Bernhard Wally: Smart Manufacturing Systems: Model-Driven Integration of ERP and MOM, Dissertation, Institute of Information Systems Engineering, Business Informatics Group (BIG), Technische Universität Wien, June 17, 2020.

Automated production systems are following a general technological trend: increasingly complex products, combined with drastically reduced lot-sizes per product variant, as well as shorter response and production times are being demanded. In order to be able to meet these expectations, modern IT systems at all levels of the automation hierarchy are required: from business related software at the corporate management level, down to the programmable logic controllers at the field level. For a well-designed coupling of systems that are located at different levels, it is necessary to find, define, and implement clear data conversion mechanisms – this endeavor is also known as vertical integration. At the same time, it is necessary to automate the inter-organizational data exchange – an aspect of horizontal integration.

If integration succeeds, and decisive systems are flexible enough with respect to their software and hardware components, we can speak of so called smart manufacturing systems that are able to adapt to new situations spontaneously. This flexibility can be related to different elements of a production system:

– Products: modern production systems should be able to produce diverse and variable products. This includes products that have not been conceived when the production system has been engineered and commissioned.

– Processes: the capabilities that are offered by a production system can evolve over time as well. New production processes can emerge, for instance, through new or re-arranged resources, as well as through novel requirements of certain products.

– Resources: are also volatile in various aspects. They can be defect, need to be maintained, are procured, retired, and replaced. Some of them are mobile and can be continuously re-arranged, etc.

Model-driven engineering, leveraging concepts from model-driven software engineering, provides a rich set of techniques, tools and methods for the formalization of domain models as well as for the loose or close coupling of them. Focusing on conceptual models and their instances enables the precise definition of knowledge that can be later on used formally, such as for data validation. Concepts, such as model transformations facilitate the conversion of design time and runtime information between different domains.

In this cumulative thesis, we are recapitulating a selection of own contributions; with respect to the conceptual models we have been employing established industrial standards, in order to facilitate industrial application. By doing so, we have focused on the upper levels of the automation hierarchy, yet we have also partially considered the lower levels. Our approaches and implementations have been successfully evaluated by a number of experiments and case studies and are therefore a contribution towards model-driven smart manufacturing systems. Our research shows that the application of model-driven engineering techniques and tools can be useful within the following scenarios:

– Vertical Integration: by providing mapping and transformation rules for elements of different conceptual models, such as those representing information technology systems of the enterprise resource planning or manufacturing operations management levels.

– Declarative Manifestations: creating explicit and structured artifacts of static (models) and dynamic (transformations) nature. In our work, we are applying thereupon transformations, validations, and code generators in order to convert between domains or doublecheck data consistency.

– Knowledge Inference: transforming declarative information into logical expressions for the efficient computation of new knowledge that can be fed back into models or other data structures for further usage. We have been using this approach for the spontaneous computation of production plans from production system models.

Advisor:        a.o.Univ.-Prof. Mag. Dr. Christian Huemer
Co-Advisor: Univ.-Prof. Mag. Dr. Manuel Wimmer