Among other things, the digital transformation in mechanical engineering has ensured that the software on production systems is playing an ever greater role - and is becoming increasingly complex. Together with the wide range of variants resulting from increasing demands on flexibility and the resulting greater modularity of machines and systems, this is increasingly leading to delays in commissioning and errors during operation.

Virtual commissioning provides a remedy. Instead of time-consuming tests on the physical system, the software is tested and optimized in interaction with the remaining mechatronic components using simulated - virtual - machines in different scenarios - often at a time when the physical system is not yet available.

Exploiting the full potential of the models

While virtual commissioning and the associated use of simulation models is already becoming increasingly widespread in mechanical and plant engineering, it is often forgotten that simulation models can have a much wider range of benefits than 'just' virtual commissioning. The development approach of model-based development, i.e. model-based design, places the model at the center of the entire development process. Models are created as early as the planning phase for the machine or system and are continuously developed further from the design stage through to commissioning.

Schematic representation of the parallelization of development through virtual commissioning

© Mathworks

The resulting parallelization can shorten the lead time from the start of the project to the start of production (see image). The simulation models play a role as the basis for digital twins over the entire lifetime of the system.
In general, it is important to ask yourself the question of return on investment (ROI) right from the start. This involves comparing a transparent list of the expected savings, such as machine downtime costs during operation, travel expenses for service calls, material and energy consumption during physical commissioning, etc., with the expenses - such as personnel costs for creating and maintaining the models, building up know-how in the team, license costs and so on.

The greater the accuracy of the simulation models, the greater the effort required for modeling, so care should be taken to start with a simple model and increase the complexity as the project progresses, if necessary. Model-based development should also not be applied to the entire machine or system in the first project, but should start with a relevant sub-component. On the one hand, this allows the added value of simulation to be demonstrated relatively quickly and, on the other, the development method can be introduced gradually without having to completely change established processes in too short a time.

The modeling

Example of a simulation model of a handling robot.

© Mathworks

The effort required for modeling depends on several components. Detailed modeling of the individual components of a system, for example drives or sensors, provides the most accurate results - but at the expense of modeling and simulation time.
A 3D behavioral model of the entire system can be helpful for the verification of logical software parts, but is not suitable for the evaluation of physical results, such as the accuracy of position control or the correct design of the drives.

Simulation environments such as Simulink therefore offer different modules for modeling, which map the individual aspects accordingly. For example, logical state models and step chains are modeled in Stateflow, while Simscape offers the option of reading in 3D CAD data as a physical model.
The simulation model of the machine or system created in this way is then used in several steps over the entire development and commissioning period to continuously verify the interaction between mechanical, electrical and electronic systems and software. This workflow has not only been established in the automotive industry for years, but is also successfully used in mechanical engineering by leading companies such as Krones, Metso and Tetra Pak.

The desktop simulation

Philipp Wallner is Industry Manager at Mathworks.

© Mathworks

The first step is to run the simulation on a desktop computer. The machine's behavior model is simulated together with the algorithms that will later run on a PLC or an industrial PC, for example sequence controls, control technology or AI algorithms. This involves running through different scenarios that would not be possible on the physical machine or would only be possible with great effort. The advantage here is that the desktop simulation does not run in real time, meaning that a large number of different tests can be carried out in a short time and sometimes in parallel. In addition, values can be recorded in the simulation that would not be measurable in practice due to a lack of appropriate sensors.

Code generation and hardware-in-the-loop

Hardware-in-the-loop simulation with Simulink and Speedgoat hardware

© Mathworks

Once the desktop simulation has been successfully completed, the next step is the automatic generation of executable real-time code. Tools such as Simulink offer the option of creating source code in various languages (ANSI-C, C++, IEC 61131-3, etc.) directly from the simulation model. This means that the software components already tested in the simulation, for example for control technology or predictive maintenance, can be converted into real-time-capable and extensively documented program code and transferred to the industrial control system without having to go through the usually error-prone and time-consuming manual programming process. Optionally, the documentation for the software can also be generated. Model-based development with Simulink supports the common industrial control platforms.

In addition, real-time-capable code can be automatically generated from the part of the simulation model that represents the physical and logical behavior of the machine or system. This then runs on real-time hardware from Speedgoat, for example, together with Simulink Real-Time. This real-time simulation of the machine model is then connected to the PLC via common industrial protocols such as Profinet, Ethercat, Ethernet/IP or Modbus. This simulates the behavior of the physical system to the industrial controller so that various tests can be carried out - this time under real-time conditions.

Digital twins - using models beyond virtual commissioning

The models developed for virtual commissioning also offer decisive added value beyond the development and commissioning phase. They are increasingly being used as so-called digital twins - i.e. virtual representations of the running system - throughout the entire service life of the machine. The digital twin continuously receives measurement data from the physical system and compares it with the values calculated from the simulation. This allows deviations to be detected at an early stage and failures or quality fluctuations to be predicted. Digital twins based on simulation models are increasingly being used in the field of predictive maintenance. The models are embedded in the IT infrastructure of the production plant. Tools such as Simulink offer corresponding options for the use of models on edge or cloud systems.

Virtual commissioning helps mechanical engineering to virtualize commissioning to a large extent in times of increasingly complex systems, thereby saving time and costs. With the help of model-based development, the simulation models created in this way are used throughout the entire development phase - and increasingly beyond in the form of digital twins. With the gradual introduction of virtual commissioning and the appropriate software tools, this step usually pays off in the very first project.