Figure 3: Model-based design workflow for the development of technical systems.

© Mathworks

Such models clearly represent the system as an executable specification and contain its components, including the connections and interfaces.

Requirements can be directly and traceably linked to components and software. At every point in the development process, tests can be used to verify whether the system works as intended and fulfills the requirements.

Each component is iteratively refined and contains well-defined interfaces. This modularity makes it possible to develop components in parallel that would normally be created one after the other, which speeds up the process. In this way, the model grows step by step into an increasingly realistic representation of the overall system, which can be tested and validated from the outset. Errors are thus detected earlier and can be rectified more easily, more cost-effectively and with less delay(Fig. 4).

Figure 4: Robot arm: Integration of CAD and electrical control to optimize trajectories and control logic.

© Mathworks

Signal flows and states can be monitored and logged at any point. Once created, test cases are used continuously and can be processed and evaluated overnight in a batch-controlled and parallelized manner. In particular, the hardware can also be easily modified without having to be available. This allows hardware and software to be optimized in a depth that would not be possible in terms of time or cost without simulation. The traceability of requirements to the model and back as well as automatic report generation ensure transparent documentation of the entire process with comparatively little effort.

Code generation and validation of the real control system

One of the key advantages of model-based design is the ability to automatically generate software from the model, such as code for structured text and ladder diagrams in accordance with IEC 61131-3. Code generators and integration tools are also available for other programming languages such as C/C++ and various hardware platforms, such as FPGAs, microcontroller systems-on-chip or web and cloud services, which enable the efficient and optimized implementation of technical systems in a wide range of application fields (Figure 5).

Figure 5: Use of models and code generation in different development phases.

© Mathworks

The platform on which this code is later executed is primarily unimportant. The starting point is always the same model - so the developers do not have to commit themselves from the outset. This flexibility makes it possible, for example, to use different control systems depending on the region without having to develop the software multiple times. It is also possible to switch to a more powerful system at a later date and reuse functions in subsequent projects with different requirements. In this way, libraries with reusable functionalities and modules can be built up successively.

Validation of the real control system against the system model

Hybrid simulations are possible in order to prove that the generated software behaves in reality as it does in the model. To do this, the application is loaded onto the physical control system and tested against a system model simulated on a real-time computer. The same test cases as in the simulation can be used for this and their results compared numerically.

Based on the Simulink model, Simulink Real-TimeTM can be used to create, control and instrument real-time applications, for example a controlled system model, which are executed on Speedgoat real-time computers. This approach is also known as Hardware-in-the-Loop (HiL).

With an additional PC running Simulink, the entire process can be monitored and logged live. Errors and problems are rectified directly in the model and new real-time code can be generated immediately to test success. Iterations are possible without interruption and in rapid succession. This configuration also allows operating parameters to be changed during the simulation. This means that the software can be pre-optimized without the need for a physical machine.

Agile development approaches

In the software industry, a separate branch of development approaches called "agile" has emerged for the type of incremental development with continuous refinement described above. It should make complexity manageable and be more efficient than traditional methods. Small teams work on one part of a piece of software at a time, such as the architecture or a component. With this iterative working method, requirements and solutions are often only worked out during development(Fig. 6).

Figure 6: Agile system development with model-based design focuses on simulation and code generation for the incremental provision of design and test artifacts, studies and evaluations.

© Mathworks

At the end of each iteration - called a "sprint" - we evaluate where we are, what is missing and what the next steps should look like. This type of development is characterized by a high degree of flexibility and resilience. Errors or changing requirements can be dealt with more easily than with conventional development.

Agile development is particularly suitable for further developing and updating products while they are already in operation with the user. With this method, known as Continuous Integration/Continuous Delivery (CI/CD), the product continues to grow and is improved or provided with security updates.

The similarities with model-based design are obvious, as here too a product is ultimately created through incremental development of the overall system. With mechatronic systems, the product of a single sprint is not necessarily ready for use, but is initially part of a model. As with model-based design, this is not monolithic, but consists of components that can be changed without disrupting the overall architecture.

In general, Agile places less emphasis on documentation or reports and more on a functioning product, which can be perceived as a gap. Model-based design tools are the ideal complement here, as they offer extensive functions for automatic documentation and report generation. Those who are already familiar with model-based design can therefore continue to use them seamlessly.

Future viability through model-based design and agile

Adding agile methods to model-based design in mechanical engineering has a number of advantages. Firstly, the working methods of hardware and software developers are aligned. New synergies and improved communication are created. Administrators are relieved because the software landscape is also standardized.

Hardware should also be increasingly flexible and upgradeable, and customers are increasingly requesting new functions at a later date. Speed of response will be one of the key competitive factors in this environment. Mastering CI/CD is one of the ways to achieve this. Companies that combine model-based design with agile are better positioned for CI/CD and new business areas. Their engineers have the necessary skills and attitudes, the internal processes support it and the appropriate software tools make it possible.

The authors

Jens Lerche is Senior Application Engineer at Mathworks.

Dr. Marc Segelken is Principal Application Engineer at Mathworks.