Deterministic Elastic-Plastic Modelling of Rough Surface Contact including Spectral Interpolation and Comparison to Theoretical Models
Type of DegreeMaster's Thesis
MetadataShow full item record
Rough surfaces can be seen everywhere in our world, such as on table surfaces, roads, or cellphone screens. The interactions of rough surfaces are essential problems because they govern friction, wear, contact resistance and other phenomenon. A finite element elastic-plastic rough surface contact model is developed in this thesis. Multiple surface areas with different roughnesses are measured by an optical profilometer, which are selected from a S-22 Microfinish Comparator. A small square area with a side length of 31 micrometers is extracted from the measured rough surfaces. The area is populated with greater resolution by a novel method called spectral interpolation. The nodes covering the area are increased from 32 by 32 to 63 by 63 at the first interpolation, and to 125 by 125 at the second interpolation while the resolution decreases from one micrometer to 0.5 micrometer and 0.25 micrometer. Visual inspection suggests that asperities of real surfaces are usually not sharp peaks but more rounded or curved. The spectral interpolation method allows for sharp peaks and intermediate points to be smoothed by adding additional nodes while keeping the surface spectrum constant. Therefore, the simulation results are presumed to be closer to cases of the real world. Then, the effects of different resolutions on the area-load relationship are studied by finite element simulation results and the results are compared to theoretical models. Additionally, the effects of the tangential modulus and different locations of the contact detection points used in the finite element method are also studied in the thesis. Different values of the tangential modulus (between 2% and 0% of the elastic modulus) result in significant changes of the area-load relationship. Moreover, this work shows that using nodal points as contact detection points is a more reasonable choice than Gauss integration points.