Real-time monitoring and control of the layer height in laser metal deposition with coaxial wire feeding using optical coherence tomography

In recent times, additive manufacturing (AM) of metal components has become increasingly important for a wide range of industrial applications. AM processes offer the distinctive characteristic of adding material layer by layer, enabling the production of near-net-shape parts based on a digital model.¹ The term laser metal deposition (LMD) describes an economically highly relevant subset of AM processes, where a feedstock in the form of powder or wire is added into a laser-induced molten pool. This deposition principle allows significantly higher build rates to be achieved than with powder bed fusion processes. Moreover, LMD can be employed for the repair of high-value components and to add features or functional coatings to semifinished products.² In terms of the feedstock, a wire offers several advantages over powder, as it is inexpensive and easy to handle, and it leads to less contamination of the process environment.³ Wire feeding can be realized laterally or coaxially to the laser beam. The coaxial variant is enabled by recent developments in beam shaping optics, ensuring a direction-independent process.⁴ 

In wire-based LMD, in general, maintaining a constant distance between the workpiece and the deposition head is critical to process stability. This represents a major challenge since even minor deviations between the actual layer height and the performed height increment can accumulate over multiple layers. In contrast, powder-based LMD benefits from the self-regulating effects of overspray in terms of process stability.⁵ Therefore, approaches to monitoring and closed-loop control of the layer height are especially relevant in wire-based LMD.

Height monitoring in LMD processes is typically realized with contactless sensors, given the high temperatures occurring in the process zone and on the part surface.⁶ Various authors utilized laterally arranged cameras to detect the part height.^7,8 While camera-based systems are comparatively inexpensive to implement, they struggle with the accurate measurement of complex geometries. Another common approach to acquiring height information of the part being built is using measurement systems based on laser triangulation.^9,10 However, as real-time height monitoring in the process zone is difficult due to the existing process emissions, the measurements are usually conducted after the deposition of a layer. These interlayer measurements naturally prolong the build time and, therefore, impair the economic efficiency of the process.

Sensors employing the measurement principle of optical coherence tomography (OCT) represent a promising alternative to the above-mentioned approaches. Using this sensor technology, the relative distance between two reflecting surfaces can be determined. For this, a beam of electromagnetic radiation is divided into two coherent beams, which travel through a reference and a measurement path, respectively.¹¹ Distance information is then obtained by analyzing the interference pattern of the reflected beams. Since the interference pattern is unaffected by electromagnetic process emissions, the measurement accuracy is not compromised.¹² Due to this distinctive advantage, OCT is already an established measurement principle in various laser material processing applications.^13,14 Furthermore, the use of OCT for height monitoring in LMD has been realized by Kogel-Hollacher et al.¹² for the powder-based process variant and by Stehmar et al.¹⁵ for the wire-based process variant.

With regard to closed-loop height control systems for LMD processes, various sensors and control approaches have been applied. In many cases, layer-to-layer control systems were implemented because only interlayer measurements were practicable. Such an approach was pursued by Tang et al.,¹⁶ who used a laser displacement sensor to acquire height information in powder-based LMD. Based on iterative learning control (ILC), the powder flow rate profile for the next layer was calculated, which enabled accurate tracking of the specified height. However, the approach required a considerable calculation time of 68.6 s per layer. For wire-based LMD, Heralić et al.¹⁷ made significant contributions in terms of closed-loop height control. They employed a laser line scanner for interlayer height measurements and an ILC algorithm to determine the wire feed rate for the next layer. The height increment was additionally adjusted based on the average height of the preceding layers. The implemented control system resulted in the production of parts with smooth flat surfaces. Another approach for a layer-to-layer height control was proposed in a previous work by Bernauer et al.¹⁸ The height profile of the part was obtained after each deposited layer using a laser line scanner. A discrete height profile was obtained by segmenting the weld beads along the deposition trajectory. For each segment, an individual proportional-integral (PI) controller defined the wire feed rate for the subsequent layer, enabling the localized compensation of significant height disturbances.

Camera-based systems were frequently used for real-time control of the layer height. For instance, in powder-based LMD, Zhou et al.¹⁹ acquired real-time height information via a charge-coupled device camera. They implemented a dynamic powder splitter to overcome the slow response time of the powder feeding system. The powder flow rate was adjusted by a proportional-integral-derivative controller, which was tuned using the Ziegler–Nichols method. When building thin walls, this enabled a uniform layer height to be achieved. For wire-based LMD, Takushima et al.²⁰ introduced a real-time height measurement system consisting of a line laser and an in-axis camera. By proportionally adapting the wire feed rate, thin-walled cylinders with different layer heights were successfully built up.

Furthermore, measurements based on OCT have been utilized to implement control systems in LMD. In order to measure the working distance in real time, Kittel et al.²¹ integrated an OCT monitoring system into a powder deposition head. This allowed reliable distance information to be obtained despite the powder flow and the optical process emissions. Based on the measurement signal, they implemented a distance control, ensuring a constant working distance when depositing on an uneven substrate. In addition, Becker et al.²² realized a closed-loop height control for wire-based LMD. The layer height was obtained via an OCT-based in-axis distance sensor with a fixed measurement spot. The wire feed rate was adjusted by an experimentally tuned PI controller, which improved the geometrical accuracy of thin-walled cylinders compared to the open-loop case. However, detailed information on the setup of the measurement system and the implementation of the controller were not provided.

In conclusion, OCT represents a promising tool for monitoring and closed-loop control of the layer height in LMD processes due to the capability of performing highly accurate real-time measurements in the process zone unaffected by process emissions. This combination of features has not been achieved by other sensors so far. However, there is remarkably little literature on real-time height control for wire-based LMD, which served as the primary motivation for this work. The objective was to establish a closed-loop height control system that compensates for local disturbances and ensures a defined layer height. In this context, an OCT-based monitoring system needed to be set up first. Subsequently, the relationship between the wire feed rate and the measured layer height was evaluated and a dedicated control system was designed. The performance of the control system was assessed in validation experiments, where a significant height disturbance in the substrate plate was compensated within just a few layers. This work thereby builds on the layer-to-layer control approach presented in Ref. 18.

A stainless steel ER316LSi welding wire with a diameter of 1 mm was used as the feedstock material. The wire was deposited on stainless steel AISI 304 plates with dimensions of 100 × 100 × 10 mm³. The plates were sandblasted and wiped with isopropanol prior to the deposition process to ensure consistent surface conditions and to remove existing contaminants. To minimize thermal distortions, the plates were bolted to a solid steel block.

Figure 1 shows the experimental setup used for the investigations. The core of the LMD system was a wire deposition head (CoaxPrinter, Precitec GmbH & Co. KG, Germany) that enabled coaxial wire feeding within an annular laser beam profile. A 4 kW disk laser (TruDisk 4001, TRUMPF GmbH & Co. KG, Germany) generated the laser radiation with a wavelength of 1030 nm in continuous wave mode. The laser beam source was connected to the wire deposition head via a fiber optic cable with a core diameter of 600 μm. For conveying the feedstock to the wire deposition head, an industrial wire feeding system from the manufacturer DINSE GmbH (Germany) was used. This system involved a front drive (FD101 LS-WB-K), a slave drive (WD 300 FD), and a control unit (FDE-EC150 HW). Furthermore, the existing curvature of the wire was compensated for by a precisely adjusted two-plane wire straightening unit. To move the wire deposition head relative to the substrate, a six-axis industrial robot (KR 60 HA, KUKA AG, Germany) with a maximum payload of 60 kg was employed. The robot had a positioning accuracy of ±0.05 mm and was actuated by a robotic control system (KR C4, KUKA AG, Germany).

To acquire accurate layer height information, the measurement principle of OCT was utilized. For this purpose, an in-process depth meter (IDM) from the manufacturer Precitec GmbH & Co. KG (Germany) was integrated into the LMD setup. The IDM operated based on the principle of Fourier-domain OCT and featured a superluminescent diode with a central wavelength of 1550 nm and an output power of 70 mW. The maximum measuring frequency was 70 kHz. A more detailed description of a similar monitoring setup and its implementation can be found in Ref. 15.

The initial processing of the height data was carried out by an integrated control unit. To configure the measurement parameters as well as the settings for data processing, the software CHRocodile Explorer v1.0.0.0 (Precitec GmbH & Co. KG, Germany) was used. The OCT measurement spot was positioned in the process zone via a scanner unit consisting of two servo-driven mirrors for the x- and y-direction, respectively. The scanner unit was actuated by a dedicated control unit (CUA32-MST, Newson NV, Belgium), which was parameterized in the software Rothor v11.0.41.8 (Newson NV, Belgium).

In the following, the specific settings and the necessary calibration of the IDM sensor unit are described. For the considerations in this work, a reference plane at a constant distance to the wire deposition head was defined. This distance corresponded to the nominal distance between the wire deposition head and the substrate surface (i.e., 103 mm). The usable measurement range in the z-direction was 10 mm, with higher measurement values indicating a shorter distance from the wire deposition head. The zero position (i.e., the farthest position to the wire deposition head within the measurement range) was set 5 mm below the reference plane. This setting with the measurement range symmetrically arranged around the reference plane guaranteed that all measurements yielded positive distance values. To obtain the deviation from the reference plane, which can generally be interpreted as the layer height, the offset of 5 mm was subtracted from the measured value.

The initial processing of the measurement signal was executed by the IDM sensor unit. The measurement frequency was set to 1000 Hz, whereby consecutive windows of 100 measurements were averaged to one value. This resulted in a smoothing of the signal and an effective measurement frequency of 10 Hz, which was fully sufficient for height measurements in wire-based LMD. The distance information was output via an analog interface.

The monitoring system providing the height data is referred to as the OCT sensor in the following. For the in-process height measurements, the lateral distance of the measurement spot to the tool center point, which is located on the wire axis, had to be taken into account. This distance is referred to as the spot distance d_ms. The position of the measurement spot in the process zone is illustrated in Fig. 2. The minimum adjustable spot distance was 3 mm, as the signal intensity decreased significantly below this threshold due to the shading caused by the wire nozzle fixture.

The measurement and control signals were processed by a programmable logic controller (PLC) to which each subsystem of the LMD setup was connected. The PLC was realized on an industrial PC (IPC) (C6030, Beckhoff Automation GmbH & Co. KG, Germany) using the software TwinCAT 3.1.4024. In the IPC, the data processing and data logging were performed with a cycle time of 1 ms.

The time sequence and the parameters of the deposition process were predefined in the KUKA Robot Language and transmitted from the robotic control system to the PLC via an EtherCAT fieldbus network. Employing the PLC, the respective commands were distributed to the wire control unit and the laser beam source, which was also carried out via EtherCAT. Using only EtherCAT for the communication between the subsystems and the PLC ensured minimal signal transfer times throughout the LMD system. The height data from the OCT sensor was transmitted to the PLC as an analog signal and saved to the internal solid-state drive in time synchronization with the other process signals. Throughout the closed-loop process, the wire feed rate was calculated using the PLC based on this height data and commanded to the wire control unit. In order to configure the OCT measurement system, the scanner controller and the IDM sensor unit were connected to a host PC via Ethernet. The longest response time within the system was determined to be less than 20 ms. Given the measurement frequency used and the melt pool expansion of several millimeters, the variables relevant to the control system could be considered quasi-continuous. In Fig. 3, the schematic structure of the overall system is presented together with the main information streams.

In the experiments, individual weld beads as well as thin multilayer walls with a length of 85 mm each were deposited. Based on previous work,²³ a constant laser power of 1500 W and a traverse speed of 1.0 m/min were used. The wire feed rate was varied as required. For the analysis of the system behavior, step changes in the wire feed rate were imposed during the deposition of individual weld beads as described in Sec. IV A.

To assess the effectiveness of the implemented control system, a groove 1 mm deep and 20 mm wide was milled into a substrate plate, causing the measured height to change abruptly. Thin walls were built up perpendicular to the milled groove using a closed-loop and an open-loop process. In the closed-loop case, the wire feed rate was dictated by the control system depending on the inline height measurements. For the open-loop comparison trial, a constant wire feed rate calculated by the feedforward control was employed.

The nominal focal position of the laser beam was set at −6 mm relative to the reference plane. Oxidation during the deposition process was mitigated by a constant shielding gas flow (argon) of 20 l/min. For each new substrate plate, the calibration sequence from Ref. 18 was utilized to ensure a parallel alignment between the substrate surface and the x-y plane of the robot coordinate system.

During all deposition experiments, height measurements using the OCT sensor were carried out. The measurement spot was always positioned on the tool path so that the weld bead peak height was measured. Furthermore, the experiments were observed using a welding camera (XVC-1000, Xiris Automation Inc., Canada) attached to the robot, which enabled a real-time evaluation of the process stability.

After implementing the monitoring system, the influence of the spot distance on the signal quality needed to be evaluated first. For this purpose, a flat substrate plate was scanned horizontally over a distance of 40 mm, whereby the spot distance was systematically varied in 0.5 mm increments. As the signal intensity dropped below a spot distance of 3 mm (see Sec. II C), this value was used as the minimum. For the control approach implemented in this work, measurements had to be taken in front of the process zone (i.e., in the positive traverse direction) to enable a reaction to height deviations starting from the first layer.

Figure 4 shows the measured height profiles for trials with spot distances ranging from 3 to 5 mm. It is evident that larger spot distances led to significantly increased fluctuations in the signal. Especially with spot distances of 4 mm and above, there was a high number of incorrect measurements. Thus, it can be assumed that the signal quality is related to the angle of incidence of the OCT measurement beam. Furthermore, the standard deviation σ of each signal is given in Fig. 5 as a measure of the signal quality. The standard deviation increased significantly from 8.2 μm at a spot distance of 3 mm to 134.7 μm at a spot distance of 5 mm, illustrating the deterioration in the signal quality with increasing spot distances. These findings coincide with the fact that a short spot distance is desirable for the closed-loop control system, as the spatial and, thus, temporal offset between the measurement and the adjustment of the wire feed rate is minimized in this way. Therefore, a spot distance of 3 mm was employed for further investigations.

Besides the requirement for minimum noise in the signal, a high accuracy with respect to the actual height deviations was desired. In order to assess the measurement accuracy, a stepped substrate plate with a height difference of 0.7 mm was scanned (see Fig. 6). The height difference was clearly identifiable in the measurement signal, closely matching the actual height difference. The surface farther from the optics yielded a lower value, making the signal easy to interpret with respect to layer height measurements.

In this work, the wire feed rate was employed as the manipulated variable of the control system. In contrast to the powder flow rate in powder-based LMD, the wire feed rate is adjustable with minimal lag and high accuracy. Therefore, determining the relationship between the wire feed rate and the layer height was critical for designing the control system. In this regard, the main concern was the static gain of the process. An experimental approach was adopted, where a straight weld bead with a total length of 85 mm was deposited. Similar to Ref. 18, a step change of the wire feed rate was triggered by the PLC after a weld bead length of 35 mm. Thereby, a positive and a negative step change between the values of 1.0 and 1.3 m/min were used. These wire feed rates ensured that the process remained within a stable process window.²³ For data acquisition, the surface was scanned with the OCT sensor in an additional measurement step. This yielded height information independent of the melt pool dynamics and the cooling process, as when measuring in front of the melt pool (see Sec. V). To account for the unevenness of the substrate plate, a reference measurement was conducted prior to the deposition process. This measurement was subtracted from the weld bead measurements.

For the identification of the static gain, only the intervals of 25 mm before and after the step change were considered. By this, anomalies in the weld bead resulting from start and stop effects of the process were eliminated. Figure 6 shows the height profiles of the deposited weld beads. In both experiments, a change in the layer height that precedes the adjustment in the control variable was observed. This phenomenon could be attributed to the distribution of the added material within the melt pool, as thoroughly described in Ref. 18. A mean static gain of K = 0.315 was estimated from the experimental data, which was used for the controller design.

The objective of the control system implemented in this work was to maintain a constant distance between the wire deposition head and the workpiece surface when building multilayer structures. To achieve this, the controller had to adjust the wire feed rate depending on the measured height of the previous layer.

For layer height monitoring, the measurement spot can be positioned either in front of or behind the melt pool. In the latter case, the height data must be stored to calculate the manipulated variable for the next layer, as it has been done in various layer-to-layer control approaches. The more straightforward approach enabled by OCT measurements is to acquire height data directly in front of the melt pool. Therefore, the control system represents no closed-loop system in the common sense. In fact, within a single layer, the approach has the character of a feedforward control, as the controller reacts to known conditions. However, as the deposition process is repeated layer by layer and the control system reacts to height deviations in the previous layer, a feedback loop is created.

Since not only the error e from the last layer needed to be corrected but also a new layer was applied, an actual feedforward control was added. The manipulated variable (i.e., the wire feed rate v_w) resulted from the addition of the signals from the feedforward control and the employed control algorithm. The structure of the overall control system is depicted in Fig. 7. In this figure, the indices i and i − 1 refer to the current and the previous layers, respectively. Due to the measurement in front of the melt pool, the surface of layer i − 1 was measured during the deposition of the current layer i. In order to obtain the height of the previous layer h_i−1, the measured height deviation from the reference plane h_d,i was offset by the specified height increment, as indicated by the transformation block. This enabled a comparison of the setpoint height h_s with the height of the previous layer h_i−1, which yielded the control error e. Furthermore, the delay block indicates that the measurement is taken in advance, considering a spatial and, thus, temporal offset between the measurement and the adjustment of the manipulated variable.

The feedforward control consisted of a gain factor, which was derived from the determined static gain of the process. To calculate the output of the feedforward control, the reciprocal of the static gain was multiplied by the nominal layer height, which corresponded to the height increment of the manipulator. In this work, the height increment was set to 0.35 mm based on previous investigations on thin-walled structures presented in Refs. 18 and 23.

Regarding the control algorithm, it had to be considered that a stationary control of the layer height is not achievable since the wire feed rate at each point of a layer is updated only once. For this reason, an integral term in the control law, which ensures steady-state accuracy in a stationary process, would not yield any benefit in this particular case. On the contrary, it might distort the manipulated variable by accumulating errors along the deposition path. Based on these considerations, an integral term was omitted. For the controller design, an instantaneous adaption of the layer height with a change in the wire feed rate was assumed, neglecting the transient behavior of the melt pool. This was justified due to the distribution of added material in the melt pool, resulting in an inherent smoothing effect.

The starting point for deriving the control law was a proportional controller. Initially, the reciprocal of the static gain was set as the controller gain, analogously to the feedforward control. As the static gain of the process and the controller gain cancel each other out in this case, a full compensation of the height deviations in the previous layer can theoretically be achieved, as long as the manipulated variable constraints are not violated. However, this approach might lead to overshoot and oscillations around the setpoint due to model uncertainties and disturbances. This was also confirmed by preliminary experiments, where an excessive intervention by the control system for minor control errors was observed. To mitigate the susceptibility of the control system to oscillations around the setpoint, the controller structure was adapted. For this, a tolerance band was introduced, which was arranged symmetrically around a control error of 0. When the control error was within this tolerance band, the controller output was reduced in a linear manner.

For the employed spot distance of 3 mm, Eq. (3) yielded a number of n = 180 cycles by which the measurement signal had to be delayed. Due to the positioning of the measurement spot in front of the melt pool, unused measurement values remained at the end of each layer. For this reason, the array was reset after each layer so that the deposition of the next layer was not affected by the stored values.

The effectiveness of the implemented control system was assessed through open-loop and closed-loop deposition experiments involving a milled groove in the substrate plate. It is noteworthy that the groove’s depth of 1 mm significantly exceeded the height of a single weld bead and, thus, constituted a significant disturbance.

In the open-loop trial, 11 layers were deposited. For the height increment of 0.35 mm, the feedforward control yielded a constant wire feed rate of approximately 1.1 m/min. The corresponding height data of the individual layers are illustrated in Fig. 8(a). After the deposition of 11 layers, there was still a significant depression in the area of the groove, as no compensation of the disturbance was performed. In the regions before and after the groove, where no disturbance was introduced, the measured part height increasingly deviated from the nominal part height indicated by the dashed lines. In particular at the start of the weld beads, there was an increasing material buildup, which was primarily attributed to the robot’s acceleration phase. During this phase, the desired ratio of the wire feed rate to the traverse speed had not yet been reached, resulting in more deposited material per unit length. These observations from the open-loop process clearly indicated that an intervention of the control system was required.

The height data acquired during a closed-loop experiment with 13 deposited layers are shown in Fig. 8(b). In contrast to the open-loop experiment, the groove was fully compensated after approximately 11 layers. In the areas unaffected by the disturbance, no increasing deviation between the nominal and the measured part height was observed. Thus, the process did not tend to instability with increasing part height. Furthermore, no oscillations of the layer height around the setpoint after filling the groove were observed as in the preliminary investigations with a constant controller gain. This was a result of the linear reduction of the controller’s intervention in the vicinity of the setpoint.

To also counteract the material buildup at the start of the weld beads, the wire feed rate was set to its lower limit of 0.85 m/min during the first 3 mm of the weld beads. In this range, no height data were available, as the measuring point was in front of the melt pool and the control was, therefore, not active. Still, the height profiles in the initial part of the weld beads appeared more irregular and slightly more material was deposited. This was attributed to the fact that the wire was elastically compressed during the initial positioning on the substrate surface. It is noteworthy that in Figs. 8(a) and 8(b), the first and last 10 mm of the walls are marked in gray, as these are dominated by start and stop effects and should be excluded for the evaluation of the control performance.

The wire feed rates commanded during the closed-loop experiment are shown in Fig. 9. In this depiction, the layer index indicates the layer currently being deposited. As the manipulated variable was calculated based on the height values of the previous layer, the colors match the corresponding measurements in Fig. 8(b).

At the groove, major adjustments to the wire feed rate were carried out by the control system. The magnitude of the control action was reduced in the upper layers as the setpoint was approached and the wire feed rate converged to the value calculated by the feedforward control.

In summary, the OCT system enabled reliable measurements of the layer height during the deposition process. In contrast to the laser line scanner from Ref. 18, the height could be measured in real time via the processing laser beam path, significantly enhancing the productivity of the closed-loop process. The spatial requirements of both setups and the additional masses to be moved by the manipulator were similar. However, it should also be noted that a commercial laser line scanner is considerably less expensive than an OCT system and provides a complete point cloud of the part surface, which is advantageous for final quality assurance.

Furthermore, the closed-loop experiment showed that an effective control of the layer height based on real-time OCT measurements is possible. In this regard, the proposed control approach offered a significant improvement over the open-loop process where the layer height diverged and there was no compensation of the disturbance. The control system exhibited good performance with regard to disturbance rejection and setpoint tracking. Since the layer height in the closed-loop process always converged to the setpoint, instability due to an incorrect working distance could be ruled out. The comparatively simple yet highly effective control approach avoided the use of more complex control algorithms such as ILC. The performance was comparable to the control system from Ref. 18, although different control algorithms were used. An even faster convergence to the setpoint was impractical due to the necessary manipulated variable constraints and would have resulted in oscillations around the setpoint for small control errors. The main advantage of the presented approach was that no additional measurement after the deposition of individual layers was required. Thus, this work represents a significant contribution to the advancement of wire-based LMD for industrial applications.

In this work, a real-time control system for the layer height in laser metal deposition with coaxial wire feeding was elaborated and thoroughly investigated. Accurate height measurements during the deposition process were enabled by an in-axis distance sensor employing the measurement principle of optical coherence tomography. After implementing the height monitoring system, its capability for measurements in the vicinity of the melt pool was examined. It was shown that the closest feasible lateral distance to the tool center point yielded the highest signal quality. Furthermore, the steady-state relationship between the wire feed rate and the measured layer height was determined. A dedicated closed-loop control system was designed, incorporating a linear reduction in the control action around the setpoint to avoid overshoot. The control system was validated for a significant height disturbance in the substrate plate, which, in contrast to the open-loop case, was compensated within just a few layers. The validation experiments also showed that reliable real-time height measurements during a multilayer buildup are feasible.

In future work, the closed-loop buildup of arbitrary geometries and solid multitrack components should be investigated. For this, an inline positioning of the measurement spot according to the respective deposition trajectory is to be realized. Inline adjustments of the measurement spot position also offer the potential to detect the height both in front of and behind the melt pool, yielding additional information on the deposition process. Moreover, the control system facilitates the deposition of layers with uneven heights, enabling the use of nonparallel slicing techniques for the buildup of overhangs. Finally, the height control system should be combined with a thermal control system, as introduced in Ref. 24, to counteract heat accumulation in higher layers.

The results presented in this paper were achieved within the AdDEDValue project, which is supported by the German Federal Ministry for Economic Affairs and Climate Action (BMWK) within the funding program “Digitalization of Vehicle Manufacturers and the Supplier Industry” (Contract No. 13IK002L) and supervised by the VDI Technology Center (VDI TZ). We would like to thank the BMWK and the VDI TZ for their support and for the effective and trusting cooperation. Furthermore, we would like to express our gratitude to our partner Precitec GmbH & Co. KG for providing the coaxial wire deposition head and the OCT system. Finally, we would like to thank Robert Gropp and Rainer Sollfrank for their valuable support in the implementation of the sensors and the hardware.

The authors have no conflicts to disclose.

Christian Bernauer: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (lead); Project administration (equal); Software (equal); Validation (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Sebastian Thiem: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (supporting); Software (lead); Validation (equal); Visualization (supporting); Writing – review & editing (supporting). Pawel Garkusha: Investigation (supporting); Writing – review & editing (supporting). Christian Geiger: Formal analysis (supporting); Writing – review & editing (equal). Michael F. Zaeh: Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (equal).