Approach for advanced working distance monitoring and control capability in laser metal deposition processing for additive manufacturing

Laser metal deposition (LMD) is an additive manufacturing (AM) process in which a material in form of powder or wire is molten by a laser beam and subsequently added to the underlaying substrate material (Fig. 1). This technology is well suited for adding coatings to workpieces, repair applications, as well as generating complete 3D additive manufacturing (AM) components.^1,2 As the requirement to keep the working distance in a valid range is highly important in LMD processing, this also applies for the high-speed variant: extreme highspeed laser deposition (EHLA)^3,4 but with stronger influence. Because of the modified configuration, the process has a higher sensitivity and an increased risk of distance drift.

A typical approach to detect or verify the working distance (tip offset) within LMD manufacturing is to perform offline distance measurements with devices such as gauges, tactile, or optical distance sensors. In recent research studies, various approaches applying contactless online measurement systems like laser spot or line scanners have been investigated for distance or surface detection.^5,6 Although offering the advantage of a contactless measurement, this technology requires a two-step sequence with a processing stop to verify the distance or surface. Furthermore, when the distance detection device is attached to the processing head of a LMD laser-cell, in the majority of cases, the measuring axis or position has an offset to the actual LMD processing zone, the tool center point (TCP). This results in an additional positioning step to determine the true distance to the surface.

To detect the working distance online, a CCD camera was installed to the LMD setup by Tuo et al.⁷ The camera was mounted laterally to the processed sample. This setup is well suited for wall structures but could result in limitations on 3D components due to the direction dependency of the fixed camera position. A sound-based distance monitoring system offering contactless online measurements for LMD processing has been investigated by Chabot et al.⁸ The achieved accuracy of this approach was found to be in the range of 0.5 mm, which is lower than that of the optical systems described before.

In the presented approach, an OCT measurement system with online measurement capability during processing has been investigated. Originally employed in biomedical imaging,^9,10 sensors based on optical coherence tomography¹¹ are increasingly used in various fields of applications such as materials research and dimensional metrology.¹² Combining spectral, confocal and coherence gating, OCT is particularly robust against external influences such as process radiation allowing its use in laser applications with harsh process conditions such as keyhole monitoring in laser depth welding.¹³ More recently, OCT was employed to monitor powder and wire based LMD^14,15 with the OCT sensor measuring point located coaxial to the laser beam axis. In comparison in this presented novel approach, the distance is detected in a path surrounding the processing zone. Rotating around the TCP allows collecting 3D surface data while traveling along the processing path compared to only a single distance point when measuring in the TCP.

To achieve the goal of an integrated LMD processing head, a laser optics for material processing was fitted with an inhouse developed OCT measurement system and LMD equipment (Fig. 2).

The basis of the developed integrated LMD processing head is a laser optics for material processing with a fixed focal length of 200 mm. The beam splitter of the optics had to be adapted to the wavelength of the measurement laser to allow transmission of the laser beam. The OCT coupling with the nutating mirror was mounted above the beam splitter and LMD equipment beneath the focusing lens.

The OCT system uses broadband radiation with a center wavelength of 835 nm, a 3-dB bandwidth of 47 nm, and 6.4 mW output power which is guided from a remote OCT unit to the processing optics via a single mode fiber. The collimated measurement beam is combined with the processing beam using a dichroic mirror and then focused using the same lens as the processing beam. Backscattered light from the surface is redirected back through the optical system into the OCT fiber, where it is brought to interference with the reference beam to obtain an absolute distance measurement within the measurement range of 20 mm.

For direction independent process monitoring and control, a nutating mirror was placed at the back focal plane of the focusing lens to scan the measurement beam along an elliptical track centered to the TCP in a telecentric configuration. By rotating around the processing axis, one section of the measurement track is ahead of the process and one section behind the process measuring the resulting topography. Additionally, the lateral displacement from the processing axis reduces disturbances caused by the process such as powder crossing the measurement beam path.

The dimensions of the half axes of the elliptic measurement track were 3.2 and 2.3 mm, respectively (Fig. 3). A rotary encoder was used to trigger distance measurements at 512 positions per revolution up to the maximum measurement frequency of 80 kHz limited by the OCT sensor.

For LMD processing, a coaxial powder nozzle with the media supply was mounted to the laser optics aligned to the processing laser axis. The powder nozzle interface includes the gas supply and the adjustment unit for positioning the powder nozzle. The processing laser beam, the powder gas jet, and the measurement laser beam are guided through the powder nozzle and exit at the nozzle tip toward the processing surface. The setup is displayed schematically in Fig. 4 as enlarged detail of Fig. 3.

The powder gas jet consists of powder particles and argon as feeding gas, which exit the powder nozzle from a ring-shaped channel forming a hollow cone.

The developed and investigated LMD processing head with the distance measurement feature was configurated for metal powder as an additive material. The experimental LMD setup for testing, validating, and processing of test samples consisted of the following main components (Fig. 5):

• 6-axis robotic system,

• 3.5 kW Diode laser system with a 1000 μm fiber,

• Laser material processing optics with a focal length of 200 mm,

• Disc powder feeder Oerlikon Twin 150,

• ILT D50 coaxial powder nozzle.

As laser radiation and metal powder are involved during LMD processing, the influence of both aspects to the measurement system have to be analyzed in respect to reliability and robustness.

All experiments were performed with Inconel 718 (IN718) powder with a specified powder particle range of 45–75 μm. IN718 is a Ni-based superalloy applied in aero-space and turbo machinery components.¹⁶ 

In the first phase of the experimental program, the processing head with the integrated OCT measurement system was built up, mounted, and validated. This included analyzing the functionality and determining the precision of the measurement system.

Trials were conducted with the LMD process active to determine the performance of the processing head under working conditions. Because of the powder particles passing through the measurement beam and the emissions from the active LMD process, there is potential risk of influences disturbing the measurement signal. First, trials were executed at different powder flow rates without laser power to determine the influence of the powder particles to the measurement result. Next, trials were performed with powder and laser beam active to investigate the overall behavior of the measurement signal.

The second phase consisted of two applications to verify the functionality and capabilities of the LMD processing head with its integrated measurement features:

• distance control

• surface detection

In a final step, the limits of the OCT measurement system were investigated.

Based on the experimental scope, the investigation results cover the development and testing of the integrated LMD processing head and the OCT sensor itself as well as the performance in LMD applications.

By activating the laser beam and the powder gas jet, two influence sources are added that need to be analyzed to prove sensitivity and robustness of the distance measurement system. In Fig. 6, the relevant interaction zones are marked by the red circles.

With the selected setup, the powder particles exit the powder nozzle, cross the measurement beam (Fig. 6, upper interaction zone), and pass on to the LMD processing zone (TCP). At the TCP, they interact with the laser beam (Fig. 6, lower interaction zone).

In a first step, trials have been performed with the powder gas jet only (no laser radiation) to investigate the powder influence. Different powder mass flow rates were covered ranging from 2.5 up to 100 g/min to record dependency.

Even at the highest rate of 100 g/min, the resulting noise level was low. According to Fig. 7, an overall measurement deviation of about 15 μm was recorded with the powder flow on. The graph represents the measured distance with the measurement beam rotating around the processing axis scanning the surface. The sinusoidal shape of the signal is caused by the elliptical scanning pattern and residual misalignments resulting in a periodic optical path difference along the measurement track (Fig. 3).

The determined accuracy level of about 15 μm represents an excellent value since a deviation at this level has no significant influence on the LMD process.

To verify the ability to measure the applied track height during the process, measurements were recorded while single tracks were applied to a flat 1.4301 sheet substrate. During LMD processing trials, the applied laser power ranged from 250 to 500 W and the powder flow rates were in the range of 3–6 g/min. Resulting data are shown in Fig. 8 where the Y Axis was reversed for visual correspondence of the measured distances to surface height.

In the beginning of the measurement, the zero position of the measurement range was placed approximately 1.3 mm above the nominal working distance with the robot at rest, periodically measuring the same part of the substrate with a standard deviation of 6.8 μm between repeated scans. The enlarged section shows the sinusoidal error introduced by the scan which was reduced to 30 μm by improved alignment. Once the LMD process starts, one section of the measurement track measures the flat substrate surface in front and the backwards facing part of the scan measures the deposited track. The larger track height at the start of the process results from the robot accelerating from zero to the programmed feed rate resulting in locally higher material deposition. Since there was no distance control applied, the measured distance decreases reflecting the tilted substrate surface. At the end of the track, the processing head moves away from the surface.

To obtain the surface height, the sinusoidal error introduced during scanning across one scan revolution was subtracted to level out deviations when calculating the working distance. Figure 9 shows the resulting surface height for a single elliptical scan. The applied track with a height of approximately 240 μm can be seen at 180°, which represents the part of the scan opposite to the processing direction. On the slopes, there are missing data points due to the limited numerical aperture of the setup.

The concept of the distance control application was to take a sample with an unknown geometry and adapt the working distance from the nozzle tip to the surface using the OCT sensor signal. To validate the distance control, the demonstrator sample M-sheet in Fig. 10 was designed and manufactured from 1.5 mm stainless steel metal sheets (1.4301) with a size of 120 by 130 mm.

The defined task was to move the processing head over the demonstrator sample with only a straight line programmed on the robot (dashed horizontal line, Fig. 11), by comparing the measured working distance at a fixed measurement position detected by the OCT system with the target working distance. The calculated distance error was transferred to an analog signal which was fed to the robot’s I/O interface. To correct the distance error, the axis compensation functionality of the robot was activated. With the distance control activated, the robot adapts the working distance according to the signal from the OCT system (Fig. 11).

For validation, two experiments were performed:

• Working distance measurement with a dial gauge installed,

• LMD processing with active distance control.

To measure the accuracy of the distance control, a dial gauge was attached to the processing head. The robot program was executed with the distance control activated but without processing laser beam and powder gas jet (Fig. 12). The recorded error from the nominal tip offset of 12.5 mm was below 0.1 mm which is less than 1% of the working distance.

As the final validation, LMD tracks were applied to the surface of the M-sheet demonstrator sample. The sequence of images from the LMD deposition process in Fig. 13 documents the successful operation of distance control. The powder nozzle was kept at a constant working distance during processing with no visible disturbance due to LMD process emissions.

As a test application, an LMD contour was deposited to a 1.4301 substrate sheet. To achieve a higher measurement, point density of the size of the scan track was reduced to elliptic half axes of 1.4 and 1.0 mm. The robot movement consisted of a starting line followed by a 360° circular track with a diameter of 25 mm. Material was only deposited on the circular track. The LMD sample contour with the measurements scan tracks of the OCT sensor are visualized in Fig. 14. The program was executed twice depositing two tracks on top of each other.

Robot position and movement direction were acquired and synchronized with the OCT data to acquire a time resolved 3d point cloud. Data processing was performed offline.

Using the direction vector of the robot, acquired OCT data were dynamically split into the measurement segment ahead of the process and the segment that is located behind the process which measures the resulting topography of the deposited track. For each data segment, surface elements were locally fitted to the resulting 3d point cloud to obtain a surface model both in forward and backward directions.

Figure 15 shows both the LMD contour and the calculated surface model, which was calculated based on the data from depositing the second material layer. The first deposited layer is represented by the forward-looking data segment. The surface of the second layer was calculated from the backward-looking data segment with both layers applied. Local defects resulted in noncontinuity of fitted surface elements which are easily identified (marked circles, Fig. 15), demonstrating the possibility of error detection during the LMD process.

As the measurement is based on the light reflected or scattered from the surface back into the OCT fiber, the signal strength is dependent on the surface structure and angle of incidence. When increasing the angle of incidence from the perpendicular position, the signal quality will decrease until a measurement is no longer possible. The experimental setup in Fig. 16 was applied to determine the influence of the angle of incidence to the measurement system. To determine the limit of the distance control, a surface scan was started on the zero position in Fig. 16 and moved in the Y direction. The control guided the nozzle tip with a constant distance to the reference cylinder surface until a distance detection was not possible anymore. The determined angle α represents the maximum possible angle offset for valid data and the operating angle range of the distance sensor.

This test sequence was performed on a sandblasted surface and a surface machined by turning.

Figure 17 displays the gradually decreasing proportion of valid data points with angle of incidence on the sandblasted surface. 90% valid data points were achieved up to an incidence angle of 8° with a drop to 10% valid data points at approximately 16°. While continuous information is not acquired at steep slopes with high data dropout rates, some applications such as distance control require data on the timescale of only 100 Hz or less. At a measurement frequency of 80 kHz, even 2% valid data points may still provide sufficient information.

With active distance control, the angle of incidence limit for a sandblasted surface was determined at approx. 25° and for the turned surface at 20°. The sandblasted surface generates a diffused reflection which results in an increased detection range.

The presented approach of developing and verifying an enhanced LMD processing head with integrated distance detection ability was successfully demonstrated. The system performance was analyzed and examined in two LMD applications.

After designing and assembling, the integrated OCT system was validated. Investigations on the two main disturbance sources during LMD processing, the powder gas jet and the emissions from the processing zone proved stable and reliable results. With the powder flow rate of 100 g/min, a distance deviation of 5 μm was measured which has a negligible influence to the LMD process. LMD trails with test tracks resulted in a standard deviation of approximately 7 μm with a laser power between 250 and 500 W and powder flow rates at 3–6 g/min. The achieved accuracy of less than 20 μm with an active LMD process can be regarded as uncritical for the majority of LMD applications.

In a first demo application, a distance control loop was integrated and the functionality of the distance control system with of the OCT sensor could be demonstrated. A simplified setup that consisted of a prismatic shaped demo-geometry with a single LMD track was selected as a 3D data analysis software module was not available. Therefore, a distance measurement at a fixed position (1D data) by the OCT system was sufficient. The calculated distance signal was transferred to the robot to compensate the distance offset. Trials with a dial gauge on a freeform test sample resulted in a distance error of less than 0.1 mm. The presented results prove that the developed LMD processing head with the integrated OCT measurement system could successfully detect and control the working distance.

For LMD processing of complex 3D geometries, a 3D distance analysis functionality is needed but presently is not available. Future steps would be the development of a real-time 3D software module to extract the relevant distance information from the 3D scan data to enable the system to handle complex 3D LMD components.

Furthermore, the presented OCT LMD processing head has the potential to be extended to a multifunctional tool as it has the capability to simultaneously detect the distance in front of and behind the processing zone while scanning the surface. This capability was successfully demonstrated in the second test applications. Additionally, when combining the machine data from the robot with the scanned distances, a 3D dataset representing the surface of the built component can be generated showing potential for advanced process monitoring and control, process automation, and quality assurance in future applications.

The research leading to these results has received funding from the Federal Ministry for Economic Affairs and Energy (Zukunft Innovation Mittelstand ZIM, Grant Nos. ZF4328109FH9 and ZF4231002FH9).

The authors have no conflicts to disclose.

Jochen Kittel: Conceptualization (lead); Investigation (lead); Methodology (lead); Resources (lead); Writing – original draft (lead). Fabian Wendt: Investigation (equal); Methodology (equal); Resources (equal); Software (lead); Validation (equal); Writing – review & editing (equal). Stefan Hoelters: Investigation (supporting); Software (supporting); Supervision (supporting). Andres Gasser: Conceptualization (supporting). Matthias Hackel: Conceptualization (supporting); Methodology (supporting); Resources (equal); Software (equal); Writing – review & editing (supporting).