Panuwat Soranansri's thesis defense

The aims of this PhD thesis were to characterize the effectiveness of surface coatings developed to combat the material transfer problems encountered during hot forming of aluminum, and to study these transfer mechanisms. The material used was an AA 6082-T6 aluminum alloy, widely employed in the manufacture of automotive components.The Hot Compression-Translation Test (WHUST) was chosen as the main tribometer for this study.

In order to precisely control test temperatures, a miniaturized WHUST device was designed to be integrated into the heating chamber of the Bruker UMT TriboLab platform. Preliminary tests with this new device showed a significant piling up of material in front of the contactor. New analytical equations have therefore been developed to identify the Coulomb coefficient of friction (COF) and the friction factor (Tresca's law) taking into account this stacking of material.

The WHUST was then used to evaluate the tribological performance of three commercial PVD coatings: an AlCrN, a TiAlN and an Arc-DLC. Experiments were conducted without lubricant, at temperatures ranging from 300°C to 500°C, under contact pressures between 40 and 100 MPa, with a sliding speed equal to 0.5 mm/s. The results showed that the Arc-DLC coating was more effective than the AlCrN and TiAlN coatings in mitigating aluminum transfer problems. In particular, the Arc-DLC coating caused less adhesion and less aluminum transfer, especially at the onset of sliding.

These results were confirmed by tests under higher contact pressures, carried out using the hot T forging test (HVGCT).
In the second part of this thesis, the Arc-DLC coating was selected to study in detail the aluminum transfer mechanisms on forming tools.

Tests were carried out with a short sliding distance (2 mm) to examine the initial stages of aluminum transfer, while tests with a sliding distance of 38 mm were used to study the evolution of transfer. Experiments were conducted at the same test temperatures (300-500°C), with two different sliding speeds, 0.5 mm/s and 5.0 mm/s, and always without lubricant.

Surface topographies and SEM images taken along the friction track showed that aluminium transfer occurs in two main stages: an initial phase mainly due to mechanical tilling, followed by a growth phase dominated, depending on temperatures and sliding speeds, by mechanical tilling or adhesion.

In the final part of this thesis, machine learning (ML) was used to study aluminum transfer mechanisms. Surface topographies and SEM images taken along the friction track were analyzed. They were classified using five simple machine learning algorithms and a customized convolutional neural network (CNN) architecture. Both ML applied to topographic data and CNN applied to SEM images were shown to identify wear modes accurately
.