Predictive Modelling of Tribological Systems using Movable Cellular Automata

In the science of tribology, where there is an enormous degree of uncertainty, mathematical models that convey state-of-the-art scientific knowledge are invaluable tools for unveiling the underlying phenomena. A well-structured modelling framework that guarantees a connection between mathematical representations and experimental observations, can help in the systematic identification of the most realistic hypotheses among a pool of possibilities. This thesis is concerned with identifying the most appropriate computational model for the prediction of friction and wear in tribological applications, and the development of a predictive model and simulation tool based on the identified method. Accordingly, a thorough review of the literature has been conducted to find the most appropriate approach for predicting friction and wear using computer simulations, with the multi-scale approach in mind. It was concluded that the Movable Cellular Automata (MCA) method is the most suitable method for multi-scale modelling of tribological systems. It has been established from the state-of-the-art review in Chapter 2 of this thesis, that it is essential to be able to model continuous as well as discontinuous behaviour of materials on a range of scales from atomistic to micro scales to be able to simulate the first-bodies and third body simultaneously (also known as a multi-body) in a tribological system. This can only be done using a multi-scale particle-based method because continuum methods such as FEM are none-predictive and are not capable of describing the discontinuous nature of materials on the micro scale. The most important and well-known particle-based methods are molecular dynamics (MD) and the discrete element methods (DEM). Although MD has been widely used to simulate elastic and plastic deformation of materials, it is limited to the atomistic and nanoscales and cannot be used to simulate materials on the macro-scale. On the other hand, DEM is capable of simulating materials on the meso/micro scales and has been expanded since the algorithm was first proposed by Cundall and Strack, in 1979 and adopted by a number of scientific and engineering disciplines. However, it is limited to the simulation of granular materials and elastic brittle solid materials due to its contact configurations and laws. Even with the use of bond models to simulate cohesive and plastic materials, it shows major limitations with parametric estimations and validation against experimental results because its contact laws use parameters that cannot be directly obtained from the material properties or from experiments. The MCA method solves these problems using a hybrid technique, combining advantages of the classical cellular automata method and molecular dynamics and forming a model for simulating elasticity, plasticity and fracture in ductile consolidated materials. It covers both the meso and micro scales, and can even “theoretically” be used on the nano scale if the simulation tool is computationally powerful enough. A distinguishing feature of the MCA method is the description of interaction of forces between automata in terms of stress tensor components. This way a direct relationship between the MCA model parameters of particle interactions and tensor parameters of material constitutive law is established. This makes it possible to directly simulate materials and to implement different models and criteria of elasticity, plasticity and fracture, and describe elastic-plastic deformation using the theory of plastic flow. Hence, in MCA there is no need for parametric fitting because all model parameters can be directly obtained from the material mechanical properties. To model surfaces in contact and friction behaviour using MCA, the particle size can be chosen large enough to consider the contacting surface as a rough plane, which is the approach used in all MCA studies of contacting surfaces so far. The other approach is to specify a very small particle size so that it can directly simulate a real surface, which allows for the direct investigation of material behaviour and processes on all three scale levels (atomic, meso and macro) in an explicit form. This has still been proven difficult to do because it is too computationally extensive and only a small area of the contact can be simulated due to the high numbers of particles required to simulate a real solid. Furthermore, until now, no commercial software is available for MCA simulations, only a 2D MCA demo-version which was developed by the Laboratory of CAD of Materials at the Institute of Strength Physics and Materials Science in Tomsk, Russia, in 2005. The developers of the MCA method use their own in-house codes. This thesis presents the successful development of a 3D MCA open-source software for the scientific and tribology communities to use. This was done by implementing the MCA method within the framework of the open-source code LIGGGHTS. It follows the formulations of the 3D elastic-plastic model developed by the authors including Sergey G. Psakhie, Valentin L. Popov, Evgeny V. Shilko, and the external supervisor on this thesis Alexey Yu. Smolin, which has been successfully implemented in the open-source code LIGGGHTS. Details of the mathematical formulations can be found in [1]–[3], and section 3.5 of this thesis. The MCA model has been successfully implemented to simulate ductile consolidated materials. Specifically, new interaction laws were implemented, as well as features related to particle packing, particle interaction forces, bonding of particles, and others. The model has also been successfully verified, validated, and used in simulating indentation. The validation against experimental results showed that using the developed model, correct material mechanical response can be simulated using direct macroscopic mechanical material properties. The implemented code still shows limitations in terms of computational capacity because the parallelization of the code has not been completely implemented yet. Nevertheless, this thesis extends the capabilities of LIGGGHTS software to provide an open-source tool for using the MCA method to simulate solid material deformation behaviour. It also significantly increases the potential of using MCA in an HPC environment, producing results otherwise difficult to obtain.