Quantitative Effect of Residual Stress on Gear Bending Fatigue

时 可实现较高的寿, 命预测精度 而修正的 S-n 曲线中, Haifeng Hu, Huaiju Liu, Caichao ZhuJournal of Mechanical Engineering 2023 · DOI

This paper, published in the Journal of Mechanical Engineering, provides a quantitative framework for understanding how residual stress—the "hidden" stress left behind by manufacturing—affects the lifespan of high-performance gears. By introducing a specific influence coefficient into standard fatigue models, the researchers have created a way to more accurately predict gear failure and unify life-cycle data across different manufacturing treatments.

TL;DR

Researchers have quantified the impact of compressive residual stress on gear bending fatigue by introducing an influence coefficient ($m_r$) into life-prediction models. They found that an influence coefficient of 0.15 for strain-based models and 0.25 for unified S-N curves provides the highest accuracy, allowing engineers to predict gear life across various surface treatments with a single model.

What Problem Does This Paper Address?

In high-demand applications like wind turbines, high-speed rail, and shield tunneling, gears are pushed to their limits. To prevent "bending fatigue" (where gear teeth snap at the root), manufacturers use treatments like carburizing and shot peening to introduce compressive residual stress (RS).

While the industry knows that RS is "good" for fatigue life, quantifying how much better it makes a gear has been a challenge. Most engineers rely on S-N curves (Stress vs. Number of cycles) that are specific to a single manufacturing process. If you change the shot peening intensity, you need a whole new set of expensive, time-consuming tests. This paper seeks to bridge that gap by creating a mathematical way to account for RS in standard fatigue life equations.

Methods

The researchers combined experimental testing with advanced computer simulations:

  1. Experimental Testing: They used 18CrNiMo7-6 steel gears (a standard high-strength alloy). They tested two groups: one that was carburized and shot-peened (high RS) and a validation group that was only carburized (lower RS). Tests were performed on a ZwickRoell resonance machine to determine the fatigue limit and S-N curves.
  2. Finite Element Analysis (FEA): Using ABAQUS, they built a 2D model of the gear teeth to calculate the precise stress and strain at the tooth root, where cracks typically start.
  3. Model Modification: They took three standard fatigue life criteria—Maximum Principal Strain, Von Mises Equivalent Strain, and Smith-Watson-Topper (S-W-T)—and modified them by adding a residual stress term ($\sigma_r$) multiplied by an influence coefficient ($m_r$).
  4. Unified S-N Curve: They proposed a "modified stress" concept ($\sigma_{Feff} = \sigma_F - n_r \sigma_r$) to see if they could collapse data from gears with different RS states onto a single, universal curve.

Datasets

  • Primary Dataset: Experimental fatigue data from 18CrNiMo7-6 carburized and shot-peened gears (Initial RS: -945 MPa).
  • Validation Dataset: Experimental data from the same gear material but without shot peening (Initial RS: -727 MPa).
  • Standards: Results were compared against the GB 3480 and ISO 6336 standards for gear strength calculation.

Key Insights and Ideas

  • The "Magic" Coefficients: The study identified that residual stress does not act exactly like mean load stress (which would have a coefficient of 1.0). Instead, for the Maximum Principal Strain criterion, the optimal influence coefficient is 0.15. For the simplified engineering S-N curve, it is 0.25.
  • Strain vs. Stress: The Maximum Principal Strain criterion was found to be the most accurate for these gears, outperforming the S-W-T and Von Mises models. This suggests that for surface-hardened gears, the local strain at the root is the primary driver of crack initiation.
  • Predicting the "Where": The simulation accurately predicted that cracks initiate at the 29.76° tangent position of the tooth root, which matched the experimental observation of 29.2°–30.7°.
  • The Unified Curve: Perhaps the most practical insight is that by using the $n_r = 0.25$ coefficient, data from both shot-peened and non-shot-peened gears fell onto the same line. This allows for "rapid assessment"—if you know the RS of a new treatment, you can estimate its fatigue life without running a full battery of new tests.

Key Results

  • Accuracy: Using the optimized $m_r = 0.15$, the life prediction for the validation (non-shot-peened) gears fell within a 3-fold scatter band, which is considered highly accurate in the stochastic field of fatigue testing.
  • Fatigue Limit: Shot peening was shown to significantly increase the fatigue limit; for example, the non-shot-peened gears had a limit of 70 kN, while shot-peened versions could handle significantly higher loads for the same lifespan.
  • Quantification: The paper successfully moved the industry from "shot peening helps" to "shot peening with $X$ MPa of residual stress improves life by exactly $Y$ amount in the model."

Limitations and Future Work

  • Material Specificity: The coefficients (0.15 and 0.25) were derived specifically for 18CrNiMo7-6 steel. The authors acknowledge it is unclear if these values remain optimal for other materials or different heat treatment processes.
  • Empirical Nature: While the models work, the physical reason why the coefficient is 0.15 (rather than a purely theoretical value) remains an area for deeper material science research.
  • RS Relaxation: The model uses the initial residual stress. However, under high loads, residual stress can "relax" or fade over time. Future work could incorporate the decay of RS into the life-prediction equations for even greater precision.

Summary updated

Abstract

摘要: 随着风电、高铁、掘进等高端装备对齿轮功率密度、服役寿命等要求的提高,齿轮的弯曲疲劳问题日益显著。为提升齿轮弯曲疲劳性能,通过渗碳热处理、喷丸强化等工艺为齿轮引入较高的残余压应力已逐渐成为工业界的标配。为揭示残余应力对齿轮弯曲疲劳性能的量化影响,在最大主应变寿命预测准则中引入残余应力影响项,通过弯曲疲劳试验确定最优残余应力影响系数,进而采用新的试验数据验证模型的准确性。基于工程应用出发,引入修正应力的概念统一不同残余应力状态下的齿轮弯曲应力-寿命(S-N)曲线,研究结果显示,最大主应变准则中,残余应力影响系数取值为0.15时,可实现较高的寿命预测精度,而修正的S-N曲线中,最佳残余应力影响系数为0.25。研究成果可用于工程实际中齿轮弯曲疲劳快速评估。
BibTeX
@article{2023quantitative,
  author = {时 可实现较高的寿 and 命预测精度 而修正的 S-N 曲线中 and Haifeng Hu and Huaiju Liu and Caichao Zhu},
  title = {{Quantitative Effect of Residual Stress on Gear Bending Fatigue}},
  journal = {Journal of Mechanical Engineering},
  year = {2023},
  doi = {10.3901/jme.2023.04.053},
}
Figure