Single Crystalline Laser Welding

Abstract

Rising demands on engines in the aerospace industry require high engine inlet temperatures in modern gas turbines in order to increase the efficiency. Single-crystal nickel-based superalloys were developed in order to meet these rising demands and confer high pressure turbine blades with the necessary wear and oxidation resistance at high temperatures. The blades of the high-pressure turbine are subject to significant wear due to the high mechanical strain under extreme conditions, which appears in the form of cracks in the single-crystal substrate material. There are no approaches for the restoration of the original material properties since the repair of cracks and erosions by polycrystalline laser cladding can be applied only to a limited extent. The aim of the subproject is the restoration of defect, single-crystal high-pressure turbine blades. This objective was achieved by developing a novel two-stage laser metal deposition process based on simulation, laser process development, process monitoring and temperature control.

1 Introduction

In order to reduce fuel costs to the airline and comply with environmental legislation on emissions, the optimization of the thermal and propulsive efficiencies of the modern aeroengine has become paramount. Even a small increase in efficiency and service interval of these engines relates to substantial savings in fuel and maintenance costs over the life of the engine (McNutt 2015).

The development of directionally solidified (DS) castings allowed for the elimination of grain boundaries transverse to the loading axis, while single-crystal (SX) investment castings eliminated grain boundaries almost completely (Donachie and Donachie 2002). This allowed for a huge improvement in creep resistance, since grain boundaries are sites of damage accumulation and crack initiation at high temperatures (Schneibel 2004).

Nickel-based superalloys consist primarily of a two-phase microstructure: A face-centered cubic (FCC) gamma solid solution matrix phase and a uniform distribution of ordered gamma prime precipitates. Advanced single-crystal superalloys contain high volume fractions of gamma prime, typically in the range of 50 to 70 volume percent. The single-crystal nature combined with the high volume fraction of gamma prime precipitates, provide superior high-temperature mechanical properties (Vitek et al. 2002).

High-pressure turbine blades are located in the hot section of aircraft engines and convert gaseous energy exiting the combustion chamber into mechanical energy. These blades are subject to high temperatures, high stress, vibration effects, centrifugal and fluid forces that can result in creep, fracture or yielding fractures (Boyce 2006).

Components in aircraft engines are exposed to extreme conditions. Due to high temperatures and pressure as well as the impact of foreign objects, defects such as plastic deformations, wear by hot gas and CMAS (Ca-Mg–Al-Si) corrosion, impact damages and cracks can occur, resulting in high costs (Alfred et al. 2018).

Due to the high cost of manufacture, repair and replacement, there is a clear financial incentive for more economical manufacturing routes. These result in surface defects as a result of erosion and corrosion, wear of the turbine blade tip and cracking in the base material.

The process of laser metal deposition (LMD) has established itself as a robust process for the repair of parts. Its flexibility with regard to materials used, near net shape of the deposited structures and control of thermal processes make it interesting for specialized applications. Microstructurally, the rapid solidification rates in the melt pool during the LMD process result in small grains (McNutt 2015) with less segregation of alloying elements when compared to casting and conventional arc welding processes. The process also allows for a high degree of control of the heat flow, thereby allowing control over the microstructure (Gäumann et al. 2001).

The Fig. 1 shows the γ and γ’ phases of typical Nickel-based superalloys (a). Figure 1b shows the fine microstructure obtained during the deposition of such alloys due to high solidification rates. Using an electron backscatter diffraction (EBSD) analysis of the deposited clad, the absence of grains and high-angle grain boundaries are detected, which is characteristic of single crystal material.

Fig. 1

a two-phase microstructure of CMSX-4, b micrograph of deposited material

1.1 Material Challenges

While the high performance capabilities of such components result in longer life spans, the repair of these parts still poses a challenge (Rottwinkel et al. 2014a, b). The processing of these Nickel-based Superalloys is challenging due to its low weldability and high susceptibility to cracking (Buchbender et al. 2020), which arises from its complex chemical structure. Many Ni-based superalloys are considered to have low weldability due to the rapid precipitation of the strengthening phase gamma prime (Wahl and Harris 2011). Further- more, an alloy is considered non-weldable if the total Al-Ti content exceeds 4 wt.% (De Luca et al. 2021).

The high aluminum and titanium content of the alloy results in a high gamma fraction, which improves performances at high temperatures, but makes it exceedingly difficult to repair via conventional fusion methods such as welding (McNutt 2015).

When filler metals with high volume fractions of gamma prime are used, weld cracking is common. They are characterized by an abundance of stray grains, which are new grains that are formed during solidification. The high-angle grain boundaries that exist in con- junction with the stray grains are weak links in the microstructure and act as preferred crack propagation paths. Therefore, existing weld technologies must use inferior filter metals with low gamma prime volume fractions (or no gamma prime) in order to produce crack-free welds (Vitek et al. 2002).

In summary, CMSX-4 is characterized as difficult to weld, resulting in material that is highly susceptible to cracking during processing. In addition, to columnar dendritic structure of these parts poses a challenge with respect to reproducing the orientation during additive repair. Balancing these requirements with contour fidelity is difficult, as the parameters that result in high single crystallinity and low cracks, do not necessarily guarantee an optimal shape (Buchbender et al. 2020).

1.2 Scope of the Sub-Project–Single Crystalline Laser Welding

In single crystalline laser welding the heat distribution has a significant impact on the formation of the micro structure. Therefore, the simulation of heat distribution during the laser metal deposition, the influence of an active substrate cooling and preheating on the heat distribution and temperature gradient in the substrate material were carried out. Building on this, a process for the repair of superficial defects using the process of remelting was developed. This allows the regeneration of cracks and defects in the substrate material located parallel and lateral to primary dendrite orientation.

Followed by the deposition of multi-layered structures, single-crystal structures by means of the two-step repair process of deposition and remelting were realized.

This involved the analysis of the influencing process factors on the properties of the deposited material as well as the quantification of the effect of the remelting parameters on the resulting microstructure.

In order to ensure a robust and reproducible process, thermal and acoustic process monitoring methods were implemented. Since the laser beam is the primary source of heat input in the process, the melt pool temperature during deposition and remelting was monitored. Empirically determined optimal temperatures for deposition and remelting were determined for a particular geometry in order to maximize single-crystal height.

The susceptibility of Nickel-based superalloys to cracking as well as the lack of information about crack formation during the process limits the application of repair process by laser metal deposition (Weingärtner et al. 2015).

The analysis of structure-borne emissions was implemented for the detection of cracking during the turbine blade repair process.

With the results from this sub-project B5 it was possible to develop a repair process under conditions determined by industrial applications. The transfer project T5 highlights these successes. Additionally, the turbine blade repair process was then implemented on an industrial machine in the scope of the automated process chain presented in system demonstrator.

2 Materials and Methods

The research was carried out on a 6-axis laser metal deposition machine. In order to melt the material a 680 W laser was used. This diode LDF 400–650 laser by Laserline GmbH emits laser light with wavelengths of 940 and 980 µm. The powder material is conveyed to the process zone using inert argon gas. An additively manufactured processing head with an annular powder deposition ring is used to generate the material deposition. The powder material was deposited coaxially with a powder focus diameter of 2 mm. The CMSX-4 powder particle diameter ranges between 25 and 75 µm. The substrates consist of similar composition to the powder material as shown in the Table 1. Additionally, the angle of primary dendrite orientation is these substrate is specified.

Table 1 Chemical composition of CMSX-4 powder and substrate material in wt.% (Buchbender et al. 2020)

2.1 Process Development

The conditions for the transformation from liquid to solid during solidification processing, such as the temperature gradient and the growth rate, vary from process to process as a function of time and space (Gäumann et al. 2001). This in turn determines the resulting microstructure in the deposited material. By varying the laser power, powder volume flow and the process speed in statistical designed experiments, the laser metal deposition process can be tuned to change the heat distribution and the thermal gradient during the solidification. The foundations for these experiments are based on numerical simulations.

2.2 Simulation

The simulation was carried out in order to analyze the temperature distribution during laser processing of CMSX-4. Ansys 12.1 was used to solve the finite element models in a 10-s transient model (Kirbach 2011).

In comparison to conventional materials like structural steel, the thermal conductivity of the nickel super alloy is lower by a factor of three. In the analyzed process duration of 10 s, the entire volume of the structural steel substrate heats up. In contrast, only 50% of the CMSX-4 substrate is heated up in the same process time. The simulations show that the isotherms in CMSX-4 are aligned in an exponential direction, while, for reference, the isotherms in structural steel are aligned approximately parallel to the components top edge. Thus, in CMSX-4 material a temperature gradient exists in several spatial directions inside the substrate (see Fig. 2). The main one-dimensional temperature gradient through the workpiece is already dictated by the thin and wide geometry of the substrate. The use preheating for the welding process is required. Otherwise, cracks are very likely to occur. With the preheating temperature of 300 °C, the temperature gradient for single- crystal solidification cannot be guaranteed for the process duration. This conflicts with necessary temperature for stress relief and crack prevention. Using a hot enough temperature field prevents cracks, but the temperature gradient needed for single-crystal solidification would no longer exist.

Fig. 2

Thermal simulation of the laser metal deposition of structural steel (left) and CMSX-4 (right)

With an added constant cooling at the bottom edge of the substrate, the CMSX-4 substrate heats up slower and a cold area is formed from below the welding, which provides homogeneous heat dissipation.

Overall, the simulation of preheating of the substrate indicated a loss of the required temperature gradient for the formation of epitaxial structures. This is in line with the recommendation by Gäumann et al. (2001) stating that the temperature of the substrate should be as low as possible.

3 Results

During the project, results have been achieved regarding the deposition of the CMSX-4 material. Material properties were investigated regarding multi-layered deposition and lateral repair. A key aspect was the integrated process monitoring and control.

3.1 Deposition of Single and Multi-Layered Structures

The resulting microstructure of the clad depends on the solidification rate and temperature gradient. The measured temperature gradient during laser metal deposition ranges between 10⁵ and 10⁷ K/m and the solidification rate is between 10^–3 and 10^–1 m/s, as shown in the Fig. 3a.

Fig. 3

a microstructure as a result of solidification rate and temperature gradient (modified from Burbaum 2011), b percentage single-crystallinity (%-SX) as a result of varying significant factors during the deposition process

These thermomechanical rates are in turn primarily determined by the laser power (W), the laser speed (mm/s) and powder feedrate (g/min) as well as the interactions of these factors amongst each other. In order to deposit structures with the desired microstructure, an analysis of the effect of the process parameters on single-crystallinity is required. Within the design of experiments, the following range of process parameters was analyzed. The laser power to create the melt pool was varied between 130 and 265 W. The laser speed ranged from 40 to 165 mm/min. CMSX-4 material has been fed with a powder feedrate from 0.8 to 2.15 g/min. The resulting interactions are dis- played in the Fig. 3b. These results indicate that a combination of high power, low laser speed and high powder feedrate result in the highest percentage single-crystallinity. Since literature (Gäumann et al. 2001) and previous work (Buchbender et al. 2020) indicate that the use of high laser power leads to the loss of the temperature gradient, by analyzing the incidence of cracks, microcracks and misorientations, a threshold for these factors can be set. If the parameters are set below this threshold, the microstructural results are desirable.

These findings formed the basis for the deposition of multi-layered structures. In order to further evaluate the effect of parameters during process development, geometrical and microstructural features of the deposited clad were chosen as analysis criteria.

The focus is set on the percentage of single-crystallinity by dividing the single-crystal area by total surface area. Furthermore, the height difference is calculated by subtracting actual height of the weld form the desired metal deposition height. With the help of computer tomography and micro sections the number of cracks, microcracks and misorientations were counted and evaluated in the statistical model.

A key finding during the research is that an additional remelting step involving a single pass of the laser beam allowed for the recrystallization of previously misoriented material within a track. This has been shown by Rottwinkel et al. (2014a, b). With this laser remelting strategy for track single-crystallinity extension has been proven a tool capable of improving and simplifying the formation of large SX-volumes by laser powder cladding. It was expected that upon remelting, a resolidification of the misoriented volume would occur, allowing for the extension of the overall single crystalline height (Rottwinkel et al. 2017). A study of builds with and without an additional remelting pass between layers showed that the former resulted in higher %-SX, fewer microcracks and higher deposition height. The largest differences between the strategies of remelting (R) and without remelting (X) were seen in the incidence of microcracks and percentage single crystallinity (%-SX) shown in Fig. 4. Deposits built with the remelting strategy showed fewer microcracks and higher percentage single-crystallinity, which are amongst the most influential factors for epitaxial height extension. Hence, the remelting strategy was chosen for further experiments (Buchbender et al. 2020).

Fig. 4

Effect of remelting strategy (R) and no remelting strategy (X) on microcrack formation and %-SX (Buchbender et al. 2020)

Using optimized parameters and remelting, a study of deposition strategies showed over- all improvement across all five analysis criteria. Percentage single-crystallinity of above 80% was achieved for all parts, with the highest being 97.9% measured by EBSD. Several builds showed zero microcracks, cracks and misorientations with a deposit height of 4480 µm can be seen in Fig. 5 (Buchbender et al. 2020).

Fig. 5

Selected cross-sectional micrographs of builds using optimized parameters and strategies (Buchbender et al. 2020)

3.2 Lateral Repair

With the creation of multi-layered structures, the repair of high pressure turbine blade tips can be targeted. For other repair situations, a lateral repair is needed. This is done by transferring the optimized parameters and strategies to substrates prepared for lateral repair. In addition to the changed thermal conditions of the meltpool, the dendrite orientation in the deposited material is also a function of the altered dendritic orientation in the substrate.

This is especially relevant for the repair of defects located lateral to the direction of primary dendrite orientation. Here, the angle of processing determines the direction of heat flow and therewith the orientation of the microstructure. As soon as the first dendritic solidification starts after the laser process, the solidifying structure takes over the orientation of the crystal structure originated from the base material. This means, that the structure grows epitaxially on the partially melted dendrites inside the base material. If the growth direction equals direction of travel of the laser beam, the orientation matches the primary dendrites. For a predefined direction of travel of the laser beam and a known position and orientation of the primary dendrite stems, the growth direction of the dendrites can be calculated (Burbaum 2011). With this knowledge and in order to develop a repair strategy for lateral defects, it is required to determine the optimal processing angle that results in the highest percentage single-crystallinity. The results of the effect of processing angle on percentage single-crystallinity and the percentage of secondary dendrites are shown in Fig. 6.

Fig. 6

a processing angle and b percentage primary and secondary dendrite orientation as a result of processing angle

These results show that at a processing angle of 20° the %-SX gradually declines, while the percentage of secondary dendrite orientation increases. This leads to the decision to advance with a processing angle of 20° for further experiments. In order to be able to process parts along the entire length of its lateral edge, a modification of the processing nozzle was carried out, since the geometry of the current processing head did not allow for flexible processing. The modification ensured the working distance of 9 mm was maintained and allowed the processing of larger parts.

3.3 Process Monitoring

Process monitoring can be employed to increase reproducibility and monitor part quality. Monitoring methods for laser processing are based on the physical phenomena that occur due to laser-material interactions. These can be acoustical, optical, electrical or thermal (Purtonen et al. 2014).

Increasing meltpool temperature and substrate temperature in the higher layers results in process irregularities, the formation of misorientations and cracks as well as the loss of the required temperature gradient for epitaxial deposition. Hence, the applicability of thermal process monitoring in the deposition of epitaxial structures was studied.

A challenge in the current state of the art of process monitoring is the inability of visual and thermal methods to detect cracking within the material during processing. Especially for materials, such as the investigated Nickel-based superalloys, that are prone to cracking and have narrow processing windows. This inhibits the industrial applicability of the process. Challenges in acoustic process monitoring are the inadequate signal to noise ratios and therefore a very high susceptibility to noise, as well as limited literature especially with regard to higher frequency bands. Within the project duration, the uses of structure-borne sound to identify the acoustic signatures of cracks were investigated. This was done by equipping the laser metal deposition process of the difficult to weld Nickel-based superalloy CMSX-4 by applying a time–frequency analysis method.

Thermal process monitoring

Pyrometry allows the measurement of the spectral intensity of a surface within a particular spectral range, which is correlated with a temperature based on the emissivity ε of the material. The applied two-color pyrometer or ratio pyrometer is a type of pyrometer that measures the temperature in two spectral ranges simultaneously and calculates the temperature by converting the ratio of spectral irradiance between the two measured wavelengths. This two-color pyrometers is used to minimize the wavelength-dependent influence of the emissivity of the measuring surface, provided the emissivity changes proportionally for both wavelengths. The schematic setup is shown in Fig. 7a.

Fig. 7

a schematic diagram of coaxial thermal process monitoring setup and b the change in penetration depth as a result of change in laser power during remelting

It was hypothesized that an empirically determined optimal meltpool temperature range exists for deposition and remelting at a constant deposition speed and powder feed rate. This allows for the maximization of single-crystallinity in the deposited material.

In order to determine the optimal meltpool temperature range, single clads were deposited onto single-crystal substrates. Keeping the powder feed rate and laser speed constant allowed for the quantification of the influence of the laser power on the microstructure. The meltpool temperature was measured with varying laser power across the length of each track for deposition and remelting respectively. Longitudinal micrographs of the clads were prepared and the ratio of SX (single crystallinity) to PX (polycrystallinity) as well as the remelting depth measured.

As shown in the previous section, the choice of remelting parameters influences the %- SX (percentage single-crystallinity) significantly.

The Fig. 7b shows the change in penetration depth during remelting as a result of a change in the laser power. Remelting power of 84 W resulted in a remelting depth of 244 µm, while a power of 140 W resulted in a remelting depth of 375 µm (Honisch 2018). This manifests an adequate remelting depth to allow complete recrystallization of previously misoriented material.

In this setup, the design of experiments concludes an optimal temperature for deposition at 1550 ± 50 °C with a laser speed of 75 mm/min and powder feed rate of 1 g/min. These parameters were determined based on the highest single-crystallinity in longitudinal micrographs of deposited clads.

The optimal temperature for the remelting process lies at 1350 ± 50 °C for a laser speed of 60 mm/min, which was determined, based on the penetration depth required to recrystallize the deposited material.

In order to achieve these temperatures during the deposition and remelting process a custom process control was implemented. By modulating the laser power, the targeted temperatures were controlled (see Fig. 8).

Fig. 8

Thermal process monitoring and control for the deposition and remelting process

Process monitoring by acoustic emission

The analysis of acoustic emissions (AE) have been used in structural health monitoring, to detect phase transformations and cracking. These acoustic emissions are elastic waves or stress waves that propagate through materials in various modes, which are characterized by the vibrations of the particles. In solids, this propagation occurs in the form of longitudinal and transverse waves. Longitudinal waves travel in the direction of sound transmission, causing local regions of compression. Transverse waves, also called shear waves, oscillate perpendicular to the direction of wave motion. The type of mode and propagation depends on the nature and geometry of the medium and are affected by numerous factors and material properties.

In laser material processing, the detection of acoustic emissions has been carried out by monitoring the optical components of the laser welding setup, the work piece or air borne sound off-axially. AE sensors can therefore be structure-borne sensors or airborne acoustic sensors. In order to analyze the emissions detected, a signal processing method must be defined to obtain useful information. Baccar (2015) mentions three approaches for the analysis of AE signals, of which the following two are most relevant to the analysis carried out during the collaborative research center.

First, a parameter-based analysis is investigated, which consists of the extraction of time- based parameters such as events, signal amplitude, rise time, count, duration and energy. Secondly, a time–frequency analysis that analyses features in the frequency domain such as power spectrum, peak frequency and dominant frequency band.

With parameter-based analysis, on one hand, the amplitude of a signal can be correlated with the severity of the occurrence. However, amplitude is affected by attenuation and hence varies with proximity to the sensor. Therefore, amplitude or hit counts based on the amplitude do not convey sufficient information about the source of the acoustic emissions. Time–frequency analyses on the other hand offer more insight. Common methods used for time–frequency analysis are amongst others the Short Time Fourier Transform (STFT), Continuous Wavelet Transform (CWT) and Hilbert Huang Transform (HHT). This study makes use of the STFT method, as is depicted in Fig. 9. In the STFT method of time–frequency analysis, the signal x(t) is multiplied by a window function g(t-τ), the FFT of the windowed signal is calculated, while the sliding window moves along the time axis, calculating the FFT for each time interval. Common window functions are the Gaussian, Hamming and Hanning windows. The width of the window has a major influence on the resolution of the resulting signal. Large windows ensure high frequency resolution and low time resolution, while small windows provide lower frequency resolution and high time resolution. The choice of window width therefore requires prior knowledge of the process events, as the window size should not exceed the time period of the events to be detected in the signal. STFT allows for the representation of the frequency information of the signal over time, upon which a frequency range can be selected in which the signal to noise ratio is large enough to detect anomalies.

Fig. 9

a process set up and signal processing chain, b time–frequency analysis of acoustic signal

For data acquisition, the commercially available piezoelectric sensor and preamplifier by QASS GmbH were used. Experiments indicated that attaching the piezoelectric sensor to the substrate mount led to the most reproducible results and favorable signal to noise ratio. All measurements were carried out with a sampling frequency of 390 kHz.

The time–frequency analysis was carried out to identify the acoustic characteristics of the laser metal deposition process (Fig. 9b).

This showed that the status of the powder (on/off) can be identified across all frequency bands. Laser on/off status can be identified by higher amplitudes in characteristic frequency bands, in this case around 25 and 100 kHz. Low frequency bands around 10 kHz showed low signal-to-noise ratios. Continuous emissions of low amplitude indicated electrical noise (as seen at 100 kHz). Furthermore, it was observed that amplitude is highly dependent on the geometry of the part, the position of the sensor and the proximity of the process zone to the sensor.

In order to detect cracking during the process of blade repair, the anomalies of the acoustic signal were further analyzed. Formation and propagation of cracks inside the material can be correlated with anomalies. The cracks are characterized by short, broad-banded, high frequency transient signals in the time domain as well as in the frequency domain (Fig. 10a). With the use of signal processing to visualize high amplitude, transient signals in the frequency domain are computed. By setting a threshold, anomalies are categorized. These classified cracks are counted in each layer (Fig. 10b) and then verified by computer tomography and micrographs.

Fig. 10

a characteristic signal in the time domain, b signal processing for the visualization of high amplitude transient signals

4 Conclusions

The rising demands on engines and efficiency in the aerospace motivated the repair of high pressure turbine blades. The single-crystal nickel-based superalloys presented challenges during the research due to crack formation and globulitic microstructure. To over- come these difficulties a novel two-step process for a single crystalline repair has been developed during the project duration (Kaierle et al. 2017; Buchbender et al. 2020). It consists of the material deposition followed by a remelting pass of the laser without additional material.

As a result, the single-crystalline repair of the three most important damage patterns is possible. With the help of process monitoring and pyrometric control the microstructure is ensured in all three damage patterns.

Despite the optimized process parameters, small cracks can occur during the laser metal deposition of the CMSX-4 material. By enhancing the process monitoring occurring cracks can be detected by means of structure-borne acoustic measurements while the welding process is carried out.

It has been shown, that Laser metal deposition is a suitable repair process for automated regeneration of the single crystalline turbine blades. In a collaborative effort with direct support of the subprojects A1–thermomechanical properties, B1–combination of brazing and cladding and B2–recontouring a high pressure turbine blade has been refurbished. Therefore, the complex capital goods can be repaired. Furthermore, the findings have been transferred in the transfer project T5, where a deposition of multi-layered structures with single-crystallinity of 98% has been achieved. The integration into the automated process chain of the system demonstrator has been realized. A repaired high pressure turbine blade, its microstructure and the final automated laser metal deposition process can be seen in Fig. 11.

Fig. 11

High pressure turbine blade with material deposition (left), the visible orientated microstructure (right) and the automated laser metal deposition process implemented in the system demonstrator