Research Projects

In industrial applications, machine-components are subjected to numerous external loads and environments during their service-life. Over a prolonged period, the structural integrity of the components degrade. This is primarily driven by inherent microscopic defects which are always present in any manufactured part. At initial stages, these material-defects are harmless. However, during continued usage of components, they amalgamate under external effects and form macroscopically visible cracks, which ultimately result in fracture. Although this process is unavoidable, it is necessary for any engineer to pre-assess the service-life of a component. This requires accurate estimation of the load-capacity and crack-size which a material can sustain. Therefore, engineers seek prior experiment testing of a component under mechanical loads and environmental conditions. However, experiments are associated with significant costs incurred by set-up and maintenance of testing machines. Moreover, they heavily consume power and oil, which adversely affect the environment. As such, a much cleaner and greener way for pre-assessing any material-failure is through a physics-based virtual computational analysis under external conditions. It requires only a computing station, which hugely reduces costs and resources required. Moreover, simulations can incorporate any complexity and can be scaled to any component-size. However, classical computational fracture analysis has mathematical instabilities and numerically unsolvable singularities, as observed by several researchers. Recently, a computational field called ‘Continuum Damage Mechanics’ has emerged to avoid such difficulties. It quantifies a damage variable to capture the loss of material-strength and stiffness. But, such computations often involve large computer-run-times, often delaying subsequent tasks. A promising way to reduce this time is the combined use of physics-informed artificial intelligence and damage-mechanics for assessing the governing material-behaviour. This technique emerged in 2019, with additional advancements in 2023. It is yet to formulated and tested for multiphysics-coupled failures of solid components. Accordingly, the primary objectives of the present research, using this combined approach are devoted to: (a) Thermo-mechanical failures (b) Chemo-mechanical failures (c) Thermo-chemo-mechanical failures. A virtual computational environment would be developed using localizing gradient-damage model and physics-driven artificial neural networks. It would serve as an effective and low-cost alternative to experimental tests and would consume significantly low power and resources. It would be a user-friendly simulator for engineers to gain prior knowledge of industrial load-capacities and defect-sizes. This is expected to speed-up industrial establishments with accurate pre-assessment of components, thus preventing any sudden failures under standard operating conditions.