Development of a design model for enhancing data quality in production planning and control through the application of data mining methods

制造企业在动荡和竞争的环境中竞争。在这种情况下，实现高度遵守承诺的交付日期可以被视为一个显着的特点。生产计划与控制（PPC）对物流目标的实现有着重要的影响。尽管作出了巨大的规划努力，但后勤目标的实现往往不能令人满意。PPC的性能受到数据质量不足的影响。现有的提高数据质量的方法不能确保足够的数据质量，因为它们要么是为了防止错误的数据定义的输入规则，要么是恳求数据收集的工作人员的纪律。因此，拟议的研究项目的假设是，它是不可能防止出现不充分的数据在PPC过程中，而缺失或不一致的值也可以随后确定。拟议的研究项目的目的是开发一个设计模型，解决生产中最关键的错误和不一致性-作为第一步，系统地识别与生产相关的主数据和交易数据中的典型错误和不一致，并通过确定其因果关系对其进行分类。之后，可以为所有优先级的错误和不一致性开发适当的算法，以估计缺失值。从其他应用领域（如保险或金融行业）转移现有算法是受欢迎的。这些算法在实验生产环境中的软件工具的形式进行了验证。

Manufacturing companies are competing in a turbulent and competitive environment. Under these circumstances, achieving a high adherence to promised delivery dates can be seen as a distinguishing feature. Production planning and control (PPC) has a significant impact on the achievement of logistical targets. Despite a tremendous planning effort, the achievement of logistical targets is most often not satisfactory. Performance of PPC is affected by an inadequate data quality. Existing approaches for increasing data quality are not capable of ensuring adequate data quality since they are either designed to prevent faulty data defined input rules or plead to the discipline of staff in data collection.Therefore, the hypothesis of the proposed research project is that it is impossible to prevent the emergence of inadequate data in PPC processes, but rather the missing or inonsistent values can also be subsequently determined.The aim of the proposed research project is the development of a design model that addresses the most critical errors and inconsistencies in production-relevant master and transaction data and improves data quality by using data mining methods.As a first step, typical mistakes and inconsistencies in production-relevant master and transaction data are systematically identified and classified by determining their causal relationships. Afterwards, adequate algorithms can be developed for all prioritized errors and inconsistencies that allow an estimation of the missing values. A transfer of existing algorithms from other fields of application, such as the insurance or financial industry, is sought after. These algorithms are validated in the form of a software tool in an experimental production environment.