How & Why Sarginsons Industries Developed Spatially Variable TYE Simulations for Aluminium Castings

How & Why Sarginsons Industries Developed Spatially Variable TYE Simulations for Aluminium Castings

Using accumulated datasets from physical testing of cast components, Sarginsons has developed an AI-driven program for predicting tensile strength (T), yield strength (Y), and elongation (E)—collectively TYE—at any point across a cast component's geometry. This is an advance on current technology which assumes homogeneous TYE across the form and is, as such, unreliable for simulations of cast component performance.

This enables Sarginsons to create a precise digital twin for variable mechanical properties, which can then be exported to a finite element analysis (FEA) mesh for virtual simulations, such as crash testing, fatigue analysis, or structural optimization.

Below is a step-by-step technical outline of the process, drawing on established methodologies in materials engineering.

1. Data Collection and Dataset Construction

The foundation is a comprehensive dataset built from historical and ongoing testing of cast aluminium components ( included A380, A356 and Sarginsons structurally enhanced recycled alloy). Key data sources include:

• Process Parameters: Melting temperature, pouring speed, cooling rate, mould geometry, initial hydrogen content, and heat treatment details (e.g., normalizing holding time).

• Microstructural Features: Measurements of porosity, secondary dendrite arm spacing (SDAS), grain size, and inclusions, obtained via techniques like X-ray computed tomography (CT), scanning electron microscopy (SEM), and optical microscopy.

• Mechanical Testing Data: Point-specific TYE values from tensile tests (per ASTM E8 standards), where samples are extracted from various locations on castings to capture spatial variability due to uneven cooling or segregation.

• Spatial Metadata: 3D coordinates or positions within the casting geometry, correlated with simulation data for features like local solidification time.

The datasets have been pre-processed to handle outliers, balance classes, normalize features, and ensure high dimensionality.

2. Model Development for TYE Prediction

Using the dataset, Sarginsons has developed an ML model to predict TYE as a function of position and process inputs. It has determined its own algorithms based on aspects of the following approaches, which it then deployed via edge computing in the foundry for production defect prediction:

• Neural Networks: Multi-layer perceptrons with backpropagation, optimized via particle swarm optimization (PSO) for faster convergence. Inputs: Chemical composition (e.g., Fe/Mn ratio, Sr content), cooling rate, porosity; Outputs: UTS (T), yield strength (Y), elongation (E). Achieves RMSE < 6 MPa for UTS and <0.1% for E.

• Ensemble Methods: Robust for handling nonlinear relationships and providing interpretability via SHAP (SHapley Additive exPlanations) values, revealing feature importance (e.g., cooling rate > porosity for yield strength).

• Hybrid Approaches: Combine physics-based simulations (e.g., cellular automaton for microstructure evolution) with ML to predict location-specific properties, enabling point-by-point TYE mapping across the component's 3D form.

Training involved splitting the dataset (80/20 train/test), cross-validation, and hyperparameter tuning (e.g., via grid search). The model infers TYE at arbitrary points by interpolating from trained spatial data, often using Gaussian processes for uncertainty quantification.

3. Exporting Predicted Data to FEA Mesh

Once TYE is predicted at discrete points, the data is mapped to an FEA mesh for virtual testing:

• Mesh Generation: Use FEA software to create a tetrahedral or hexahedral mesh of the cast component, with element sizes refined in high-stress regions (e.g., 0.5-2 mm).

• Property Assignment: Interpolate predicted TYE values onto mesh nodes/elements using kriging or nearest-neighbour methods. This creates a heterogeneous material model where properties vary spatially.

• Data Export Format: Output as .inp (Abaqus input) or .cdb (ANSYS) files, with TYE assigned via user-defined material subroutines (e.g., UMAT in Abaqus) that incorporate plasticity models like Johnson-Cook for crash simulations.

• Integration Workflow: The AI program exports a point cloud of TYE data, which is imported into FEA pre-processors for field mapping. This enables full virtual testing, such as explicit dynamic simulations for crashworthiness, predicting failure modes with 90-95% accuracy compared to physical tests.

4. Approach Benefits

The new approach represents a significant advancement in materials engineering, particularly for aluminium foundries and component designers. Unlike traditional uniform property assumptions, this granular prediction allows for more accurate virtual testing via finite element analysis (FEA), enabling engineers to foresee performance under stress, fatigue, or impact. Benefits span cost savings, design innovation, and sustainability, as evidenced in modern foundry practices.

• Reduction in Physical Prototyping

One of the primary advantages is the substantial reduction in the need for physical prototypes, which are traditionally costly and time-intensive to produce. By simulating TYE variations across the casting's geometry, engineers can virtually test multiple iterations for defects like shrinkage, porosity, or hot tearing before committing to tooling or production. For instance, casting simulation software can replicate molten metal flow, solidification, and cooling in hours rather than weeks, identifying issues like air entrapment or uneven feeding early.

This can cut prototyping cycles by up to 50%, lowering material and labor costs while accelerating time-to-market. In aluminium foundries, where alloys like A356 are common, simulations predict location-specific properties tied to microstructural features, reducing trial-and-error from 10-20 physical runs to just a few validations.

Overall, this shifts resources from reactive fixes to proactive design, with reported ROI improvements through fewer scrapped prototypes.

• Rapid Iterative Design

Point-by-point TYE simulations facilitate rapid iterative design by allowing quick adjustments to process parameters (e.g., pouring temperature, gating systems) and immediate re-simulation of outcomes. This enables engineers to optimize designs in virtual loops, testing scenarios like varying cooling rates or alloy compositions without physical rework.

For example, if a simulation reveals low yield strength in a thin section due to rapid solidification, designers can iteratively refine the mould or add risers, achieving balanced flow and superior yield in days instead of months.

In automotive applications, this speeds up development for complex parts, integrating with CAD/FEA tools to predict fatigue and stress, fostering innovation like tailored microstructures for enhanced ductility.

The result is shorter design cycles—often 30-50% faster—boosting competitiveness in industries like aerospace and EVs.

• Topological Optimization for Lightweighting

Topological optimization, enhanced by TYE simulations, is key for lightweighting, where material is strategically placed only where needed for strength, reducing weight while maintaining performance. By mapping variable TYE, simulations identify load paths and stress concentrations, allowing algorithms to "remove" unnecessary material virtually, achieving up to 50% weight reductions in aluminium components.

This is particularly effective in EVs, where lighter structures extend range; for instance, optimizing for higher tensile strength in high-stress zones while allowing lower elongation in others minimizes mass without compromising safety.

Combined with AI-driven models, it enables counter-intuitive geometries that traditional uniform-property assumptions couldn't validate, leading to fuel efficiency gains and reduced emissions.

• Minimizing Scrap

TYE simulations minimize scrap by predicting defects like porosity or inclusions that lead to weak spots, allowing preemptive process tweaks. By forecasting TYE variations, foundries can optimize parameters to ensure uniform properties, reducing rejection rates from 10-20% to under 5%.

For aluminium castings, where scrap from failed tensile tests is common, simulations correlate process data (e.g., cooling rates) with outcomes, preventing over-pouring or misfeeds.

This not only cuts material waste—critical for sustainability—but also lowers energy use in remelting, with studies showing 20-30% scrap reduction through virtual defect anticipation.

In low-carbon foundries, this aligns with zero-waste goals by enabling first-pass success.

• Different Casting Techniques (Sand, Gravity, Low Pressure)

The flexibility of TYE simulations lies in using technique-specific datasets, tailored to methods like sand casting (slow cooling, higher porosity), gravity die casting (moderate flow, better density), or low-pressure die casting (LPDC, reduced turbulence for superior integrity). These datasets incorporate unique parameters—e.g., sand's thermal conductivity affects local yield strength, while LPDC's pressure minimizes gas entrapment for higher elongation.

Machine learning models trained on these datasets predict TYE accurately (e.g., RMSE <6 MPa for tensile), allowing cross-technique comparisons to select the optimal method. This customization enhances quality control, with sand datasets focusing on shrinkage prediction and LPDC on fatigue resistance, ultimately improving yield across diverse production scenarios.

• Predicting Structural Integrity to Facilitate Single Piece Assemblies (Mega Castings)

TYE simulations are pivotal for mega castings—large, single-piece aluminium structures replacing multiple components—by predicting structural integrity through variable property mapping. This eliminates welds (potential failure points), enhancing crashworthiness via controlled load distribution and unified geometry.

Simulations forecast TYE in critical zones, ensuring high yield strength for impact absorption while reducing weight by up to 30%, boosting EV range. Benefits include lower assembly costs, improved scalability, and repairability—e.g., designed failure zones prevent total write-offs, cutting repair costs vs. steel structures.

In EVs, this supports sustainability by simplifying recycling and reducing manufacturing emissions.

• Predicting & Countering Deformation in Heat Treatment

Heat treatment (e.g., solutioning, quenching, aging) can induce distortions in aluminium castings due to thermal stresses and phase changes, but TYE simulations that can predict these on a point-by-point basis, enable counter-intuitive designs that compensate for them.

By modelling residual stresses and plastic deformation (e.g., via creep models), engineers can pre-distort geometries in CAD to achieve final shapes post-treatment, allowing asymmetric or thin-walled features that defy traditional intuition.

For A356 alloys, simulations using heat simulation tools can predict quenching-induced cracks or warpage with 90% accuracy, optimizing support frames to minimize deviations.

This unlocks innovative designs, like mega castings with integrated cooling channels, ensuring structural integrity while reducing post-processing needs and enabling lightweight, high-performance parts

5. Who Benefits?

Point-by-point TYE prediction simulations offers transformative advantages in casting, from cost savings to innovation. Below is a list of key stakeholders who would benefit, along with explanations grounded in the described benefits like reduced prototyping, lightweighting, and structural optimization.

• Aluminium Component Manufacturers: They gain from minimized scrap (up to 30% reduction), rapid iterative design, and technique-specific datasets (e.g., for sand vs. low-pressure casting), leading to higher yields, lower production costs, and sustainable operations with less waste and energy use.

• Automotive and Electric Vehicle (EV) Producers:

Companies like Tesla or Ford benefit from topological optimization for lightweighting (up to 50% weight reduction) and mega castings, enabling single-piece assemblies that enhance structural integrity, improve fuel efficiency/range, and predict heat treatment deformations for innovative, counter-intuitive designs that cut assembly time and repair costs.

• Aerospace and Defence Industries:

Entities such as Boeing or SpaceX would leverage precise TYE mapping for high-strength, lightweight components, reducing physical prototyping cycles by 50% and ensuring reliable performance under extreme stresses, which accelerates development of aircraft parts or satellites while minimizing material use.

• Design Engineers and Product Developers:

These professionals benefit from rapid iterations and virtual crash testing via FEA integration, allowing counter-intuitive designs that account for heat treatment deformations, fostering creativity and faster time-to-market without the risks of physical failures.

• Sustainability and Environmental Organizations:

Groups focused on green manufacturing, like those in the EU's Green Deal initiatives, gain from low-carbon processes that minimize scrap and enable efficient mega castings, reducing overall emissions and material waste while promoting recyclable, single-piece designs over multi-component assemblies.

Sarginsons believes that it’s groundbreaking work in casting simulations will transform the cost, capability and performance of aluminium casting and is currently working on the next stage of this development, the integration of these simulation technology with metal solidification, digital twins and topological optimisation to create a one stop design programme for structural cast components.