Smit Thermal Solutions created the control software for its latest oven with a limited amount of manual coding. The software was then thoroughly tested before the hardware was available by running it on a digital twin of the machine. These advancements were to a large extent enabled by the model-based low-code approach supported by the Cordis Suite development platform.
“We’ve taken some significant strategic steps in recent years,” says Bob Veldt, software architect at Smit Thermal Solutions, the Eindhoven-based supplier of customized thermal equipment and processes. It started a couple of years ago with the relocation of most of the machine assembly to China, followed by the adoption of a modular hardware approach (see box). Recently, Veldt introduced a model-based low-code approach to software engineering, based on the Cordis Suite platform.
“When I joined Smit in November 2020, we were writing PLC code by hand, some of it even in ladder diagrams,” recalls Veldt. “Having worked for Cordis, I immediately saw that we could make a huge step there as well. Although we don’t have many competitors, we strive to reduce project lead times to maintain our competitive edge. This requires us to be more predictable, which includes minimizing the risks in our software development. Using Cordis Suite has allowed us to streamline our development by generating close to 100 percent of our machine control software, while at the same time boosting our code quality.”
Smit’s core competencies include applying deposition layers on a wide range of substrates, both rigid and flexible, using processes like thermal evaporation, (plasma-enhanced) chemical vapor deposition and atomic layer deposition. “Solar panels are a big market for us,” illustrates Veldt. “By evaporating a metallic powder, we create an active layer onto a glass substrate. Our ovens can be anywhere from 20 to 75 meters long, depending on the required throughput.”
The panels are transported through the machine with a 5-millimeter gap between them. On their way, they’re subjected to different process steps, including vacuumizing, heating, deposition and cooling, executed by the different subsystems. “For each subsystem, we have a separate cabinet, containing such components as power supplies for servo motors, solid-state relays for heating loops and IO for sensors and actuators,” details Veldt. “An Ethercat network connects the components to a central industrial PC, running Windows next to a PLC on a separate processor core.”
For the new modular setup, Smit has standardized on Beckhoff hardware. “Beckhoff is supported by Cordis Suite whereas our previous suppliers aren’t, so that further facilitated our switch,” explains Veldt. Currently, Cordis Suite provides code generation for five hardware platforms: Beckhoff, Bosch Rexroth, Lenze, Siemens and C# on PCs, with four more in the pipeline.
Low-code and digital twin
Development started with creating a low-code UML model of the system. In Cordis Suite Modeler, the machine was broken down into its constituent parts. Veldt and his software engineers took this decomposition and defined for each component the states it can be in and the possible transitions between these states. They then specified the actions to be taken in the states and the conditions for executing the transitions. From the resulting model, the code for the Beckhoff PLC was generated in full using Cordis Suite – at the push of a button.
The generated code was first tested on a simulation of the machine being built. Veldt: “We’ve created a digital twin of the transport subsystem by connecting a simplified CAD representation in Siemens NX MCD to a soft PLC. Cordis Suite provides the gateway that acts as an intermediate: based on the modeled behavior, it relays sensor data from the transport subsystem to the PLC and control commands vice versa. For the temperature and vacuum subsystems, we’ve hooked our own simulator to the Cordis Suite Gateway. With these elements, we could simulate the entire machine and parallelize software and hardware development as much as possible: we were able to do most of the testing already on the digital twin, before the hardware was available.”
Integrating the software with the actual hardware still required some effort, but that was mainly due to the platform switch to Beckhoff. “Two-thirds of the work was commissioning the new devices,” says Veldt. “The remaining one-third was integration testing. Here, the Cordis Suite Dashboard was very instrumental. Using our models, it allowed us to view and check any part of the control software in real-time, down to every command and every state. The Cordis Suite Dashboard is like our machine’s user interface, only with much more detail.”
Despite his earlier positive experiences with Cordis Suite, it still surprised Veldt that many parts of the software came out ‘first time right.’ “Take the sequential positioning of the panels on the conveyors,” he gives as an example. “After having completely tested this functionality on our digital twin, we ran it on the actual hardware and we measured a consistent inter-panel gap of 5-6 mm – it worked, straight away! Of course, it wasn’t all hunky-dory. For example, we didn’t catch all exceptions, but with the help of the Cordis Dashboard, those problems were quickly found and fixed.”
Adoption of Cordis Suite helped improve the software development process at Smit in other ways as well. “It’s great to have a graphical software design: it’s much easier to read than code and we can share it on a big screen. This makes the process transparent to all, we can keep the overview and learn from each other much faster. The focus has completely shifted from ‘traditional’ coding to where it should be: the actual design. The automatic code generation further increases the reliability of the process, while the integration with simulation and the ensuing parallelization with hardware development allow us to reduce the time to market.”
Veldt foresees additional gains in future projects. “During this latest development, we’ve put a lot of work in laying the foundations: learning to model, familiarizing ourselves with new hardware devices, creating models for them and combining everything into a completely new platform. In our next project, we can build on this. I expect that the reuse of components and knowledge is going to save us a lot of time and effort, on top of the improvements we’ve achieved already.”
This article was written in close collaboration with Cordis.