other application areas with similar problem characteristics. Core objective is the development of concepts respectively architectures and the incorporating methods which are able to solve the described complex problems efficiently and effectively within this structure. The conceptual basic components shall be:- A modular optimisation framework consisting of a generator, optimiser, simulator and analyser- An iterative, heuristic initialisation, optimisation and process concept using agent-similar and meta-heuristic functionalitiesThe key methodological objectives are:- The development of methods for the efficient handling of large search spaces- The exploration of a value system as a method for evaluating and comparing solutions produced under consideration of differentiated objectives and constraints.