The heart of the problem lies in a quantity engineers call the crack growth rate, abbreviated da/dN, which describes how fast a crack lengthens with each loading cycle. Paired with the crack growth threshold, the stress intensity range below which cracks essentially refuse to grow, these numbers feed directly into models of structural degradation. The trouble is that measuring them properly has always been awkward. The standard protocol, laid out in the ASTM E647 standard, requires small, precisely shaped specimens that fit inside a laboratory fatigue testing machine. But the upper main bearing wall of an HT300 cast iron diesel engine block is large, geometrically complex, and heavily loaded. You cannot simply bolt a whole engine block into a servo-hydraulic testing rig and expect clean data.

Engineers have tried two workarounds, and both fall short. The first is to cut small pieces out of the component and machine them into standard single-edge notched bend specimens, known as SENB specimens, which are slender bars with a notch machined into one side. These tests capture the material properties of the cast iron but completely ignore the geometry and load distribution of the real part, which turn out to matter enormously. The second approach is online monitoring, attaching high-precision sensors to the running component to watch cracks grow in service. That sounds ideal, but the equipment is expensive, difficult to install on large complex castings, and can only observe the surface or localized regions, missing micro-cracks and non-uniform stress fields hidden inside the structure. Neither method delivers an accurate da/dN model for an actual complex component.

The Chinese research team took a different route: build a scaled replica of the bearing wall that preserves the physics that matters. Using the component scaling method, they first performed a scaling analysis to ensure that the stress field inside the part would be identical before and after shrinking. At a ratio of 1:7, the massive bearing wall became a model small enough to clamp into an MTS 307 servo-hydraulic fatigue testing machine. Crucially, the scaled models were not fabricated from fresh material. Sixteen scaled models and SENB specimens were cut by wire electrical discharge machining directly from actual upper main bearing walls, and the six pairs with the most similar microstructural distributions were selected for repeated testing. That choice minimized the risk that differences in graphite flake distribution or pearlite spacing, rather than geometry, would skew the comparison.

The test campaign followed ASTM procedures with a stress ratio of 0.1 and a sinusoidal loading waveform at 10 hertz. For the crack growth rate tests, loads were applied in steps matched to the stress intensity factor: 6 megapascals when the stress intensity ranged between 6 and 10 MPa·m^1/2, 10 megapascals between 10 and 15, and 20 megapascals between 15 and 20. Crack length was monitored continuously, producing streams of paired data linking cycle count to crack length, which the team fitted with least-squares polynomial curves and differentiated to extract da/dN. Threshold tests used the standard load-decreasing method, starting from a high cyclic load and stepping down until the crack effectively stopped growing. The shape factor for the scaled model, the term that translates geometry into stress intensity, was computed using a complex curve establishment method developed by Sijia Ren and colleagues.

The results exposed how badly traditional specimens mislead. Across the full crack propagation stage and the near-threshold region, the da/dN curves of the scaled models sat significantly below those of the SENB specimens, while the threshold values of the scaled models were notably higher, averaging 7.2 MPa·m^1/2 against 6.2 for the SENB bars. In plain terms, the real component-shaped geometry resists cracking better than the lab bar made of the same material. The researchers attribute this to the tortuous crack path forced by the complex geometry and load distribution: as cracks wander around structural features, they consume more energy per unit of advance, slowing growth. Local strengthening effects from the complex structure and discontinuities in the material microstructure further bolster the component’s resistance, a phenomenon that flat standard specimens simply cannot reproduce.

With accurate crack growth data in hand, the team built a refined da/dN model based on the Trantina–Johnson formulation, which explicitly incorporates the crack growth threshold and handles variable-amplitude loading realistically by ignoring load cycles below the threshold. They then constructed two structural degradation models for the bearing wall, one fed by the scaled-model data and one by the SENB data. Degradation was framed through Miner’s linear cumulative damage theory and modeled mathematically as a Wiener process, a stochastic framework that treats degradation as a steady drift with random fluctuations described by Brownian motion. From the fitted model, the researchers derived probability density and reliability functions, and computed the expected remaining life of the component over time, yielding a degradation curve that charts how strength and stiffness decline as loading cycles accumulate.

Then came the moment of truth: full-scale fatigue tests on real cylinder block bearing walls. Using a hydraulic pulse-type fatigue testing rig developed at Zhejiang University and sinusoidal loading at 30 hertz across three typical working conditions, the team recorded actual fatigue lives and plotted the genuine degradation behavior. Judged by four statistical metrics, root mean square error, coefficient of determination, mean absolute relative error, and mean absolute error, the degradation model built from the scaled model consistently tracked the experimental curves more closely than the SENB-based model. For example, the mean absolute relative error dropped from about 0.0150 to 0.0118 for one bearing wall and from 0.0132 to 0.0102 for another. Averaged fitted parameters told the same story: the scaled model yielded a coefficient C of 1.11×10^-8 and exponent n of 2.76, versus 1.3×10^-8 and 3.03 for the SENB specimens, confirming slower predicted crack growth in the true geometry.

Fractography added a microscopic epilogue. After cleaning the fracture surfaces ultrasonically, the researchers examined them under a scanning electron microscope and, in parallel, prepared metallographic sections ground with silicon carbide paper up to 2000 grit. Macroscopic inspection revealed flat fracture surfaces, coarsened grains near the fracture zone, pronounced flake graphite segregation, and widened pearlite interlamellar spacing, hallmarks of final instantaneous fracture. At points 3, 6, and 18 millimeters from the crack initiation site, the surfaces were dominated by dimples, the cup-like voids that form when ductile metal stretches and tears. The verdict: the bearing wall fails primarily through ductile fracture dominated by dimple rupture, a characterization that grounds future degradation monitoring in the actual failure mechanism rather than an assumed one.

The implications reach well beyond one casting in one engine family. Because the scaling approach preserves the stress field of the original component, it offers a general recipe for extracting laboratory-grade fatigue data from parts too large or too intricate to test directly, a recipe that could extend to aerospace structures, wind turbine components, and railway axles. More accurate degradation models mean maintenance can be scheduled on evidence rather than conservative guesswork, keeping trucks, ships, and aircraft in service longer while reducing the risk of sudden structural failure. The authors note the method is intended to support maintenance strategies for critical components and improve the utilization efficiency of transportation systems. For a field that has long accepted the compromise of testing coupons instead of components, a seven-times-smaller replica that tells the truth about how metal actually fails is a quietly transformative result.

Subject of Research: Prediction of fatigue-driven structural degradation in diesel engine cylinder block main bearing walls using component scaling and crack growth modeling

Article Title: Accurate prediction of structural degradation in diesel engine cylinder blocks based on component scaling methods

Article References: Huang, W., Xu, C., Liu, M., & Fu, Y. (2025). Accurate prediction of structural degradation in diesel engine cylinder blocks based on component scaling methods. Frontiers in Aerospace Engineering, 4, Article 1604213. https://doi.org/10.3389/fpace.2025.1604213

Image Credits: AI Generated

DOI: 10.3389/fpace.2025.1604213

Keywords: diesel engine, cylinder block, fatigue, crack growth rate, component scaling, structural degradation, cast iron, SENB specimens, Wiener process, fractography, remaining useful life, main bearing wall