Decision Diagrams for the Verification of Consistency in Automotive Product Data

Berndt R (2016)

Publication Language: English

Publication Type: Thesis

Publication year: 2016

Publisher: FAU

City/Town: Erlangen

Pages Range: 134



The automotive manufacturing domain of these days is faced with several challenges. According to Maslow's hierarchy of needs (American psychologist, 1908-1970), the desire for individuality will steadily increase along with the growing prosperity of people. This is why the concept of mass customization has been evolving since the 1990s, i.e. "producing goods and services to meet individual customer’s needs with near mass production efficiency". Adding to this comes the progressing globalization which involves new competitors and saturated markets, especially in Europe and Northern America. In turn, car manufacturers have not only continuously developed their product range but also reduced the time to market (TTM). To achieve this, the simultaneous engineering (SE) paradigm for the development cycles of cars was adopted. On that score, well-performing flows of information are considered to be among the fundamental prerequisites. In order to implement them, a high degree of standardization is required. As a side effect, many companies struggling for markets suffered a loss of flexibility—but flexibility is essential to efficiently react to both rapidly changing framework conditions and the fast technological progress. The multitude of different concepts established by various business units is usually reflected by a high complexity within the underlying data structure. Due to this and the limited technical options regarding the design of information technology (IT) systems that prevailed in the past, in many cases a truly diverse and interlaced system topology has been evolving. While the flows of information keep getting more complex and non- transparent, the consolidation of IT system architecture is becoming ever more difficult. Nevertheless, the interdependence of business processes and IT systems is continuously growing as well. As a result, the influence of data w.r.t. decision-making and process coordination is steadily rising, thus leading to a new relevance of data quality (DQ). This is why diverse methods and management approaches to ensure high DQ were developed. But due to the application-oriented nature of many DQ problems, these methods often remained abstract and theoretical. Faced with these circumstances, the AUDI AG started a project to replace the paper- based human-machine interfaces (HMIs) controlling the assembly process of cars by an IT system. Its big advantage is given by the reduction of complexity for the operators. Still a problem is that the new system heavily relies on correct input data. Being aware of that, the AUDI AG established this doctoral research study to ensure high quality of the involved data. In the course of this thesis, a new technique is introduced focusing on the verification of consistency within the different data sources. For this, a formal model of the data including a set of consistency requirements will be developed. Based on that, multi-valued decision diagrams (MDDs) will be deployed, on the one hand to encode large sets of configurations (also: customer orders), and, on the other hand, to efficiently carry out set operations which are the basis for the verification process. The verification method requires an MDD-based representation of the feasible design space, i.e. the set of all valid configurations. But the compilation of the configuration problems of highly customizable cars leads to high memory consumption—determined by the number of nodes. The latter, in turn, can be influenced by the variable ordering, i.e. the arrangement of levels within the diagram. Therefore, both well-known and new variable ordering approaches will be applied to keep the size of the MDDs practicable. In addition to that, during the compilation of the configuration problems, the sequence of processing the constraints determines the dynamics of memory consumption. In order to keep the growth of memory consumption low, two new algorithms will be introduced. It will be demonstrated that the combined application of the presented methods allows to construct the MDD representation of the feasible design space of all investigated products. The results of the MDD-based verification technique will be validated by using a SAT- based technique (from 'Boolean satisfiability problem') which also allows to compare the performance of both approaches.

Authors with CRIS profile

How to cite


Berndt, R. (2016). Decision Diagrams for the Verification of Consistency in Automotive Product Data (Dissertation).


Berndt, Rüdiger. Decision Diagrams for the Verification of Consistency in Automotive Product Data. Dissertation, Erlangen: FAU, 2016.

BibTeX: Download