Advances in materials science and computational capability have led to the multiscale modelling and simulation-based design of materials and systems for high value manufacturing.
Integrated computational materials engineering (ICME) is an emerging multi-disciplinary field that aims to establish causal relationships between material microstructure, manufacturing routes and properties. It seeks to integrate models defined over a range of temporal and spatial length scales to accurately represent material structures and their influence on properties, in order to provide a physics- based approach to manufacturing and performance. The benefits of such an approach are to improve component performance as well as reducing the cost and timescale of introducing a part to market through reductions in manufacturing trials and component level testing. The latter are considerable in industrial sectors such as transport, in particular, aerospace, and energy e.g. nuclear engineering. The Partnership for Research in Simulation of Manufacturing and Materials at the University of Birmingham is developing the tools needed to apply this new paradigm for manufacturing in industry.
Our vision is to deliver validated software solutions for component design and manufacture to industry through the integration of computer modelling, advanced experimentation and novel manufacturing technology͟. The integrated computational materials engineering approach allows this by focussing on modelling material microstructure, mechanical properties and manufacturing combined with experimentation to enable validated analysis of component material behaviour for design, development, production and use. It offers huge economic potential and the successful implementation of integrated computational materials engineering will revolutionise the way components are being designed and manufactured.
Integrated computational materials engineering is unique in that it is not dependent on a specific production technology but can be applied across the manufacturing sector. This drives innovation and increases competitiveness in the manufacturing sectors by reducing the time to market and increasing component performance.
This vision is being realised through fundamental and applied research aiming to establish the causal relationships between microstructure and properties to reduce phenomenology and allow location-specific property predictions. A framework has been developed for the implementation of computational tools that span a number of length scales and cover the entire manufacturing cycle for the optimisation of process routes and the provision of validated physics-based methods for product design and component lifing.
Tools are being developed and implemented to provide increased efficiency, shorter lead times, decreased material input, reduced operation times, increased dimensional accuracy and better performance by addressing the following challenges:
• Rapid insertion of materials and technology
• Near-net shape manufacturing
• Towards near-zero-defect manufacturing
• Location-specific properties
The initial exploitation of integrated computational materials engineering will be in the aerospace sector and the UK has a 17% global market share in aerospace industry revenues, making it the largest in Europe and second to the USA worldwide. In 2012, the industry had a turnover of some £20bn ** (~€2.43bn) and consequently the financial impact of integrated computational materials engineering is potentially highly significant for the UK alone. Studies of the benefits of modelling clearly show the significant returns on investment that can be achieved (see Fig. 1).1
The successful implementation of integrated computational materials engineering in industry will revolutionise the way components are designed and manufactured, thereby offering huge economic gains. One example of the advantages of integrated computational materials engineering is seen in an aerospace industrial implementation which has shown significant business benefits,2 resulting in:
• The reduction of casting scrap and US $5m ** (~€4.4m) savings per year
• A 90% reduction in the forming trials required for hot isostatic pressing (HIP) developments
• The shortening of development time to market of new product introduction and redesign efforts
The tools that address the challenges of rapid technology insertion and process design optimisation are fundamental to cost and component performance prediction. In addition, inclusion of material variability into the design optimisation allows improved risk analysis on component performance and life. The gains associated with component lifing are hard to quantify but it has been shown that the application of location specific lifing can result in doubling critical component lives.3
Application – a case study of Additive Manufacturing
The applicability of metallic powder-based production methods such as additive manufacturing (AM) are restricted by an inability to define the process parameters with sufficient accuracy to provide the quality required for industrial production due to a lack of understanding of both the gas-liquid phase interactions and the effect of the solid state phase transformations that occur in the relevant alloy systems. Traditional solutions, based on practical trials and physical assessment, are both costly and time consuming and for the long service lives encountered in the energy and propulsion industries are not feasible while empirically based phenomenological modelling approaches cannot provide the required fidelity. Thus, the recent developments in the direct fabrication of parts from metal powder using additive manufacturing (AM) techniques, such as selective laser melting (SLM), provide an opportunity to apply integrated computational materials engineering techniques in a closed manufacturing cycle.
However, the physical metallurgy of additively manufactured engineering alloys has been shown to give rise to complex non-trivial microstructure variations. These have a profound effect on the mechanical behaviour of AM builds resulting in location specific properties as well as playing a fundamental role in process-induced cracking. Understanding the causal links between manufacturing parameters, microstructure variations and property scatter is fundamental in the development of AM strategies for the high strength nickel-based superalloys used in the aerospace industries. Furthermore, establishing such causal links will provide a physically-based framework for the development of lifing methodologies and therefore a route towards uncertainty quantification of aerospace AM components. This case study presents a computational materials framework for the prediction of AM property scatter based on process-induced microstructures variations.
The modelling framework
A multi-scale materials modelling framework has been developed that seeks to establish computationally the causal relationships between process parameters, microstructure and the mechanical behaviour of an SLM build. An overview of the proposed SLM model framework is presented in Fig. 2. The framework is divided into seven broad computational domains:
• Powder deposition
• Melt-pool dynamics
• Solidification microstructures
• Representative volume element (RVE) reconstruction
• Full field simulations
• Location-specific property prediction
Associated with these domains are a number of computational material methods. The interconnection between these computational domains is presented in terms of a bottom-up hierarchy of models: discrete element methods (DEM), computational fluid mechanics, crystal plasticity, homogenisation and state variable approaches.
Simulations of the melt-pool evolution require an instance of the powder deposition layer to be created. This is achieved by a discrete element method (DEM) that simulates the adding of powder particles onto an existing surface, which may initially be the substrate or the deposition layer. In the present work, the DEM capabilities provided in the commercial finite element software ABAQUS have been utilised.
A volume of fluids approach has been adopted for simulating the SLM process that extends the Stokes-Navier field equations to multi-phase problems, where tracking phase interfaces is required. This allows the simulation of the powder melting and the prediction of the movement of the solid-liquid and liquid-vapour interfaces (see Fig. 3).
Schemes for predicting solidification structures (e.g., dendrites and grains, see Fig. 4) and solid state transformation (nucleation, growth and dissolution of precipitates) have been derived. A cellular automata (CA) finite difference (FD) scheme has been developed for the simulation of dendritic structures at lower length scales (in the nm to m range) to agglomerations of thousands of grains (from ** m to mm scale).
The grain structures generated by the CA-FD are used to constructive representative volume elements (RVEs) that aim to capture the localised microstructure variations associated with the ALM process (see Fig. 5). Since the thermal histories can be predicted the simulation of solid state precipitation reactions can be carried out. In nickel-superalloys these include precipitation of the intermetallic phase as well as the orthorhombic phase. From this statistical information of the process-induced microstructural features can be extracted such as precipitate and grain size distributions as well as their mean and standard deviations.
Full field simulations
During the deposition of process or under in-service conditions, an ALM build will undergo plastic distortions, which are determined by the microstructure distributions defining the RVE. Plastic deformation is driven by crystallographic slip and the appropriate kinematic formulations are required to capture the development of plastic strains at the microscale. Crystal plasticity provides such a formulation and allows the prediction of physical microscale fields, such as the process-induced residual stresses between and within grains.
Location specific properties
The term ‘location specific properties’ refers to the scatter in material properties within a manufactured component. These variations are linked to non-uniform distributions of process-induced microstructures as shown in Fig. 6. Through the homogenisation of these microscale fields over the RVE the ** macroscale/component level properties can be derived. From these it is possible to derive the causal relationships between the process-induced microstructure and the final part properties.
The way forward
A key requirement for systems-based approaches to digital manufacturing and alloy design is the integration of fundamental science, technology and innovation to ensure the adoption of the emergent and cross-cutting technologies needed for materials design, improved manufacturing and increased competitiveness. To achieve these aims, new inter-disciplinary fields have emerged including the integrated computational materials engineering (ICME), as discussed here, and the science of big data.
The development of multiscale materials modelling and the characterisation methods required for the validation of the integrated computational materials engineering approach has become a major activity of research that seeks to establish the relationships between chemistry, microstructure, process routes and the properties of materials. Establishing these causal links is fundamental in developing capabilities for system-based approaches that allow rapid insertion of new technologies as well as the development of novel materials for aerospace and energy sectors. Because of the range of length scales investigated, ranging from the atomistic to component level, a number of different characterisation methods are being used and include:
• Imaging of the internal structure of materials (macro-CT, micro x-ray CT, neutron TM, X-ray diffraction microscopy (10um-1m), Nano X-ray CT (1um), TEM tomography (0.1nm-1m))
• Mapping chemistry variations and atomic structure (XRD, PDF, XPS, SIMS, EPMA, HRTEM STEM EDX, Atom probe tomography)
• Mapping mechanical fields (Macro in-situ deformation, in-situ indentation, micro in-situ deformation, digital image correlation (DIC) and AFM)
Similarly, the simulation of material microstructures and properties is achieved through the application of a range of theoretical and numerical tools. These include interface tracking methods (level-set, phase fields), cellular automata, finite element and fast Fourier schemes. Full field simulations of the micromechanical and composition fields over representative volume elements of 100microns x100microns100microns are characterised by millions degrees of freedom.
However, major challenges exist in the development and usage of such multiscale characterisation and modelling capabilities such as:
• Large volume of data generated by these methods (processing of the raw data can take up 12 months in some cases)
• Gaps in software capability for transferring data to different analysis tools
• Computational facilities for large scale simulations
To meet these challenges in ‘Big Data‘ and large multiscale simulations co-ordinated efforts across disciplines is a requirement. The application of artificial intelligence (AI) methods will also be instrumental in extracting insights into the interactions between the various length scales using both experiment and simulation.
1: G. Goldbeck, ‘The economic impact of molecular modelling of chemicals and materials’. **it Goldbeck Consulting Report 2012, http://dx.doi.org/10.5281/zenodo.44350
2: Bolcavage **it et al. **it Integrating Materials and Manufacturing Innovation, 2014, 3:13
3: Brooks **it et al., ‘Probabilistic Property Prediction of Aero-Engine Components for Fatigue’, **it Proceedings of ASME Turbo Expo 2010, GT2010-22708