Every 3D-printed part is a lottery ticket. Instead of money, the prize is service life: how many times you can load the part before a crack appears. The ticket number is set by the random grains and pores inside the metal. Engineers used to take a single number with a large safety margin, but now they can see all possible outcomes.
Scientists printed steel cubes and examined them under a microscope and with X-rays. Each cube was divided into 64 parts, and for each part they calculated how easily a crack might form there. The most demanding check of a single part took almost 18 hours, so all calculations ran for more than 240 days.
From all the calculated numbers, they obtained a probability distribution. Most parts last around 29,300 cycles, but some reach up to 31,400. If this distribution is used to plan inspections, a part can be operated for 4,200 cycles longer than with the old safety-factor approach. This method is useful in aviation, energy, and medical implants, where it is important to know not just the average lifetime but also the spread of possible outcomes.
The lottery ticket of a 3D-printed part can now be read in advance, rather than waiting for it to be drawn. This helps decide when to inspect a part and when to replace it, without relying on crude safety margins.
The crystal plasticity simulation for a single subcube took almost 18 hours on a 28-core processor, whereas the elastic calculation finished in 355 seconds, and the graph search for the crack path in 91 seconds. For 64 subcubes and five cases, the total time for plasticity alone exceeded 240 days of machine time.
In Kim Stanley Robinson's novel 'Aurora', the interstellar ship suffers from fatigue cracks in printed structures that cannot be predicted in advance. This work offers precisely the tool the characters lacked: a probability distribution of fatigue life that allows repairs to be planned before a crack becomes critical.