Predicting material properties Machine Learning designs cheaper and rust-proof steel for 3D printing

Related Vendors

HASCO Hasenclever GmbH + Co KG

AMPCO METAL Deutschland GbmH

Renishaw plc

A machine-learning strategy has generated a new class of ultra-high strength and ductility steel for 3D printing that costs less, resists rust, and requires only a fraction of the usual processing time.

In International Journal of Extreme Manufacturing, a new study demonstrates that integrating artificial intelligence with the fundamental physical and chemical properties of elements can rapidly identify optimal alloy recipes. The resulting metal achieves a rare balance of extreme strength and ductility, solving a persistent bottleneck in heavy manufacturing and aerospace engineering.

Currently, producing ultra-high strength and ductility steels through 3D printing requires heavy doses of expensive elements like cobalt, molybdenum, or high levels of nickel. Even with these premium ingredients, the printed parts must undergo complex, multi-step heat treatments in industrial furnaces to reach their final strength, and they often remain highly vulnerable to corrosion in harsh environments.

To bypass this trial-and-error chemistry, a research team from the University of South China and Purdue University turned to an "interpretable machine learning" model. Instead of treating the AI as a black box that simply guesses combinations, the team fed the algorithm 81 fundamental physicochemical features of various elements, such as their atomic radius, electron behavior, and how fast sound travels through them.

The algorithm calculated that a specific blend of iron and chromium, mixed with precise, small amounts of cheaper elements like silicon, copper, and aluminum, would form the ideal internal structure. After 3D printing the metal Fe-15Cr-3.2Ni-0.8Mn-0.6Cu-0.56Si-0.4Al-0.16C (wt.%) using a laser-directed energy deposition technique, the researchers baked it in a single-step tempering process at 480°C for just six hours.

The physical testing matched the algorithm's predictions. The resulting steel withstood stresses of 1,713 MPa and stretched by 15.5 percent before breaking. This represents an approximately 30 percent increase in strength over the metal's raw, printed state, accompanied by a doubling of its ductility.

The team investigated the metal's internal architecture to understand the mechanics behind this performance. They found that the short heat treatment forced the metal to grow a dense network of nanoscale particles, including copper and nickel-aluminum.

Toward cheaper, tougher tool parts

3D printing tungsten carbide-cobalt with less waste

When physical stress is applied to the metal, these tiny particles act as roadblocks that pin down structural defects and stop them from spreading, drastically increasing the force required to break the part. Simultaneously, microscopic pockets of a softer phase, known as austenite, act as shock absorbers by changing their crystalline shape to soak up energy, a phenomenon that prevents the steel from snapping under tension.

Rust resistance

The AI-designed recipe also solved the rust problem inherent to many high-strength alloys. In typical steels, the formation of carbides drains chromium from the surrounding metal, creating vulnerable, chromium-depleted zones where corrosion takes hold. The researchers found that the nanoscale copper particles in their new steel effectively expelled chromium during their formation, forcing it to remain evenly distributed throughout the surrounding matrix. In salt-water tests, the new alloy degraded at a rate of just 0.105 millimeters per year, significantly outperforming standard commercial stainless steels like AISI 420.

While the interpretable machine learning approach successfully cut costs and processing times, the researchers note that the methodology relies on datasets that are highly specific to certain manufacturing techniques. Because different 3D printing methods heat and cool metals at drastically different rates, data from one fabrication process is often incompatible with another.

In future work, researchers will need to re-screen these fundamental physical features when applying the AI to entirely new material classes. However, the study provides a clear blueprint for moving away from slow and empirical testing, offering a rapid pathway to design custom, high-performance components.

Original Article: Interpretable machine learning integrated with physicochemical feature for developing additively manufactured ultra-high strength and ductility steel; International Journal of Extreme Manufacturing; DOI:10.1088/2631-7990/ae5006

(ID:50807225)