Titanium parts for aircraft, medical implants and complex equipment are already being printed on 3D printers, but selecting the printing mode often turns into lengthy and expensive trials. A new study shows how artificial intelligence can narrow the search before experiments.
The method in question is selective laser melting (SLM): a laser fuses metal powder layer by layer. The scientists focused on the Ti6Al4V alloy, a common titanium material valued for its strength at low weight, corrosion resistance and compatibility with the body.
The study involved specialists from Istanbul Technical University and Gazi University (Turkey), as well as partners from the University of Engineering and Technology in Lahore, the Pakistani-Austrian University of Applied Sciences and Technologies, the University of Pittsburgh and the University of Udine. The paper was published in the International Journal of Precision Engineering and Manufacturing.
The problem is that even small changes in laser power, scanning speed, layer thickness, hatch spacing, laser spot size and powder particle size affect whether a part will be strong, ductile or non-uniform in its properties.
The researchers assembled 201 experimental records from previously published works. In them, six printing parameters were linked to three tensile properties: yield strength (the stress at which the metal begins to deform irreversibly), ultimate tensile strength (the maximum load a specimen withstands before failure) and elongation at break (how much a specimen stretches before failure).
After comparing several machine learning algorithms, the team found that an artificial neural network gives the most stable predictions for all three properties. They then applied SHAP analysis, an explainable AI method that shows how much each printing parameter contributed to the model's prediction.
The most practical part is the inverse design interface. Unlike models that only predict properties from given settings, this tool allows the user to enter the desired strength and ductility and obtain recommended parameters for further verification. Narrowing the range of promising modes before experiments can reduce unnecessary trials, saving time, material and money.
"Metal additive manufacturing still requires careful process control, especially for critical applications such as aerospace structures and biomedical implants," noted Huseyin Kizil of Istanbul Technical University, who led the work. "Our goal is to give engineers a data-driven decision-making tool to narrow the process window before experimental verification."
The authors emphasize that the tool does not replace physical testing, quality control and certification. It can help at the pre-production assessment stage by directing experiments toward more promising combinations of parameters and improving understanding of the relationship between printing conditions and material properties.
The approach could be useful for companies and research groups developing titanium components where reliability, precision and material savings matter. After further validation on data from specific machines and in real production conditions, AI suggestions could shorten development cycles, reduce waste and make the output of high-performance metal parts more consistent.