An AI-based asset maintenance system accesses a variety of data sources related to an entity to analyze data regarding one or more damage mechanisms corresponding to the entity thereby identifying and implementing corrective actions that mitigate the effects of the damage mechanisms within the entity. The accessed data is stored using a parameterized data model that represents the entity. A trained parameter model identifies the most significant operating parameters for a given component of the entity for the damage mechanism affecting the component. A projection model is used to perform ‘what-if’ analysis of the most significant operating parameters for determining the instances of minimum and maximum degradation due to the damage mechanism. Corrective actions for mitigating the degradation due to the damage mechanism can be determined based on analysis of the operating parameters and other attributes corresponding to the best and worst case degradation scenarios.