رویکرد جامع برای کنترل شناختی از سیستم های تولید

Increasing dynamics and a turbulent environment force industrial enterprises to ensure a highly efficient production. The field of production planning and control (PPC) and the sustainable optimization of its methods are hereby of utmost importance. This paper introduces a concept for a cognitive production planning and control system, in which so-called smart products store knowledge about the production process and its current state. The RFID (radio frequency identification) technology presents a promising approach to realize those smart products, to enhance the information management on the shop floor and to offer a precise image of individual product states in the production process. The knowledge on production sequences is represented in a graph-based model. The developed concept represents the executable production of every single resource in capability profiles that are used for the allocation of production steps to resources. Material transports are realized by an anticipatory transport control, which updates its model parameters autonomously. During runtime, the product-specific operation times are measured and stored on the smart product, which is subsequently used to update the overall planning data. Thus, the introduced production planning and control system is able to react to unforeseen events (e.g. missing material, insufficient product quality) and autonomously adapts the planning data to the actual elapsed values of the real production. First experiments showed promising results for the approach to provide and process information directly on the shop floor: the idleness of resources due to errors was reduced by 41% from 19.4% to 8.0% during a 3 h test run. The waiting time of resources caused by missing material can be reduced in specific cases by 17.7%.

In recent years, global economic competition and a shift from seller markets to buyer markets have induced increasing dynamics and a turbulent environment for industrial enterprises [25]. Propagated concepts such as mass customization and individualization promised the creation of unique products that satisfy the needs of nearly every customer [29]. This trend has been accompanied by uncertainties for production enterprises in terms of an increasing number of products, product variants with specific configurations, large-scale fluctuations in demand and random dispatching of orders [18] and [28]. Therefore, companies can only compete successfully if they offer products and services that meet the customer’s individual requirements without sacrificing cost effectiveness, product quality and on-time delivery [7], [14] and [27].

Increasing requirements with regard to reactivity, adaptability and traceability in production and by extension in the supply chain can be observed among products, processes and clients all over the product lifecycle. In order to cope with these challenges, new methods for an accurate, reliable and adaptive production planning and control are necessary. The integration of modern sensor technologies presents a promising approach to enhance information management on the shop floor. Hence, current states of both resources and individual products can be captured. The paper presents the framework of a cognitive production system, which shall foresee the consequences of its action and thereby be able to influence the environment in a way that optimizes the expected performance in a robust and flexible manner. To reach this goal, perception, knowledge and planning capabilities of a production system are required on different levels. The approach is detailed by taking the example of the cognitive production planning and control, the anticipatory transport control and the model update and optimization. Although, the concept developed shows significant results, it also contains several risks and limitations: the RFID-technology is not suitable for every production, as it is susceptible to external influences. In addition, the methods used to solve the single decision-making steps during the production may still be improved or even replaced by more general algorithms. Nevertheless, the overall concept seems to be a contribution to improve the production planning and control in high-variant production scenarios.