How forty years of material-selection wisdom are being rewritten by machine learning, sustainability add-ons, and cloud-native data intelligence
Material selection has always been the decision that locks in the majority of a product’s cost, weight, and environmental footprint. For decades engineers solved it with Ashby charts — plotting modulus against density, strength against embodied energy — and picking the material that landed on the right Pareto front. In 2026 the Ashby logic is still sound, but the data behind it has exploded and machine learning has moved from augmenting selection to driving it. This article walks through the evolution from Ashby to AI-native selectors, highlights the major platforms (including Ansys Granta MI 2026 R1), and shows how Simreka makes sustainability a first-class dimension of material selection, not an afterthought.
The Ashby Legacy: Still Right, Still Incomplete
Michael Ashby’s material-selection methodology, developed through the 1980s and 1990s, is the conceptual backbone of modern material-selection software. The idea is simple: identify the performance index that matters (e.g., specific strength for a lightweight beam, thermal shock resistance for a crucible), plot candidate materials in a 2D property space, and pick those on the upper-left Pareto boundary. Granta Design (founded by Ashby and Mike Hobday) turned this methodology into commercial software that became the industry standard — and that software is now the Ansys Granta MI product family, which in 2026 marked 20 years of materials intelligence deployments across aerospace, automotive, energy, and electronics.
The Ashby approach is still correct, but its classical form has three limits: it treats material properties as point values rather than distributions, it struggles with more than two objectives at a time, and — until recently — it did not natively incorporate sustainability metrics like embodied carbon, water, and end-of-life recyclability. Each of those limits is being addressed by 2026-era tooling.
What Changed: AI Enters the Selection Loop
Granta MI AI+ brings machine learning into the Granta environment directly: ML algorithms can be applied to Granta MI datasets to predict properties, infer needed process parameters, and surface material candidates that no engineer would have thought to compare. Ansys Granta Selector 2026 R1 adds expanded sustainability and restricted-substances data, with updated EU REACH and regulatory coverage baked in, so compliance filtering happens as part of selection rather than as a post-hoc check. The 2026 R1 release also introduces radar plots, enhanced record comparison with color-coded differences, and Record Tree browsing — visual upgrades that sound cosmetic but materially accelerate engineering decision-making.
The broader picture: material selection is moving from lookup and chart-plotting to an interactive, AI-augmented decision environment. Selectors now incorporate uncertainty ranges, multi-objective Pareto navigation across 3–5 dimensions, and on-the-fly LCA scoring. The engineer still owns the decision, but the tool narrows a 10,000-material catalog to a 10-candidate shortlist in seconds.
The New Dimensions: Sustainability, Supply Chain, Regulation
Modern material-selection data models extend far beyond mechanical properties. The axes that matter in 2026 include:
| Dimension | Typical Metrics | Why It Matters |
|---|---|---|
| Mechanical / functional | Modulus, strength, toughness, conductivity | Classical performance — still the first filter |
| Embodied impact | kg CO&sub2;e/kg, MJ/kg, m³ H&sub2;O/kg | Drives Scope 3 reporting and regulatory filings |
| End-of-life | Recyclability, biodegradability, disassembly | Circular-economy compliance, ESPR requirements |
| Regulatory | REACH SVHC, TSCA, RoHS, PFAS restrictions | Market access and legal risk |
| Supply chain | Geographic sourcing, concentration, volatility | Resilience against trade disruptions |
| Economic | $/kg, price volatility, scalability | Product-cost impact and viability |
| Process compatibility | Machinability, moldability, weldability | Manufacturing feasibility with existing assets |
The Ashby instinct still applies — look for the Pareto frontier — but now the frontier is in a 7-dimensional space, and you need algorithms rather than intuition to find it. This is the space where AI-driven multi-objective optimization outperforms human spreadsheet triage by an order of magnitude.
A Concrete Example: Selecting a Lightweight Automotive Housing
Suppose an OEM wants to select a material for a battery housing: target mass ≤ 3.5 kg, stiffness class A, maximum 2.5 kg CO&sub2;e/kg, no SVHC substances, ≥ 30% recycled content, and a cost under €4/kg. A traditional Ashby chart plot of density vs. specific stiffness gives a starting shortlist (fiber-reinforced polymers, aluminum alloys, magnesium alloys). Feeding all seven constraints into an AI-powered selector narrows that further to three candidates: a 30%-rPET long-glass-fiber composite, a recycled aluminum AlSi10MnMg alloy, and a hybrid PP/flax-fiber option. Each comes with a quantified embodied-carbon estimate, regulatory status, supply-chain concentration index, and LCA score. The engineer chooses among three, not three hundred.
How Simreka Layers on Top of Selection Tools
Commercial selectors like Granta MI are excellent at catalog lookup and Ashby-style comparison, but they rarely design new formulations. That is where the Simreka AI-Formulator fits in: once selection narrows to a class (say, a fiber-reinforced biopolymer), Simreka optimizes the exact recipe — fiber loading, compatibilizer, plasticizer, stabilizer — against multi-objective targets. The Simreka LCA & Impact Assessment module makes sure every recipe carries a live embodied-impact score. The Simreka Regulatory Compliance module filters against REACH, TSCA, SVHC, and PFAS lists in real time. And the Simreka Recycled & Alternative Materials module surfaces bio-based, rPET, and mechanically recycled feedstocks that can replace virgin inputs without compromising performance. Selection and design become a single continuous workflow.
Pitfalls That Still Trip Up Data-Driven Selection
Three pitfalls consistently undermine otherwise excellent selection programs. First, stale data: supplier property sheets from 2015 are worse than useless if you’re designing a 2027 product. Keep your material database on a refresh cadence. Second, missing sustainability fields: a selector is only as green as the data it knows about, and many legacy records still have no embodied-impact entry. Enrichment is a one-time investment that pays out forever. Third, false precision: ML models happily report predictions to four decimal places even when the underlying data has 20% noise. Always look at uncertainty, not just central estimates.
Conclusion
The Ashby chart was one of the most influential ideas in 20th-century engineering, and its logic still holds. What has changed is the data density, the number of dimensions you can handle, and the availability of ML tools that can find Pareto frontiers in seven-dimensional spaces. Ansys Granta MI 2026 R1 and its AI+ extensions bring this capability into the engineer’s daily workflow; platforms like Simreka extend it all the way from selection to recipe design to LCA scoring. The organizations that win the next decade of sustainable-materials work will be the ones who treat selection as a live, data-driven, multi-objective optimization — not a one-time lookup.
Frequently Asked Questions
Q1. Do I still need Ashby charts if I have ML-powered selectors?
Yes — they remain the best tool for intuitive 2D exploration and for educating non-specialists on tradeoffs. ML selectors complement, not replace, them; the Simreka MatIQ AI copilot can render Ashby-style projections from its multi-objective output on demand.
Q2. How reliable are embodied-carbon figures in selection databases?
Quality varies. Top-tier databases (ecoinvent, Granta sustainability add-on) use peer-reviewed LCA methods with documented uncertainty. Older or supplier-declared figures can be optimistic. Always check method and scope (cradle-to-gate vs. cradle-to-grave), and prefer sources whose data is harmonized inside the Simreka Databank.
Q3. Can small teams access this kind of tooling?
Yes. Ansys Granta EduPack, the Simreka AI-Powered Formulation Generator, and open datasets let small teams run data-driven selection workflows without enterprise licensing. Start small, prove value, then scale.
Q4. How do I handle materials with no data in the public databases?
Combine supplier-declared sheets with transfer-learning ML models that predict properties from composition and process. Uncertainty will be wider than for cataloged materials; plan lab validation accordingly, ideally inside the Simreka Virtual Experiment Platform so each prediction carries an uncertainty band.
Q5. What is the right frequency for refreshing material data?
Annually at minimum for supply-chain and regulatory fields, biannually for LCA (methods and impact factors evolve), and as-needed for performance fields tied to supplier revisions. The Simreka Databank automates this refresh cadence so you don’t have to chase it.
Q6. How should selection integrate with CAD/CAE workflows?
Tightly. Granta MI integrates directly with Ansys Mechanical, Abaqus, Siemens NX, and others, pushing selected material properties into simulation without manual re-entry. Losing this integration reintroduces the transcription errors that materials intelligence platforms exist to eliminate — request a Simreka demo to see the same integration applied to sustainable-formulation design.
Bibliographical Sources
- Ansys. Granta: Materials Information Management. https://www.ansys.com/products/materials
- Ansys Blog. Granta MI Software Marks 20 Years of Materials Intelligence. https://www.ansys.com/blog/ansys-granta-mi-software-marks-20-years
- Ansys. Granta Selector — Materials Selection Software. https://www.ansys.com/products/materials/granta-selector
- Ansys. Granta MI Enterprise — Material Data Management. https://www.ansys.com/products/materials/granta-mi
- Ansys. Granta EduPack — Software for Materials Education. https://www.ansys.com/products/materials/granta-edupack
- Ansys Webinar. Deep Dive on Eco-Design Powered by Materials Intelligence with Granta MI. https://www.ansys.com/webinars/deep-dive-on-eco-design-powered-by-materials-intelligence-with-ansys-granta-mi
- EDRMedeso. Granta MI Materials Intelligence Platform. https://edrmedeso.com/products/ansys-granta-mi/
- Cadfem. Ansys Materials Product Family. https://www.cadfem.net/en/our-solutions/ansys-simulation-software-the-product-family/ansys-materials.html
Make Sustainability a First-Class Selection Axis
Stop choosing materials on mechanical properties alone. Request a Simreka Demo → and see multi-objective, LCA-aware material selection in action.
