Applicability of Machine Learning Techniques to Forecast Remaining Useful Life on Structural Components

Most of the major machines and structural components are designed for fatigue life and at same time it is important to design structural components for no premature fatigue failure. The performance of major machines and structural components are usually tested in controlled environment but in real life components are subjected to fluctuating loads known as fatigue loads which are common causes of failure. Fatigue cracks are common indicators of potential structural failure, and an early stage of crack initiation phase often goes undetected until noticeable performance degradation or failure to the component occurs resulting in a machine downtime. Early detection of Failure and understanding remaining useful life of a component is increasingly more important to customers as it helps in preventive maintenance by timely replacement of a component. This would also result in reducing costs by forecasting time to failure. With recent advancement in science, available data can be analyzed easily to build analytical models and develop a machine-based algorithm to learn from data, identify time to failure and make decisions with less human intervention. This paper aims to research the apparatus and methodology used to detect fatigue cracks and forecast Remaining Useful Life (RUL) on Excavator Structural component.