TL;DR
This study investigates how surface roughness (Sa) affects the friction and wear of GCr15 bearing steel under varying loads with oil lubrication. They found that increasing roughness generally increases friction, and wear initially decreases then increases with roughness, with an optimal Sa of 0.5 µm for minimal wear. A backpropagation neural network (BPNN) was developed to predict wear performance, achieving an average prediction error of 10.64%.
What Problem Does This Paper Address?
Bearings are critical components in many industries, and their performance, including friction, wear, and lifespan, is heavily influenced by surface quality. Surface roughness, in particular, plays a significant role in lubrication effectiveness, stress distribution, and ultimately, bearing durability. While the general importance of surface roughness is known, there's a need for a deeper understanding of its specific impact on GCr15 bearing steel, a widely used material, especially under varying operational loads. Furthermore, predicting this wear behavior efficiently is challenging due to the complex interplay of factors. This paper aims to fill this gap by systematically analyzing the effects of surface roughness and load on GCr15 steel's tribological performance and by developing a predictive model.
Methods
The researchers conducted experimental tests using GCr15 steel specimens with controlled surface roughness values (Sa ranging from 0.01 µm to 1.5 µm). These specimens were prepared through various grinding and polishing processes.
Key experimental techniques included:
- Surface Profilometry: A 3D surface profiler was used to measure surface roughness (Sa) and analyze wear track morphology.
- Microscopy (SEM/EDS): Scanning Electron Microscopy (SEM) was employed to examine the wear surfaces and identify wear mechanisms, while Energy Dispersive Spectroscopy (EDS) analyzed the elemental composition.
- Microhardness Testing: A microhardness tester was used to assess the material's hardness.
- Wettability Measurement: Contact angle measurements were performed using PAO-6 base oil to understand how surface roughness affects lubricant adhesion.
- Tribological Testing: A ball-on-disk tribometer (UMT-2) was used to measure friction coefficients and wear under controlled loads (15 N, 25 N, 35 N) and with adequate oil lubrication.
- Wear Area Calculation: Wear area was quantified by integrating the wear track profiles, a method illustrated schematically.
A significant methodological contribution is the development of a Back Propagation Neural Network (BPNN) model. This model was trained using experimental data (Sa, load, and wear area) to predict the wear performance of GCr15 steel. The BPNN utilized a Sigmoid activation function and was optimized to have five hidden layer neurons for minimal error.
Datasets
The study utilized a custom-generated dataset for training and validation of the BPNN model. This dataset consisted of 12 data points, each representing a combination of specific surface roughness (Sa) values (0.01, 0.1, 0.5, 1, and 1.5 µm) and applied loads (15 N, 25 N, and 35 N), along with the corresponding experimentally measured wear area. This dataset is not a publicly released benchmark but rather a specific collection of experimental results generated for this research.
Key Insights and Ideas
- Roughness-Friction Relationship: The study confirms a direct correlation between increasing surface roughness (Sa) and a higher friction coefficient under oil-lubricated conditions. This suggests that smoother surfaces are generally better for reducing friction.
- Optimal Roughness for Wear: There's a "sweet spot" for surface roughness in terms of wear. While very smooth surfaces might experience more direct contact and wear, and very rough surfaces can lead to increased stress, a moderate roughness (Sa = 0.5 µm) was found to yield the minimum wear area and optimal wear resistance. This is attributed to a balance between lubricant film stability and the ability to trap wear debris.
- Load's Impact on Wear: Higher applied loads consistently lead to larger wear areas, indicating that increased stress exacerbates material removal.
- Wear Mechanism Dominance: Under oil lubrication, abrasive wear appears to be the dominant mechanism for GCr15 steel, especially at higher loads. However, adhesion also plays a role, particularly at lower roughness values.
- BPNN as a Predictive Tool: The development and validation of the BPNN model demonstrate its potential as an efficient and cost-effective tool for predicting bearing wear performance across various surface roughness and load conditions, reducing the need for extensive experimental testing.
Key Results
- Friction Coefficient: As Sa increased from 0.01 µm to 1.5 µm, the average friction coefficient generally increased. For example, at 15 N load, average friction values ranged from 0.080 µm (Sa=0.01) to 0.120 µm (Sa=1.5). At higher loads (25 N and 35 N), the friction coefficient became less sensitive to roughness, hovering around 0.09 for most Sa values except the highest (1.5 µm).
- Wear Area: The wear area exhibited a non-monotonic trend with increasing Sa. At 15 N load, the minimum wear area was 763 µm² at Sa = 0.5 µm, while the maximum was 1590 µm² at Sa = 1.5 µm. Similar trends were observed at 25 N and 35 N loads, with Sa = 0.5 µm consistently showing low wear.
- BPNN Model Accuracy: The developed BPNN model achieved an average prediction error of 10.64% for wear area. Independent validation tests showed specific errors of 7.19% and 14.1% for tested conditions.
Limitations and Future Work
- Lubrication Conditions: The study focused on adequate oil lubrication. The behavior under different lubrication regimes (e.g., boundary lubrication, dry friction) might differ.
- Material Specificity: The findings are specific to GCr15 bearing steel. Further research could explore if these trends hold for other bearing materials.
- Load Range: While multiple loads were tested, exploring a wider range of loads, including very low and extremely high loads, could provide a more comprehensive understanding.
- Environmental Factors: The study did not explicitly consider environmental factors like temperature or humidity, which can influence tribological performance.
The authors suggest that the BPNN model provides valuable insights for bearing design, failure analysis, and fault diagnosis. Future work could involve expanding the BPNN model to predict other tribological parameters (e.g., friction coefficient) or incorporating more complex wear mechanisms and environmental factors.
Summary updated