Operational and online planning of maintenance and production

生产和维护的核心冲突是维护措施必须确保机器的高可用性。然而，为了实现它们，通常需要在短时间内停机。特别是在复杂、多阶段和高利用率的生产系统中，生产和维护计划的冲突行为尤为明显，因为机器停机会迅速导致整个生产系统的生产损失和故障后果。现有的科学方法不能解决这些问题，因为实际生产系统的复杂性是不充分的，并且是在非常限制性的假设下绘制的（例如单机方法，确定性和静态决策环境）。在dfg资助的科学项目mk - proist的框架下，通过生产工程和机床研究所（IFW）的离散事件模拟（DES），开发并实施了一种协调生产和维修计划的动态规划方法。这种方法可以对生产过程中实施维护措施时产生的故障后果进行可靠的预测和量化。此外，规划备选方案可以推导和评估（例如，备选维修开始时间）。对该方法的验证表明，生产和维护成本可降低9%。此外，由于仿真技术在决策中考虑了实际生产系统的复杂性、动态性和随机性，大大提高了规划质量。该方法的工业利用的基础是减少模型创建和适应所需的时间，并在早期阶段显示低模型有效性。知识转移项目的目标是利用动态规划方法为各种公司的高效在线应用奠定必要的基础。为此，现有的规划方法将扩展到基于BDE / mde系统（例如随机数据）的数据的面向过程的仿真模型的学习和自参数化方法。此外，将建立适当的反馈回路，以及时识别低效度的模型。这些方面是在实践中高效地创建和适应模型以及可操作（在线）应用程序的基础。

The central conflict of production and maintenance is the fact that maintenance measures have to ensure a high availability of machines. However, for their implementation a machine downtime is often required at short notice. Especially in complex, multi-stage and high utilized production systems, the conflicting behavior of production and maintenance planning is particularly pronounced, since the machine downtime can quickly cause production losses and failure consequences for the entire production system. Existing scientific approaches do not solve these issues, since the complexity of real production systems is mapped inadequate and under very restrictive assumptions (e. g. single machine approaches, deterministic and static decision environment).In the framework of the DFG-funded scientific project MK-ProInst, a dynamic planning approach to coordinate the production and maintenance planning was developed and implemented by means of discrete event simulation (DES) at the Institute of Production Engineering and Machine Tools (IFW). This approach allows a robust prediction and quantification of the resulting failure consequences that arise when implementing a maintenance measure during production. Moreover, planning alternatives can be derived and evaluated (e. g. alternative maintenance starting time). A validation of the approach showed that the production and maintenance costs can be reduced by up to 9 %. Furthermore, the planning quality is significantly increased since the complexity, dynamics and stochastics of real production systems are considered using the simulation technology in decision making. The basis for industrial exploitation of the methodology is to reduce the time needed for model creation and adaption and to show a low model validity at an early stage.The goal of the knowledge transfer project is the exploitation of the dynamic planning approach to achieve the essential foundation for an efficient and online application for a variety of companies. For this, the existing planning approach will be expanded with regard to methods for learning and self-parameterization of process-oriented simulation models based on data from BDE-/MDE-Systems (e. g. for stochastic data). In addition, appropriate feedback loops will be developed to the timely recognition of a low model validity. These aspects are the basis for time-efficient model creation and adaption and the operational (online) application in practice.