Production planning and control (PPC) is facing more and more requirements due to the increasing dynamics and complexity in highly saturated markets. For example, a purely economic evaluation of the production process is no longer sufficient – in addition, ecological and logistical requirements must also be met. In this area of conflict, the decision that results in the greatest possible economic success for the company per unit of resources used must inevitably be prioritized.
In the future, intelligent agents for production control will be used to support this decision and the best possible positioning between the three conflicting requirements.
Production control and its challenges
Production control has the task of dispatching production orders and controlling them during the production process. The primary goal is to process the production orders in accordance with the existing production plan and to take into account any changes that may be necessary. In order to achieve this goal, the orders must be generated, released and put into a sequence according to a uniform logic.
The SFB 1153 is developing innovative process chains for the production of hybrid high-performance components using tailored forming technology. These are evaluated economically, ecologically and logistically in the area of tension shown in order to quantify the advantageousness of the technology. To achieve the overall goal of producing load-adapted hybrid solid components using joined semi-finished products, the IFA is developing an integrated PPC process.
Advantages of AI over established approaches
Established approaches to production control primarily focus on economic analyses – in addition to a logistical evaluation of decisions. These established methods are usually based on expert knowledge and heuristics. The key advantage of novel AI methods used in this research approach is the simultaneous consideration of a large number of relevant characteristics (features) while incorporating extensive data sets.
The integration of ecological factors represents a major challenge in the development of the intelligent approach. Especially against the background that a later application in industry requires a simultaneous consideration of all restrictions, sustainability aspects have to be integrated in addition to the traditional economic and production logistic factors. This results in a gap between increasing dynamics and complexity on the one hand and growing requirements on the other. By using modern AI approaches, the IFA aims to bridge this gap.
AI in production control
AI provides a variety of methods that have the potential to counteract the complexity described above. Existing master data and transaction data in companies, generated, for example through the use of modern ERP and MES systems, can be efficiently used and seamlessly integrated into production control through the use of such methods and approaches.
AI enables the rapid processing of large volumes of data and automated decision-making. As part of Industry 4.0 developments and related technological advances such as Radio-Frequency Identification (RFID), an essential foundation has been laid for cost-effective data collection, which in turn forms the basis for the use of AI applications.
Intelligent agent learns to make decisions
The use of reinforcement learning (RL) is one such innovative approach to dealing with the complexity of production control and is being advanced in SFB 1153. This is a method from the field of machine learning in which an intelligent agent learns to make sequential decisions through interactions. Within the framework of SFB 1153, the IFA is currently developing such an agent. This agent is trained with the help of Q-learning or policy gradients for a holistic production control, which takes into account economic and logistical as well as ecological restrictions.
First analyses have shown that the underlying approach is able to fulfil requirements better than established industry standards and thus has the potential to improve the logistic performance of production control and to realize a better positioning in the field of conflicting target variables.
Currently, the identification and exploration of additional potential measures and influencing factors is underway to further develop and optimize the intelligent agent in a step-by-step process. These explorations aim to increase the effectiveness and performance of the intelligent agent. The incremental improvement emphasizes the continuous progress in AI research and development as well as the increasing knowledge gained in the context of SFB 1153. The optimization process continuously extends the adaptation possibilities of the agent to increase its efficiency and applicability. The aim is to be able to manufacture ever more efficient, lighter and smaller components in production, taking into account economic, ecological and logistical goals.