Monitoring the condition of city bus engines by analysing used oil using PCA and K-Means clustering

Why Bus Engine Oil Tells an Important Story

City buses work hard: they creep through traffic, sit idling at stops, and run long hours every day. All that effort takes a toll on their engines, and the oil that keeps them moving quietly records everything that happens inside. This study shows how reading the “fingerprints” left in used engine oil can reveal which buses are healthy, which are wearing in, and which may be drifting toward trouble—far more accurately than just watching the odometer.

From Simple Mileage Counts to Smarter Care

For decades, most fleets have decided when to change oil by counting kilometers. But urban driving is messy. Two buses can travel the same distance while living very different lives: one glides along open roads, another crawls through stop‑and‑go traffic, idling at red lights and bus stops. Earlier research has shown that under these conditions, mileage alone often fails to predict how “tired” the oil really is. This work tackles that problem by treating oil not as a generic consumable, but as a rich source of data about how each engine is actually aging.

Turning Used Oil into Data

The authors analyzed 165 samples of used oil from a municipal bus fleet using the same kind of infrared light scanners common in industrial labs. They focused in detail on one widely used oil, Lukoil 10W40, to reduce confusion from different formulations. For each sample they measured how thick the oil was at two temperatures, how acidic or basic it had become, how much it had oxidized or reacted with combustion gases, how much soot and fuel had crept in, and how many microscopic metal particles—from iron to copper and lead—were floating in the fluid. They also tracked the levels of protective additives, such as zinc, phosphorus, and calcium, that slowly get used up over time.

Watching Oil Age Inside Real Engines

Looking at one variable at a time confirmed a familiar pattern: as buses remained on the same oil fill, the fluid gradually thickened, turned more acidic, and picked up more sulfur‑ and nitrogen‑based by‑products, while its ability to neutralize acids declined. Yet some expected relationships broke down under real‑world conditions. Surprisingly, oils that had been in service longer did not always show more wear metals; in this fleet they even showed a weak negative link with distance on oil. The likely reason is human behavior: buses in better mechanical shape may be allowed to stretch oil changes, while engines that raise concern are serviced earlier, reshaping the simple picture that “more kilometers means more metal in the oil.”

Finding Hidden Groups of Engine Behavior

To uncover deeper patterns, the team treated all these measurements together using two statistical techniques that can find structure in complex data. First, they used a method that condenses dozens of related measurements into a handful of underlying “axes” describing overall chemical aging, metal wear, and additive loss. Then they fed those condensed descriptions into a clustering method that groups similar samples. Out of the jumble emerged four clear profiles: a group of urban buses that suffer from soot and chemical aging linked to stop‑and‑go idling; a group of brand‑new engines in their break‑in phase shedding unusually high amounts of copper and other metals; a group of mature engines operating in a relatively stable, well‑maintained regime; and a lone outlier bus whose oil showed extreme contamination and additive imbalance, matching a known oil‑leak problem.

Translating Patterns into Maintenance Decisions

The researchers went a step further by training a simple machine‑learning model to recognize these groups from the oil measurements alone and to highlight which measurements mattered most. This confirmed that what truly separates the profiles is not how far the bus has traveled, but the chemical state of the oil—signatures of oxidation, acidity, soot, metals, and the health of key additives. Together, these findings suggest that fleets can safely lengthen oil‑change intervals for older, stable engines, shorten them for new engines during their delicate early life, and quickly flag buses showing unusual contamination, even when basic numbers like viscosity still seem acceptable.

What This Means for Everyday Riders

For riders, this kind of oil‑based health check is invisible, but its effects are not. By listening to what used oil reveals, transit agencies can move from fixed calendars to condition‑based maintenance, catching problems earlier, avoiding unnecessary oil changes, and extending engine life. The paper’s central message is simple: the dark liquid drained from a bus is more than waste—it is a diagnostic report. Reading it carefully can make city buses more reliable, maintenance more efficient, and the daily commute a bit more dependable.

Citation: Duarte, M.O., Margalho, L.M., Gołębiowski, W. et al. Monitoring the condition of city bus engines by analysing used oil using PCA and K-Means clustering. Sci Rep 16, 9392 (2026). https://doi.org/10.1038/s41598-026-39045-x

Keywords: engine oil analysis, bus fleet maintenance, condition-based maintenance, diesel engine health, predictive maintenance