Dynamic reliability assessment of aero-engine components using dependence-aware PNET and mechanism-informed aging correction

Inside the heart of a modern jet engine, a small collection of metal parts works under conditions that would melt most materials. These components, known as the hot section, sit between the burning fuel and the spinning turbine. They endure extreme heat, violent vibrations, and the crushing force of air moving at supersonic speeds. Over time, this brutal environment causes the metal to weaken, crack, or wear down. Engineers must predict exactly when these parts will fail to ensure the engine remains safe. If they guess too early, they replace parts that are still good, wasting money and grounding aircraft. If they guess too late, the engine could fail in flight. The challenge is that these parts do not age in a simple, straight line. They suffer from many different types of damage at once, such as cracks from vibration, wear from rubbing, and slow stretching from heat, all happening simultaneously and influencing one another.

For decades, engineers have tried to calculate the safety of these engines by looking at each type of damage separately. They would estimate the chance of a crack forming, the chance of wear occurring, and the chance of heat damage, then multiply these numbers together to get a total safety score. This approach assumes that each type of damage happens independently, like rolling separate dice. However, in the real world, the same forces that cause a crack also cause wear. A single burst of high temperature might speed up both processes at the same time. When engineers treat these linked problems as if they were unrelated, they often count the same risk twice, leading to a confusing picture of how safe the engine really is. Furthermore, simply counting the hours an engine has flown does not tell the whole story. An engine flown in a hot, humid climate with frequent takeoffs and landings ages much faster than one flown in a calm, cool environment, even if the clock shows the same number of hours.

A team of researchers has developed a new way to solve this puzzle by combining two ideas that are usually kept apart. First, they created a method to recognize when different types of damage are actually connected, so they do not count the same risk twice. Second, they introduced a way to translate the raw hours an engine has flown into a measure of actual damage, accounting for how hard the engine was working. They tested this new system on a high-pressure turbine blade and disk assembly, a critical part of a jet engine that connects the spinning blades to the central disk. This specific part is made of special metal alloys and is held together by a complex, tree-root-like structure called a fir-tree attachment. It is a perfect example of a component where heat, vibration, and rubbing forces all collide.

The researchers started by identifying seven major ways this part could fail. These included cracks from low-speed bending, cracks from high-speed vibration, wear where the blade touches the disk, cracks in the disk itself, peeling of the protective coating, damage from foreign objects hitting the blade, and slow stretching from heat. Instead of treating these seven problems as seven separate enemies, the team looked at how they influenced each other. They built a system that measured how similar the risks were. If two failure modes were driven by the same forces, like the same heat or vibration, the system recognized them as a pair. It then grouped these related failures together and picked the most dangerous one to represent the whole group. This process, which they call a dependence-aware screening, meant they no longer had to calculate the safety of seven separate items. Instead, they could focus on a few key representatives that captured the true risk of the entire group.

In their analysis of the turbine assembly, the researchers found that the rubbing wear at the connection point was the most important representative for a large group of related failures. This single mode of damage, known as fretting wear, stood in for the cracks and stresses happening in the same area. As the engine flew more hours, the group of related failures remained stable, but the secondary group of failures changed. Early in the engine's life, the second most important failure mode was slow stretching from heat. However, as the engine aged and the metal fatigued, the focus shifted. The second most critical risk became a crack forming in the slot of the disk itself. This shift showed that the most dangerous part of the engine changes over time, and a static list of risks would miss this evolution.

The second part of their new method addressed how fast the engine was actually aging. The team realized that an engine flying in severe conditions accumulates damage much faster than one flying in mild conditions. They created a way to convert the calendar hours of flight into an equivalent damage time. If an engine flies through a very harsh mission profile, the clock effectively speeds up. One hour of severe flying might count as more than one hour of damage. By applying this correction, they found that the engine was aging faster than the standard flight hours suggested. When they combined this faster aging rate with their new method of grouping related failures, they got a much clearer picture of the engine's true reliability.

The results showed that the old way of calculating safety, which treated every failure mode as independent and ignored the harshness of the flight, gave a misleadingly low safety score. It counted the same risks multiple times. The new method, which grouped the related failures, showed that the engine was actually safer than the old calculation suggested because it stopped double-counting the risks. However, when they added the correction for how fast the engine was actually aging, the safety score dropped again. This drop was not because the engine was worse, but because the calculation finally accounted for the fact that the engine was working harder than the simple flight hours indicated. The final, most accurate estimate came from using both methods together: grouping the linked failures to avoid double-counting, while also speeding up the clock to reflect the true intensity of the damage.

This study does not claim to have solved every problem in engine safety, nor does it replace the need for detailed physical models of how metal breaks. The researchers used data that had been stripped of sensitive details, and their method relies on average estimates of how fast damage occurs rather than a second-by-second simulation of every vibration. However, the approach offers a powerful new tool for engineers. It provides a way to make sense of complex, interacting failures without getting lost in the math of trying to predict every single possibility. By recognizing that some failures are twins and that time moves faster under stress, this framework helps engineers understand the true health of an engine. It suggests that the most critical thing to watch is the wear at the connection points, as this single factor drives the majority of the risk in the early and middle stages of the engine's life. As the engine gets older, the focus must shift to the cracks forming in the disk slots. This dynamic view allows for maintenance that is not just based on the calendar, but on the actual story the engine is telling about its own wear and tear.