Smart robot-enabled remaining useful life prediction and maintenance optimization for complex structures using artificial intelligence and machine learning

Abstract

To replace current legacy inspection/maintenance methods with autonomous real-time health status tracking , the paper proposes a smart robotic system with integrated remaining useful life (RUL) prediction tailored for complex components, structures and systems (CSSs). Capabilities like artificial intelligence (AI)/machine learning (ML) utilizing sensing data along with other monitoring data assist in maintenance optimization. The designed system is based on the state-of-the-art reinforcement learning (RL) and deep learning (DL) framework, which consists of an input, modeling, and decision layer. To achieve better prediction accuracy with higher autonomy, a novel active robot-enabled inspection/maintenance system is deployed in the input layer to collect whole-field infrastructure sensing data and inspect critical CSSs. The deep RL approach is integrated with failure diagnostic and prognostic algorithms to train a risk-informed AI-based agent for controlling the robots. With the data collected from the input layer, the modeling layer first conducts data fusion and predicts RUL of components using an efficient Bayesian convolutional neural network (BCNN) algorithm. In the decision layer, a resilience-driven probabilistic decision-making framework will be developed to control the robot for automatically detecting local damage, e.g. defects, degradation, and recommend mitigation/recovery actions for the health management of infrastructure under uncertainty. The combined layers comprise a AI-risk-driven sensing system (AIRSS) which was tested on an Aero-Propulsion System turbofan engine.