Bolt Looseness Detection in Flanged Pipes Using Parametric Modeling

Detecting bolt looseness in flanged pipes at an early stage is critical for ensuring safety in the oil and gas industry. Failure to identify such looseness can lead to severe incidents such as leaks and explosions. One effective indirect method for detecting looseness is through vibration analysis of the structures connected by bolted joints. Changes in vibration parameters can indicate looseness, which affects the bolted joints’ structural stiffness. This study presents a novel parametric modeling algorithm for detecting bolt looseness in flanged pipes. Autoregressive (AR) model parameters serve as the feature vector for a Mahalanobis distance–based indicator, facilitating accurate looseness detection. Validation was conducted using AR and time-varying autoregressive (TAR) models adapted to the stationary and nonstationary vibration signals of flanged pipes, respectively. The structure was excited using white noise (stationary state) and a moving mass inside the pipe (nonstationary state) to ensure practical applicability. The results demonstrate the method’s effectiveness in detecting flange looseness at early stages using an output-only approach.