BMBF-Projekt zur Unterstützung der Arbeitsplanung bei kleinen und mittleren Unternehmen gestartet

Support for process sequence determination in work planning through machine learning (VorPlanML)

The task of work planning is to plan the manufacturing processes for each product in order to enable its economical and production-oriented manufacture. This is done by creating a schedule and order-neutral work plan. Based on the product description, the work plan determines the raw material to be used, the sequence of operations, the means of production, the standard times and the wage groups. Work planning thus forms the transition between the development and production of a product and determines the time, costs and quality of production. Ever shorter product life cycles and the trend towards customized products are increasing the workload of work planning in the industrial environment, especially in small series. Against this backdrop, supporting work planning with software offers great potential for savings. However, this potential has not yet been fully exploited. Although the relevant software offers support during work plan creation, there is still room for improvement in the integration of implicit knowledge from previously created work plans into the software. The lack of mapping of implicit knowledge in the software means that manual intervention in work planning is often necessary. This applies in particular to determining the sequence of operations, which is often based on the experience of skilled workers and other aids such as documents, guidelines and work plans for similar products. Accordingly, the integration of implicitly available knowledge into software leads to a reduction in the frequency of manual interventions when determining the sequence of operations and thus to an acceleration of the work planning process.

The aim of the project is to predict the steps of the operation sequence by extracting implicit knowledge from existing work plans using machine learning algorithms and making it usable within a software demonstrator. To achieve this goal, the first step is to define the object of consideration and a concept for the data-driven and formal recording of operation sequences. For this purpose, the process of process sequence determination is analyzed, described and a process model is created. The available data records are then prepared by describing the quantity, structure and quality of the data. Once the data has been prepared, machine learning models are developed within a three-stage process, which enable the prediction of valid process sequence steps. These machine learning models are then integrated into a software demonstrator, which makes the use of the machine learning algorithms tangible and operable for the production employee via a user interface. This software demonstrator is integrated and validated in the production environment of the cooperation partners. For validation, a test concept will be drawn up to evaluate the demonstrator's suitability for practical use and a corresponding user study will be carried out

The project is being carried out within a consortium that includes FBK, the companies up2parts GmbH and KWS Kölle GmbH as well as the Machine Learning working group. The consortium also includes the associated partners Lauscher Präzisionstechnik GmbH and Wagner Maschinenbau GmbH.

The project (VorPlanML, FKZ: 01 IS 21 0 10) is funded by the Federal Ministry of Education and Research as part of the "Research, Development and Use of Artificial Intelligence Methods in SMEs" program.