One of the common misconceptions in additive manufacturing is that it is a “plug-and-play” process, in which a 3D model is divided into layers (slicing), which are then converted into motion instructions, or g-code which is uploaded to a printer. With such a setup, the flow of information between the digital model and the printed geometry stops there; and the ability to adaptively update the position of a 3D printer in response to the printed geometry is lost. In reality, there are often differences, especially in WAAM, some of which may be quite significant, between a design geometry and what is actually printed. These variances may include, but are not limited to, discrepancies in layer heights or distortion of the printed structure during printing. Therefore, the ability to adjust robotic motion paths, layer-heights and welding parameters is critical to ensure a stable printing process.
In response, this research establishes a workflow from design to production (Fig. 3). The digital model, generated with the design tool Rhinoceros 3d, contains information on process and input parameters, welding layer geometries and material properties. The weld paths are generated by means of parametric robot programming (PRP), a coordinate determination based on mathematical functions (see chapter Parametric Robot Programming). A feedback of information from the column in production with the digital model by the robot controller enables adaptive updating of the weld paths according to the required production.
During manufacturing, data like welding, cooling, measuring and movement times, as well as the measured layer heights of the manufactured structure are recorded for each layer. The final evaluation of the recorded data, supplemented by 3D scanning for a target/actual comparison, is used to generate welding and milling trajectories for a surface finish, which is not part of this article, as well as to determine, for example, optimized movement sequences or cooling and measuring processes. All stored information results in a rudimentary Digital Twin within the design tool. This was essential in ensuing proper process control as it allowed for a partially real-time feedback of information between the initial 3D geometry and data generated during the printing process.
Process and input parameters
A layer deposition under a stable welding process depends significantly on the selected welding and input parameters, which influence the geometric dimensions of a weld layer and the subsequent material properties [12]. The material deposition is decisively controlled by the defined travel speed (TS), the wire feed speed (WFS) and the CMT process regulation [13]. Fronius offers the "CMT Cycle Step", a welding process in which the ratio of material application (CMT cycles) to the predefined pause interval for the solidification of the melt can be selectively controlled. This targeted control is essential for the additive manufacturing of free-formed column structures with overlaying layer deposition, for which dripping of the molten material (Fig. 6) must be prevented. Table 1 lists the welding parameters for a Fronius CMT Advanced 4000 R welding source with a CMT Cycle Step characteristic and relevant process parameters.
Table 1 Process and input parameters for CMT Cycle Step To prevent the weld metal from dripping a cooling, explained in detail in chapter Process Sequence Parameters, is set after each welding so the structures can cool down.
Material characteristics
To determine the load-bearing capacity of an additively manufactured structure, the material properties were first determined. Table 2 lists the minimal material properties of the wire electrodes Weko 2 G3Si1 and Weko 2 G4Si1 which serve as a reference [14]. Then the measured average material characteristics and standard deviation for each tensile specimen sample line – consisting of three samples each - according to DIN 50125 - E [15] are given. They were recorded in tensile tests on a universal testing machine of the Institute for Steel Construction and Materials Mechanics of the Technical University of Darmstadt with a maximum tensile or compressive force of 100 kN. The specimens samples 1Seam-milled-45 and 1Seam-milled-90 were milled out of cantilevered wall structures (shown in Fig. 4d). They consist of one welding seam per layer with an inclination of 45° and 90° to a vertical material deposition. More detailed information can be found in [16]. The specimens of 2Seams-as built-0 (b) and 2Seams-milled-0 (c) were printed vertically with two welding seams in each layer (a), providing milled and as built surfaces.
Table 2 Material characteristics (Average and standard deviation) In addition to the yield strength Re and the ultimate strength fu, the elongations at fracture were determined according to DIN EN ISO 6892-1 eq. 6. However, the focus of these investigations is on the ultimate strength.
The results show, that the minimum yield strength and the minimum ultimate strength are almost equal or slightly above (except 1Seam-milled-45) the wire manufacturer's specifications within the range of the standard deviation. The same applies to the measured elongation at fracture. The consistent results allow using the material properties of the wire electrode as parameters for the rudimentary digital twin while they should be supplemented by means of further tensile tests.
Parametric Robot Programming
Central to the ability to update welding paths according to recorded information is the use of Parametric Robot Programming (PRP). This process, which has already been illustrated in [17], differs from conventional robot programming in that coordinates are generated by means of mathematical functions which are defined by information recorded during the printing process. Due to the fact that in conventional slicing (Fig. 5a), every single coordinate has to be precalculated and sent as a line of code, the data files generated by PRP (Fig. 5b) are, in most cases, far smaller than conventional ones. In PRP, a series of loops are used to define each layer and equations (right) illustrate how an ellipsoid with varying cross-section can be defined by two loops using Cosine and Sine angles.
The manufactured columns are characterized by a layer deposition with continuously changing transverse offset which influences the layer height. Fig. 6a schematically illustrates the material deposition in vertical orientation and for different overhang angles αt shown in Fig. 6b to d. Up to a horizontal offset of about 10 % of the weld thickness t, a stable process results with angles αt and in layer heights h depending on the selected offset (Fig. 6e)
Above α0.25t, the molten material starts to drip off under neutral torch position [16, 18], so that an unforeseen loss of material and height occurs. In addition, the contact tip to work distance (CTWD) which describes the distance between the contact tip of the welding gun and the work piece changes. In case of significant deviations, this has an influence on the layer width and height to be applied, which would further cause an unplanned manufacturing of the column geometry and could lead to an interruption of the welding process. The PRP prevents this error development during the manufacturing process by dynamically adjusted weld paths through permanent monitoring of the geometric height development of the column structure.