Level 1 - Absolute Beginner

Rocket engines get very, very hot. NASA made a special metal for the hottest part of the engine. The metal is called GRCop-42. It is a mix of copper, chromium, and niobium.

GRCop-42 is hard to print with a 3D printer. Workers must try many settings. This takes a lot of time and money.

Scientists at Washington State University had a new idea. They used artificial intelligence, or AI, to help. The AI looked at more than 100 million possible settings. It picked the best ones to try.

The team only needed 40 real tests. The AI found six settings that worked well. One setting used very low power. This means more printers can use it. Small companies can now print this strong metal too.

rocket engine

The powerful machine that pushes a rocket up into space

alloy

A metal made by mixing two or more other metals together

3D printer

A machine that builds an object by adding thin layers of material one on top of another

artificial intelligence (AI)

Computer programs that can learn and make smart choices

setting

A control or option on a machine, chosen before you use it

laser

A strong, narrow beam of light used to melt metal in some 3D printers

power

How much energy something uses

combustion chamber

The part inside a rocket engine where fuel burns

Level 2 - Elementary

Inside a rocket engine, one part gets hotter than almost anywhere else: the combustion chamber, where fuel burns to push the rocket forward. NASA developed a special copper alloy called GRCop-42 to survive that extreme heat.

The problem is that GRCop-42 is very difficult and expensive to 3D print. Finding the right printer settings, like laser power and speed, usually takes many rounds of trial and error, and each failed attempt wastes time and material.

Researchers at Washington State University decided to let artificial intelligence handle the search. Their AI model looked at more than 100 million possible combinations of settings and estimated which ones were most likely to succeed.

The AI then guided a small number of real test prints, mixing settings it felt confident about with ones it was still unsure of, learning a little more with each round. After only 40 real experiments, it had identified six working configurations, including one that used a record-low laser power of just 500 watts, which means more ordinary, widely available 3D printers could handle the job.

combustion chamber

The part of a rocket engine where fuel and oxygen burn together to produce thrust

copper alloy

A metal mixture that has copper as one of its main ingredients

trial and error

Trying different attempts, learning from mistakes, until something works

configuration

A specific combination of settings used to complete a task

laser power

The strength of the light beam a 3D printer uses to melt metal powder

watt

A unit used to measure how much power something uses

adaptive

Able to change and improve based on new information

specialized

Made or designed for one particular purpose, often at extra cost

Level 3 - Intermediate

Rocket engines that burn liquid propellant subject their combustion chambers to some of the most punishing conditions in engineering: extreme heat, intense pressure, and rapid thermal cycling. To withstand this environment, NASA developed GRCop-42, a high-performance copper-chromium-niobium alloy prized for its strength and thermal conductivity at high temperatures.

That performance comes at a cost. GRCop-42 is notoriously difficult and expensive to 3D print, because manufacturers typically have to discover workable printer settings through repeated, costly trial and error, with each failed attempt consuming material, machine time, and money.

A team at Washington State University's School of Electrical Engineering and Computer Science and School of Mechanical and Materials Engineering took a different approach, applying an AI-driven adaptive experimental design method. Rather than testing settings blindly, the model estimated the probability that an untested configuration would succeed, then guided small batches that deliberately mixed promising options with more uncertain ones, refining its predictions after every round.

The payoff was striking: after only 40 real experiments (a small fraction of the more than 100 million possibilities the model considered) the system had identified six successful printing configurations. One of them operated at a record-low laser power of 500 watts, well below typical settings, meaning the alloy could potentially be printed on more widely available commercial machines rather than highly specialized equipment. The work, published in the Proceedings of the AAAI Conference on Artificial Intelligence, received the conference's Innovative Deployed Application Award.

thermal cycling

Repeated heating and cooling that puts stress on a material over time

thermal conductivity

A material's ability to carry heat away efficiently

experimental design

A planned strategy for deciding which tests to run and in what order

probability

A mathematical measure of how likely something is to happen

configuration

A specific, complete set of settings used together for a process

commercial printer

A 3D printer widely sold and used by businesses, as opposed to rare, specialized lab equipment

liquid propellant

Liquid fuel and oxidizer burned together to power certain rocket engines

deployed application

A technology that has moved beyond theory into practical, real-world use

Level 4 - Advanced

Among the harshest operating environments in modern engineering is the combustion chamber of a liquid-propellant rocket engine, where extreme heat flux, cyclic thermal loading, and high pressure converge on a component that must nonetheless remain dimensionally precise and structurally reliable across repeated firings. NASA's answer to that challenge is GRCop-42, a copper-chromium-niobium alloy engineered specifically to retain strength and thermal conductivity under conditions that would degrade conventional materials.

The alloy's very properties, however, make it a demanding candidate for additive manufacturing. Identifying viable laser powder bed fusion parameters has historically required extensive, costly trial and error, since the parameter space, encompassing laser power, scan speed, layer thickness, and related variables, is vast and the relationships between settings and print success are not easily modeled analytically.

Researchers at Washington State University's School of Electrical Engineering and Computer Science and School of Mechanical and Materials Engineering addressed this by applying an AI-driven adaptive experimental design framework. Rather than sampling the parameter space exhaustively or arbitrarily, their model assigned success probabilities to untested configurations and then selected small, deliberately mixed batches, some drawn from high-confidence predictions, others from regions of greater uncertainty, updating its estimates after each round in a closed-loop refinement process.

The resulting efficiency gain was substantial: from a search space exceeding 100 million theoretical configurations, the system converged on six viable printing parameter sets after just 40 physical experiments, a fraction of what conventional trial and error would typically demand. Notably, one configuration operated at a record-low laser power of 500 watts, suggesting the alloy could be manufactured on comparatively modest, widely available commercial systems rather than the highly specialized equipment the field has generally assumed necessary, a finding with direct implications for cost and accessibility across the smaller aerospace manufacturing sector. The work, published in the Proceedings of the AAAI Conference on Artificial Intelligence, was recognized with the conference's Innovative Deployed Application Award.

heat flux

The rate at which thermal energy passes through a given area of a surface

additive manufacturing

The formal term for 3D printing, in which objects are built up layer by layer

laser powder bed fusion

A 3D printing method that uses a laser to melt and fuse layers of metal powder together

parameter space

The full range of possible combinations of settings that could be used for a process

closed-loop refinement

A process in which results from each step are fed back to improve the next step