Characterization of Grey Cast Iron-Steel Tribo-system under Starved Lubrication Conditions : Failure, Failure Detection and Laser Surface Texturing

Reducing friction and wear is essential for building efficient systems with low energy consumption and a long lifetime. Surface texturing is one of the methods to reduce friction and wear, especially in oil-lubricated systems. However, there is still a lack of knowledge about the fundamental aspects involved in the improvement of tribological performance, especially for the mechanical parts sliding under starved lubrication conditions. Moreover, failure in such lubricated contacts is poorly understood and there are little agreements regarding the mechanisms leading to scuffing. The present research focuses on three main objectives. The first is to characterize a cast iron-steel tribo-system under starved lubrication conditions, and to improve its tribological performance by laser surface texturing. All tribo-tests are performed using a flat-on-flat set-up with reciprocal movements. This set-up simulates specific industrial operating conditions of a semi-journal bearing used in a cutting machine. A Design of Experiments (DoE) approach is selected to investigate the effect of the different geometrical micro-texture parameters on the coefficient of friction (COF) and the lifetime of the cast iron samples. An optimum micro-texture geometry leading to a low COF and a long lifetime is determined based on a fractional factorial design. The second goal is to investigate the failure mechanisms of the selected tribo-pair. Failure of the tribo-system is characterized for both textured and un-textured surfaces. Scuffing is found to be the failure mechanism for the un-textured cast iron samples. For the textured cast iron samples, in contrast, two different failure mechanisms are observed depending on the distance between the micro-textures in direction of sliding (DMS). The textured surfaces with a DMS > 3.5 mm fail via the scuffing mechanism, whereas the surfaces with a DMS < 3 mm fail by a different mechanism named the oxidation mechanism. In contrary to the scuffing mechanism, the oxidation failure is found to be more a gradual failure. Based on the experimental observations, two hypotheses are suggested for the scuffing mechanism, and one for the oxidation mechanism. Finally, predicting the lifetime of a tribo-system is of the utmost importance to save costs. The DoE model developed for the lifetime was found to be a weak predictive model due to the sudden nature of scuffing. The accuracy of this model could only be improved by performing large statistics, which is not time and cost effective. Consequently, the last objective of this work is to employ an acoustic emission (AE) technique to detect precisely the surface state of our tribo-system. The application of wavelet packet decomposition, as an advanced signal processing method used for AE signals, is found to be promising for early detection of scuffing. The extension of this work using random forest regression shows the possibility of predicting scuffing 5 min before it occurs.