Dimensional and geometrical accuracy in generative laser cladding

Contact

Name

Reinhart Poprawe

Management

Phone

work
+49 241 80 40419

Email

E-Mail
 

In powder-feed Laser Metal Deposition (LMD), dimensional and shape accuracy as well as the resulting microstructure play a decisive role with regard to the quality and properties of the manufactured or repaired component. The formation of the melt pool, including the movement of the three-phase line along the solidification front, determines the geometry of the weld track and, in the case of overlap processing or multilayer welding, the resulting geometry after each additional layer. For an industrially suitable process technology in the field of generative processes for metals, the requirement for LMD is to build up a component according to CAD data within a dimensional and form accuracy of a few tenths of a millimeter allowance and to avoid undersizing. Values for dimensional and shape accuracy before the start of the SFB were > 0.3 mm. The process control required to build up special geometry features (edges, corners, overhangs) is complex and in many respects not yet fully understood. For example, at the beginning of the SFB, there was a lack of understanding of the interactions between powder particles and laser radiation, the formation and instabilities of the melt pool geometry, and the Marangoni convection in the melt. In addition, the formation of residual stresses and thermal distortion as well as the microstructure that forms as a function of solidification conditions were unclear.

Therefore, the aim was to develop a basic methodological procedure for the buildup of defined geometry elements and the setting of a defined microstructure, and to develop the necessary process control in advance with regard to the essential process strategies and parameters, so that only fine-tuning is required experimentally. This is why in phases 1 and 2, the influence-effect relationships between process control and dimensional/shape accuracy were investigated on selected geometry elements (e. g. tracks with varying width, edges, corners, radii), which represent the spectrum of geometry-relevant elements for the near-net-shape structure. Among other things, process diagnostics by means of high-speed videography and thermography and model-theoretical analysis and process simulation were used for this purpose.

In this way, the geometric and microstructural formation of the weld track or the generated volume was derived and the relationships with the process management determined. This process knowledge was systematically collected to serve as a basis for improved path planning with regard to shape and dimensional accuracy in the third phase. For example, the database contains complete documentation of the results obtained with regard to track geometry and layer height as a function of the process parameters. In addition, special cases such as single track with varying width but constant height are included. These can be used, for example, to produce pointed geometries without stair-case effect. For materials that are not yet recorded in the database, or laser beam intensity distributions or powder particle densities that show significant deviations from the process parameters stored or recommended in the database, a simulation tool has been developed that can be used to analyze the process computationally in advance. The investigations cover both the application case of a repair on already existing material and the complete build-up of an entire component.

By investigating and quantifying the dependencies of track height and width on the process parameters, tracks can now be manufactured with continuously varying width and constant height. This possibility opens up further degrees of freedom in path planning, which are used in particular for varying cross-section widths and pointed components.

 
  Contour plot for a) track width and b) track height as a function of laser power and feed rate, c) weld track with variable width and constant height Copyright: © SFB 1120 Figure 1: Contour plot for a) track width and b) track height as a function of laser power and feed rate, c) weld track with variable width and constant height
 
 

The process influence on geometry and residual stresses was investigated in the production of thin-walled structures by LMD. For example, residual stresses can be reduced by higher laser powers, but this results in more strongly varying wall thicknesses and thus deteriorates the geometric accuracy.

 
  left: Residual stresses as a function of laser power, right: Variation of track width as a function of laser power and powder mass flow rate Copyright: © SFB 1120 Figure 2: left: Residual stresses as a function of laser power, right: Variation of track width as a function of laser power and powder mass flow rate
 
 

The effects of the weld pattern, i.e. the sequence and arrangement of the weld tracks, on the resulting distortion for a single-layer weld on thin (2 mm thickness) substrates were investigated using in situ Digital Image Correlation (DIC). The widely varying distortions must be taken into account in web planning for components susceptible to distortion.

 
  Absolute distortion along measuring lines in x- and y-direction for single-layer welds a) spiraling inward, b) spiraling outward, c) meandering Copyright: © SFB 1120 Figure 3:Absolute distortion along measuring lines in x- and y-direction for single-layer welds a) spiraling inward, b) spiraling outward, c) meandering
 
 

When welding on free-form surfaces, the effects due to inclined surfaces (varying track height and width depending on inclination and welding direction) must be taken into account during path planning. For this purpose, the resulting surface geometries after n layers must be anticipated. The methods commonly used so far for surface prediction assume a translation of the surface in z-direction (opposite to gravity) or in surface normal direction (z- or n-offset method). For inclined surfaces, the real layer thickness is overestimated (z-offset method) or underestimated (n-offset method). By combining the two methods, the surface geometry can be predicted much more accurately and the deviations from the real geometry can be reduced by 70-80%.

 
  ACTUAL and TARGET profiles after eight layers in path planning based on a) z-offset method, b) n-offset method, c) average method. Copyright: © SFB 1120 Figure 4: ACTUAL and TARGET profiles after eight layers in path planning based on a) z-offset method, b) n-offset method, c) average method.
 
 

The melt pool model developed in phase 1 was validated using a broad range of parameters (welding on inclined surfaces and variation of process parameters). The measured three-dimensional particle distribution (from the nozzle exit to the processing plane) and the beam acoustics are used as input variables for the free boundary problem in laser cladding.

 
  Simulated temperature profile (left) and experimental measurement (microscop) (right). Relative good comparativeley: relative error < 10 % with regards to melt pool depth, width and height. Copyright: © SFB 1120 Figure 5: Simulated temperature profile (left) and experimental measurement (microscop) (right). Relative good comparativeley: relative error < 10 % with regards to melt pool depth, width and height.
 
 

A simple statistical interaction model is used to calculate the particle temperatures, the individual trajectories and the resulting shading of the laser radiation (by the particles). These quantities serve as additional boundary conditions for the melt pool model. In this model, under conservation of mass flow, the free melt pool surface as well as the transient temperature distribution is calculated. The former is relevant for predicting track geometries (influence in path planning) and the latter is used to determine the solidification conditions on the phase front. In general, the solidification conditions cannot be determined experimentally.

 
  Calculation of the solidification conditions during LMD. Copyright: © SFB 1120 Figure 6: Calculation of the solidification conditions during LMD.
 
 

From the solidification conditions, a non-equilibrium phase field simulation can be performed to determine the microstructure properties in advance. There are still minor discrepancies in the prediction of the dendrite arm spacing (measure for the fineness of the microstructure). However, the occurring phases are predicted correctly.

 
  Phase field simulation: Modeling of the solidification processes on µm scale. The fineness of the microstructure is measured via the primary dendrite arm spacing (PDAS) and scaled with the macroscopic, mechanical properties of the material. Copyright: © SFB 1120 Figure 7: Phase field simulation: Modeling of the solidification processes on µm scale. The fineness of the microstructure is measured via the primary dendrite arm spacing (PDAS) and scaled with the macroscopic, mechanical properties of the material.
 
 

The nickel-base superalloy Inconel 718, which is frequently used in LMD, requires a strength-increasing heat treatment after the manufacturing process. This must be adjusted to the initial microstructure but is currently optimized for cast or forged microstructures. The aim of the heat treatment is, on the one hand, to homogenize the matrix and release brittle phases such as the Laves phase and, on the other hand, to precipitate strength-increasing γ'' phase in a controlled way. The comparison of non-solution annealed samples (Direct Aging, DA) with solution annealed samples shows a smaller hardness in the DA samples. This can be explained by the presence of Laves phase precipitates and a smaller volume fraction of γ'' phase.

 
  a) time course of hardness of specimens with solution heat treatment (1h at 980 °C) after aging at 760 °C or 720 °C and specimens without solution heat treatment (direct aging, DA) after aging at 720 °C, b) SEM image of a specimen after aging for one hou Copyright: © SFB 1120
 
 

Figure 8 :a) time course of hardness of specimens with solution heat treatment (1h at 980 °C) after aging at 760 °C or 720 °C and specimens without solution heat treatment (direct aging, DA) after aging at 720 °C, b) SEM image of a specimen after aging for one hour

 
 

Approaches for Phase 3

  • Path planning (conventional and machine learning supported): In the second phase, the fundamental cause-and-effect relationship between track geometry and the process parameters, taking into account the laser beam caustic and the shape of the powder gas jet, were identified and quantified. In the third phase, this knowledge and the created model will be used to optimize the path planning for complex components (thin-walled structures, repair applications on free-form surfaces). For this purpose, in the first step, the collected knowledge will be statistically evaluated and fed into a classical path planning tool (deterministic: according to "simple" rules, such as LMDCAM2 of Fraunhofer ILT) or used with one. In parallel, modern machine learning based approaches are used to solve the same problem (i.e.: optimized path planning with geometry adapted process parameters). The path planning will be tested, improved and demonstrated using complex geometries.

  • Temporal and spatial modulated laser radiation: Another goal is to vary parameters that have (usually) been set as constant so far. This includes firstly the temporal modulation of the laser power and the spatially adjusted intensity distribution. The melt pool model from phase 1, 2 must be validated for the temporally modulated power and adjusted if necessary. The optimal intensity distribution for the laser cladding process (currently Gaussian beam profiles are mostly used) is determined by solving the inverse heat conduction problem. This allows for a greater variety in terms of solidification conditions and geometry of the weld tracks.