Real-time data visualization of welding robot data and preparation for future of digital twin system


The framework, which uses a digital twin to present the current welding robot movement and record the motion and welding parameters—for parameter estimation, real-time system monitoring, seam sequence optimization will be fine-tuned18,19,20,21,22. The ideal weld parameters will also be supported by mathematical modelling23,24. The framework will become complete when the optimal values are proposed, and autonomous when adapted. In addition to the real welding of the parts, the virtual robot is working in parallel, which is monitored on the connected computer. The extended framework now includes a comparison with the recommended values and recommendations, which will be detailed in a future paper. The creation of the digital twin starts by detecting the connectivity options and activating them. The connection consists of two parts, firstly the control of the robot and secondly the connection of the power supply of the welder to the test system.

Connecting the robot controller and the power source in the manufacturing system

The following architecture was given as an example of how to build the system. The diagram of the Fig. 3 can be used to present the ERP-MES levels already mentioned in the introduction. The area of each level of the production system is indicated by a dotted line since the production processes are displayed at these levels. Elements of the IT architecture that have program functions were identified at the design stage. These are the RobotVis and RobotComm applications. The connection of the IT elements with the relevant levels of the automation architecture is as shown in the figure. Thus, the Robot as a tool attaches the information from the field and sensing levels back to the supervisory level.

Figure 3
figure 3

Integration of IT components into the automation structure.

The RobotVis program aggregates all the data, can create the record, and also performs the visualization of the current state. Later, the replay of the recorded data can create new opportunities for optimization. The possibilities are also expanding along the lines of cost efficiency. To design the test system, the first step was to design the connectors. The control of the robot can be connected to a PC via a network cable. The following types of devices were used to assemble the system:

  • The robot arm Yaskawa AR1440—Article number: 9519201 organised in cells; two-axis rotator: Yaskawa MT-1-250S2D Article number: 124614100

  • Welding machine: Fronius TPS400i—Article number: 4,075,179

Both the robot and the welding power source use standard ethernet and TCP/IP protocols respectively. Finally, by using a switch and configuring the devices, an external network was set up through which both the power source and the robot were able to communicate with each other. In contrast, an external computer was able to communicate with them. This made it possible to read the motion parameters from the welding robot in real time. Which are the three (actual) motion coordinate values (x, y, z) and the rotation values (Rx, Ry, Rz) along the axes associated with them. To connect the power source of the welder to the PC, it was also necessary to use a power distribution unit, as there were not enough connection points. On the display of the equipment, the values to be applied during welding can be selected as templates to be followed during Windowsing process and digitally recorded in the log file being created. The output signals include, in addition to current and voltage, welding speed, wire feed rate, and the time elapsed since the robot controller was started. Another important piece of information is the active/passive index of the actual welding process, i.e. whether welding is taking place or not.

Create a virtual robot (preparation of digital twin)

After aggregating the data, it is possible to create the initial steps of the digital twin. This required a software package that could handle windows and graphical user interfaces, display 3D models, and communicate via socket. A framework for this purpose was already under development at the university and was chosen. Since a dynamic library .dll loader module and a python interpreter package linked through it were available, the development time could finally be accelerated. The complete system finally consists of two separate programs, named RobotComm and RobotVis (Fig. 4). Communication between each subpart of the software stack takes time, so it was important to create the simplest architecture possible.

Figure 4
figure 4

System configuration process.

Using the operating system’s socket programming interface, the RobotComm program links the robot to the RobotVis program (Fig. 5). Which currently only requests and transmits position data. Ideally, this functionality would be part of RobotVis, but the communication software library available for the welding robot was developed using the .NET framework and was not currently directly embeddable due to limitations.

Figure 5
figure 5

The main purpose of RobotVis (Fig. 6) is to aggregate, process and display data. After start-up, it automatically establishes a connection with RobotComm and, on user request, with the welding power source. From here, the data is read by the unit’s built-in web interface, which automatically updates the values displayed in the browser via WebSocket. A direct re-implementation of this communication was planned, but a solution with less error potential was (temporarily) developed. Currently, the data is read by a browser test and automation package called Selenium, using a Python interpreter coupled with the dynamic library loader functionality of the operating system.

Figure 6
figure 6

RobotVis software in process.

Recording of DATA

Data collection and recording was done in files with the extension .csv. When the TorchSignal changes from 0 to 1, this is when the actual seam creation takes place. After comparing several .csv files, a group repetition of the data series is shown, selecting a recommended welding template, and the recorded parameter series periodically show characteristic properties. The recorded .csv files consist of 11 columns of nearly 1000 rows each, with a size of 100 kbyte (Table 1).

Table 1 Data table generated from the measurement data.

In total, 150 parts and 450 seams were produced, which means 150 files (.csv) with all the associated data:

  • The time elapsed since the system was switched on;

  • Coordinates of movement—6 columns;

  • Process of welding (yes/no);

  • Welding parameters—3 columns.


Leave a Reply

Your email address will not be published. Required fields are marked *