1 Introduction
2 Digital twin and process chain concepts in MPBF
2.1 Overview of a digital twin concept
2.2 Overview of process chain simulation concept
3 Laser powder bed fusion (LPBF)
3.1 Micro and meso scale models
3.2 Component scale models
4 Electron beam powder bed fusion (EBPBF)
-
Uncoupled thermal models, which simplify the physics and reduce the computational resources required by only considering the main heat transfer phenomena. These consist of conduction between the powder particles, conduction between the powder bed and substrate, and radiation from the powder into the chamber. Heat dissipated by viscous forces and convection in the melt pool are ignored [89, 90].
-
Thermomechanical models, which incorporate mechanical properties to predict the residual stresses induced in the material by the EBPBF process. This results in more informative models at the cost of longer run times [93].
5 Topology optimisation models
6 Post-processing operations
6.1 Heat treatment
6.2 Machining operations
6.3 Surface engineering
6.4 Inspection
7 Data acquisition and transfer in manufacturing process chain simulation
7.1 Data acquisition
7.2 Data transfer
8 Uncertainties in modelling, validation and data transfer
8.1 General overview
-
Identification of uncertain input quantities and quantification via appropriate probability density functions (PDFs);
-
Propagation of the PDFs through the model;
-
Summarising the results statistically.
8.2 Input parameters and assigning distributions
Information available | PDF derived using the principle of maximum entropy |
---|---|
Lower and upper limits a and b | Rectangular: R(a, b) |
Inexact upper and lower limits a ± d, b ± d | Curvilinear trapezoid: CTrap(a, b, d) |
Sinusoidal cycling between lower and upper limits a, b | Arc sine (U-shaped): U(a, b) |
Best estimate x and associated standard uncertainty u(x) | Gaussian: N(x, u2(x)) |
Best estimate x of a non-negative quantity | Exponential: Exp(1/x) |
8.3 Uncertainty propagation and sensitivity
8.4 Uncertainties in data transfer
-
change of formats between different software packages;
-
change of physics;
-
change of geometry or spatial distribution;
-
change of time distribution;
9 Maturation of MPBF modelling technologies
-
Mechanical properties: prediction of properties that govern the mechanical behaviour of the part, such as the Young’s modulus.
-
Physical properties: prediction of properties that govern the non-mechanical behaviour of the part, for example thermal, electrical conductivity or density.
-
Micro-cracks: the ability to predict small cracks that form and propagate at the micro and meso scales.
-
Surface roughness/texture: the surface finish quality of an MPBF component following a build.
-
Micro-structure: the microscopic structure of the material, including grains and precipitates, that evolves as it cools and solidifies.
-
Porosity: the ability to predict the amount of non-material space in the part, such as gas pores or other voids.
-
Powder distribution/raking: the morphology and distribution of metal powder as it is spread over the build chamber bed.
-
Melt pool temperature: the variation in the temperature of the molten material in the melt pool with time.
-
Lack of fusion: areas of the part where the powder has failed to fully melt.
-
Overheating: the temperature of a component and the risk of it becoming too high during a build.
-
Geometric distortion: the deviation of the part geometry from that which is intended due to stresses and strains that arise during the build process.
-
Recoater collision: the risk of the powder recoating blade within the machine hitting the part due to geometric distortion.
-
Residual stresses: stresses remaining in the part following manufacture that arose as a result of the manufacturing process.
10 Conclusions
Technology | Domain | Concluding remarks & future perspectives |
---|---|---|
Metal powder bed fusion (MPBF) | Modelling of powder | Rain models and models based on the discrete element method have been used to simulate powder raking and distribution of powder. Models should be advanced by incorporating non-spherical shapes. The benefits of the models are that they could support the design of the recoater blade and its operation. Another research challenge is to predict potential defects based on the powder distribution. This challenge would require further research effort by incorporating uncertainty models to describe the input randomness of the initial powder distribution. |
Prediction of temperature | Models for the prediction of the melt pool temperature are available, but the validation of the predicted melt pool temperature history remains a challenge. Prediction of the thermal history at component scale is a challenging due to high computational demands. Analytical models have been developed, as well as method to lumps layers together, but there are still no strong validation procedures, which could be a focus of research. | |
Prediction of material properties | Prediction of material properties is heavily dependent on the accuracy of the predicted temperature field which is used for the prediction of the microstructural evolution. There are existing theories based on multi-scale approaches to predict the mechanical properties in MPBF. However, their application at component scale is still a challenge due to the challenges in predicting the thermal history. It is expected that more research emerges in this field in the near future. | |
Prediction of distortion | Despite the fact that distortion is depended on the thermal history, macro-scale approaches (e.g. inherent strain method) have been successfully applied to predict distortion at a component level. Sufficient accuracy has been demonstrated and commercial software tools are available. However, for distortion compensation where high accuracy is required, the macro-scale predictive methods for distortion might show weaknesses for some geometries and materials which would require to understand the limitation of those modelling techniques. | |
Prediction of residual stresses | Bulk residual stresses have been predicted with macro-scale approaches. There are evidences that the prediction of residual stress field can be reliable. However, the predictions of residual stress distribution need to be further understood at micro-scale and how it affects the development of micro-structure and defects. | |
Topology optimisation | Topology optimisation tools for mechanical loads have been widely used in light-weighting applications against proof and fatigue performances. Manufacturing constrains have been also researched and incorporated in the topology optimisation algorithms for MPBF. Embedding manufacturing constrains from post-processes into topology optimised algorithms is an area where further research can be conducted. Optimisation of surfaces for fluid flow and electro-magnetisms is also an area where further research is needed. | |
Post-processing | Heat treatment | Modelling of temperature history in a furnace is possible. Stress relief models have been also validated and applied to MPBF. Validation of furnace models for prediction of the thermal history in the part could be enhanced. Rapid micro-structural changes in a stress field during the heating could be further research to avoid cracks due to brittleness effects. |
Machining operations | Modelling of machining on additively manufactured parts has been researched, including modelling of bulk material removal and the tool-workpiece interaction. Prediction of microstructural changes and material properties at the surface is an area where further research can be dedicated. | |
Surface engineering | Physics-based modelling of surface finishing with conventional and non-conventional technologies remains a challenge due to computational demands. Empirical-based modelling techniques have been primarily used. Surface hardening with shot-peening and laser shock peening have been modelled to predict the residual stress effect. Residual stresses profiles have been superimposed on surfaces using mathematical models. | |
Digital twins | Manufacturing process chains | Research on modelling and simulation of manufacturing process chains has been demonstrated where distortion and residual stresses have been transferred from the LPBF to other post-processes (e.g. heat treatment, machining and surface hardening). However, there is a lack of validation of the predicted residual stresses and distortion for the entire manufacturing process chain. There is also a gap in estimation of errors and uncertainty quantification due to data transfer between models. Modelling the evolution of micro-structed and the non-homogeneousness of the material has not been studied in the context of manufacturing process chains. There is also a lack of modelling capabilities to predict defect and their evolution in the manufacturing process chain. The integration of process models into manufacturing process chains requires further development and validation to enable the emergence of simulation-based process chain digital twins in the near future. |
Interoperability | The data transfer between models in the process chain has been addressed to some extent. However, there is no standardised data transfer formats to enable the wider usability within the research community. Also, links between inspection data and predictive models should be developed. Development of architectures for enabling seamless interoperability should be addressed in the development of future digital twins. | |
Computational performance | The computational power is still a demand in modelling the physics in LPBF and post-processes. The development of surrogated models by using simulation data generated with validated physics-based predictive models is an area where long computational efforts can be overcome in applications requiring real-time calculations (e.g. deep machine learning techniques). Future research for the development of hardware and fast software algorithms is needed. |