In our everyday practice of model-based systems engineering in various branches of industry, we repeatedly encounter the question of precise and easy-to-handle versioning of models. This question is becoming increasingly important due to legal regulations, stricter safety regulations and the significant increase in product variants. Therefore, we are responding to a frequently expressed wish and answering the question in this white paper as to whether clean versioning is possible at all when modelling complex cyber-physical systems. To do this, we first look at what constitutes a complex system and why clean release is more difficult today when modelling such systems. Then we turn to practical matters by analysing the versioning process in more detail and describing the resulting requirements for tools. Finally, in the summary you will find the recommendations for achieving clean model versioning based on our decades of experience. So follow us on our journey into the land of unlimited variants and learn how to always keep them safely under control with the help of clean versioning.

With the proliferation of cyber-physical systems and rapidly increasing digitalisation, companies today are facing major changes. Mechanical and electrical requirements are increasingly taking a back seat, while software is taking over most of the value creation. However, the systems are also becoming more complex and challenges are arising that were previously unknown. A central solution approach for this is model-based systems engineering (MBSE). Here, information about a system to be developed is no longer based on documents, but on models. These models are usually created on the basis of the UML or SysML specification and help to better cope with central challenges such as safety, agile development processes and the work of distributed development teams.

Since many versions and variants of models naturally arise in such an agile and complex environment, clear and secure versioning is particularly important. In order to strengthen trust in model-based systems engineering, it has to be clearly demonstrated that clean versioning is guaranteed and controllable at all times in the development process.

Precise versioning becomes particularly important when using agile methods, since new versions are created at short intervals and you still have to keep track of everything. Agile software development refers to approaches in the software development process that increase transparency and the speed of change and are intended to lead to faster use of the developed system. The design phase is shortened and attention is paid to achieving executable software as early as possible. This is regularly coordinated with the client in order to increase customer satisfaction.

Agile software development is characterised by self-organising teams and a step-by-step approach. In international companies, there is a clear trend towards virtual teams that work together across regional, national and cultural borders as well as time zones.

For the practical implementation of clean versioning in the development of complex systems, suitable tools are needed. In this paragraph, we will show which requirements for these tools result from the circumstances described above.

In the development phase of complex systems, innumerable model versions naturally arise. Therefore, the most important function of a tool is the detailed comparison and easy merging of different versions. This requires a precise and fine-granular recognition of changes that have been made. Conventional approaches in software development use line- and text-based applications for this purpose, but they are not sufficient for graphical models. In addition, knowledge of the semantics of the modelling language is also important. A change report at database level or through XML files is neither user-friendly nor does it allow for a clean merge of changes.

In order to version well with a tool, it is not only important to make precise and complete changes at the model level, but also to present the changes in a way that is easy to understand. Model elements can be moved, changed, deleted or added. Moreover, such elements are often represented in several diagrams. In order to maintain an overview, the changes must be easily recognisable. Only in this way the developer is able to get a good picture of what is happening and also understand the changes made by other contributors.

When several people make changes in models, conflicts can arise. Therefore, a tool must calculate these conflicts precisely. After all, a conflict should only be reported if it is not automatically resolved during the merge and human interaction is necessary. Then, with the help of the tool, it must be possible to easily merge these changes and thus eliminate the conflicts. In this process, it is the task of the tool to reliably ensure the validity of the model. Since proven processes, workflows and tools are already used in practice, models must be able to interact with them. The modelling tool must therefore enable seamless integration into existing versioning systems (SVN, Git, PTC, etc.) and industrial application lifecycle management (ALM) tools.

Today, entire supplier chains are involved in the development process of complex systems, each of which contributes only a part to the overall system. In classic „product line development“, it is therefore inevitable that parts of models can also be processed by different suppliers with one tool and then reintegrated into the overall system. This process is further complicated by countless product variants and the trend towards so-called „platform strategies“ (e.g. one floor assembly for different vehicle types).

This means that it must be ensured that different elements and parts of the overall system can be worked on by different development teams at different times and in different places and ultimately be integrated back into the overall system. Thus, an overall system is actually a system of many subsystems. In this scenario, a tool must ensure that there are no system conflicts or that information is not lost either when model parts are detached or reintegrated.

In the course of the increasingly agile orientation in software development, our customers ask us again and again whether modelling and agility go well together. Our opinion on this is quite clear and has also been proven in various studies: Only with properly applied modelling can agile processes be implemented at all in the face of increasing complexity, if you also want to document all regulations and requirements in a comprehensible way. As an example, in this white paper we have dealt with the question of whether clean versioning is possible at all in complex systems.

Our answer to this is clearly yes, but it requires corresponding experience and suitable tools. According to our decades of experience, the tool chain we recommend for implementation around Enterprise Architect, LemonTree and Git is currently the best prerequisite for realising clean versioning at the cutting edge of technology. In our opinion, the potential of this tool chain is currently unique on the market and opens up completely new possibilities for our customers.

LemonTree has been specifically designed to meet the above requirements for tools:

• Accurate change detection and merging
• User-friendly representation of changes
• Precise handling of conflicts in changes and their interdependencies
• Editing of partial models
• Reuse of proven processes (Git, DevOps etc.)

The central task of LemonTree is to support team-based collaboration in today‘s complex projects in the best possible way, and it is constantly being expanded with new releases. For example, a partial model (or the entire model) can still be managed as EAP(x) in versioning systems such as Git, which effectively supports the work of distributed teams. If a component of the model has evolved during parallel editing, it can be easily re-imported thanks to the intelligent merge function. The dependency analysis integrated in LemonTree is used to define model parts more precisely before export. Mutual dependencies can be examined or eliminated if necessary. Simple dependencies and cyclical dependencies between model packages are clearly shown.

Contact LieberLieber!

by phone under +43 662 90 600 2017