DIGITALIZATION OF A ROBOT-BASED DIRECTED ENERGY DEPOSITION PROCESS FOR IN-SITU MONITORING AND POST-PROCESS DATA ANALYTICS

To enhance quality and productivity as well as reduce trial and error in directed energy deposition (DED) processes, data analytics are essential. Collecting different signals like melt pool images, melt pool temperatures, environment conditions, and robot data during production, enables insights into the process. In general, the collected data can be either used during the process for in-situ monitoring or for post-process analytics anytime after the print job. Hence, in order to generate a digital twin of the process, the software architecture has to cover both cases, real-time data processing as well as data storage and access for post-process analysis. For this purpose, a software architecture consisting of four different layers, namely the physical-, distribution-, storage- and application layer was developed. The physical layer takes care of the sensor integration in the process and the real-time connection between the sensors and the data collection device. This device hosts at the same time the developed in-situ monitoring software ("DED Process Monitor"). The distribution layer is the connection layer between the physical- and storage layers by organizing the data transfer between the sensors and the database used by potential applications. The storage layer is embodied by a time-series database. A web- based software tool ("DED Process Analyzer"), representing the application layer, makes the database accessible for multiple users on different devices e.g. computers, notebooks, tablets, and smartphones. Through this architecture, a robot-based directed energy deposition process was digitalized including in-situ monitoring as well as postprocess data analytics. The DED Process Monitor is useable for the operator through a human-machine interface at the robot cell to observe the current production. This supports him in reacting to process deviations by adjusting process parameters or early stopping the process. In contrast to the DED Process Monitor, the DED Process Analyzer is realized through a decentralized web-based software tool. Therefore, the tool is useable for every employee with a suitable device connected to the internal network. Thereby multiple data engineers can use the collected data simultaneously for their analysis. Hereby different data analytic options are available. The scope starts from simple 2D data visualizations and ends up in post-process robot path simulations as well as complex 3D-Geomapping of sensor data. Through the 3D-Geomapping process sensor values are considered together with robot position data. This increases the extractable information content enormously. This method allows a geometrical representation of the sensor data over the contour of the printed part. By mapping sensor data to specific positions given by the robot, the sensor values are mapped to geometrical areas in the printed part. Thereby critical areas in the part can be identified. Using this information, later quality tests can focus on these areas instead of the total part. This saves time and costs. Furthermore, the DED Process Analyzer can be used to easily perform data preprocessing tasks like filtering and cleaning to create data sets which can then be used to train machine learning models.