Sensor-based prognostics and structured maintenancepolicies for components with complex degradation

Replacement models;Spare parts inventory;Maintenance policies;Markov decision processes;Reliability;Degradation models

Elwany, Alaa H. ; Industrial and Systems Engineering

University:Georgia Institute of Technology

Department:Industrial and Systems Engineering

关键词: Replacement models;    Spare parts inventory;    Maintenance policies;    Markov decision processes;    Reliability;    Degradation models;   

Others  :  https://smartech.gatech.edu/bitstream/1853/37198/1/Elwany_Alaa_H_200912_Phd.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

We propose a mathematical framework that integrates low-level sensory signals from monitoring engineering systems and their components with high-level decision models for maintenance optimization.Our objective is to derive optimal adaptive maintenance strategies that capitalize on condition monitoring information to update maintenance actions based upon the current state of health of the system.We refer to this sensor-based decision methodology as "sense-and-respond logistics".As a first step, we develop and extend degradation models to compute and periodically update the remaining life distribution of fielded components using in situ degradation signals.Next, we integrate these sensory updated remaining life distributions with maintenance decision models to; (1) determine, in real-time, the optimal time to replace a component such that the lost opportunity costs due to early replacements are minimized and system utilization is increased, and (2) sequentially determine the optimal time to order a spare part such that inventory holding costs are minimized while preventing stock outs.Lastly, we integrate the proposed degradation model with Markov process models to derive structured replacement and spare parts ordering policies.In particular, we show that the optimal maintenance policy for our problem setting is a monotonically non-decreasing control limit type policy.We validate our methodology using real-world data from monitoring a piece of rotating machinery using vibration accelerometers.We also demonstrate that the proposed sense-and-respond decision methodology results in better decisions and reduced costs compared to other traditional approaches.