An increasing variety of production goods and ever higher quality requirements mean that the production machines required for this are also becoming increasingly complex. Many manufacturers are pursuing a modular approach - machines and systems are assembled from a repertoire of fully developed machine modules according to requirements. However, this raises the question of how the modules and their interaction can be tested without the need to physically assemble them. This is where virtual commissioning comes into play.

The idea is simple: with the help of a computer simulation - a virtual machine - the interaction between mechanics, machine software and product can be simulated and verified in a wide range of variants. As most of the time-consuming - and in some cases quite costly - tests have already been carried out in the simulation, this shortens the 'real' commissioning of the physical system.

Modeling the machine model

The fact that virtual commissioning is not always so easy to implement in practice is shown by the wealth of different approaches. However, all approaches to virtual commissioning have one thing in common - they are based on a simulation model of the machine or system, the implementation of which has a decisive influence on the quality of the results. The more accurately the model replicates the functional behavior of the real system, the more meaningful the results from virtual commissioning will be. However, simulation is never a completely accurate representation of reality, especially as the greater the accuracy, the greater the effort required to create the simulation model. Therefore, the user must first consider which aspects of reality must be reproduced in the model and which can be neglected.

In contrast to traditional sequential development, virtual commissioning (VIBN) allows parallelization and early testing of the machine software.

© Mathworks

For example, it is not necessary to describe the physical behavior in detail for the verification of logical step chains, such as for starting the machine or for error handling. For the determination of suitable controller parameters from virtual commissioning, on the other hand, precise modeling of the functionality is absolutely essential. The presentation of the simulation results also varies from application to application. A graphical 3D representation can certainly be helpful for verifying the logical machine behavior. For control engineers, however, the 'measured' values and curves from the simulation are more important. Finally, the various solutions for virtual commissioning also differ in terms of the different connections between the machine model and the control software. Three successful approaches have emerged here - the simulation of the machine and control logic in a common model, the coupling between the virtual machine and the controller and the generation of real-time code from the machine model and subsequent hardware-in-the-loop simulation.

Selection of the level of detail

The first and decisive step for the use of virtual commissioning is the correct choice of the level of detail. This is derived directly from the expected results.

A common use case is the testing of complex logical step chains on the virtual machine. These are usually implemented for machine sequences such as starting the machine, switching between different operating states or shutting down in the event of an error. In conventional machine development, the sequences are programmed manually directly on the industrial controller, i.e. a PLC or an industrial PC, and tested on the physical system. Even with manageable step chains, comprehensive testing in particular often turns out to be a Sisyphean task. This results in exploding commissioning times and direct costs in the form of faulty goods. Some scenarios cannot be carried out at all, as the corresponding tests would pose a risk to people and machines. As machine software becomes increasingly complex, it is simply impossible to comprehensively process all test cases, meaning that more and more untested software ends up in the field, leading to production downtime or even accidents.

The detailed evaluation of simulated curves is carried out directly in MATLAB and Simulink and enables (virtual) measurements to be taken in places where they would be impossible or difficult to take in reality.

© Mathworks

Virtual commissioning provides a remedy for this. With a model that emulates the logical behavior of the machine, even extensive logical step chains can be tested comprehensively and, as a rule, completely automatically. Manufacturers such as Mathworks also offer solutions for a coverage analysis of the existing test scenarios and for the automatic creation of reports.

At the other end of the scale are models with a high level of detail for the design and parameterization of controllers. In this use case, virtual commissioning also helps to save time and costs during commissioning on the real machine. By simulating the controller software and virtual machine, both the most suitable control algorithm and optimized parameter values for the controller can be identified. Determining suitable parameters on the machine is often a lengthy process, especially in the case of increasingly sophisticated algorithms that are intended to further improve product quality. In practice, the parameters obtained from virtual commissioning represent good starting values that significantly shorten real commissioning. In this application, it is not enough to create a logical image of the machine. Models for the virtual commissioning of controllers require a detailed description of the physical behavior of the machine limited to the components of the machine that are relevant for control.

Verification of the control software

To verify the control software - be it a complex logical step chain or a control algorithm - the virtual machine must be linked to the control program. There are three frequently used variants for this:

• In the first, the machine model and control system are merged into a common system model. Tools such as Simulink not only allow the creation of models for mapping the machine or system, but also support the developer in the design of control logic and control algorithms with subsequent generation of real-time code for the PLC.
• In the second case, the user connects the virtual machine to the control program. This is done via protocols such as OPC UA or UDP, for example, or via proprietary interfaces from the various manufacturers.
• The third variant is virtual commissioning using hardware-in-the-loop simulation. Here, the machine model is converted into executable code by means of code generation and executed on a real-time system. Through the physical

connection to the industrial controller - for example via a fieldbus - the machine software is made to believe that it is connected to the physical machine. By emulating the behavior of the system, different test scenarios can be carried out while the physical machine is still being set up.

There are also different approaches for virtual commissioning with regard to the presentation of the results - depending on the task at hand. Graphical representations of the machine or system - often even in 3D - give the developer the feeling of having the physical machine in front of them. This approach helps to make initial statements about the logical behavior of the system.

Presentation of the resultsDo all modules basically function as expected? Are there any jams or collisions between the products in the system? Does the machine actually stop when the shutdown is activated?

However, when it comes to the detailed evaluation of a system's functionality, virtual sensors and the data they record are the method of choice. In contrast to the real machine, unlimited measuring points can be added to the virtual machine in the model - even in places where this would not be possible in reality, for example due to space restrictions. This allows sequences to be recorded during the automated tests, which show the commissioning engineer whether the system is actually running as it should or support him with the controller parameterization.

Virtual commissioning is not so different from conventional commissioning on the physical system. Here too, an initial look at the system helps to verify the basic behavior. Measured values read in by the PLC or recorded with the help of measuring devices are then used to test the detailed machine behavior.

Virtual commissioning based on model-based development in Simulink is the key to keeping commissioning times and costs low, even as the complexity of machines and systems continues to increase in the future. However, it is important to note that not all virtual commissioning is the same. The right choice of solution approach is ultimately responsible for success or failure.

Author:
Philipp Wallner is Industry Manager for Industrial Automation & Machinery at Mathworks.