To Jack or Not to Jack: Inside the Minds of Pipe Jacking Operators

Limited available land is leading to the emerging demand for trenchless methods for the installation of buried pipelines. Pipe jacking is one such method. The prediction of jacking forces is crucial in assessing the performance of the pipe jacking process, thus many efforts have been made to predict the jacking forces. Current approaches are based on the negotiated geological conditions, either through empirical or mechanistic approaches. However, such methods are only useful during the planning and design stages. During the construction stage, the microtunnel boring machine (mTBM) operators base their decisions on the operation parameters. This is further complicated when pipe jacking is performed through highly weathered lithology, i.e. ‘soft rocks’. There continues to be a gap in this domain of modelling jacking forces for soft rocks based on operation parameters. In the current study, deep learning techniques were applied to pipe jacking field-measured operation parameters obtained from the Kuching Wastewater Management System. The deep learning models helped to relate the time series input operation parameters to the jacking forces. Subsequently, a study on different moving average periods towards the input features was applied to understand the changes in the prediction performance of the jacking forces using the deep learning model. From the study, the results showed that the cumulative parameters which are the drive length, the cumulative days elapsed, and the cumulative lubricant used have the highest moving average periods. This suggested that along the pipe jacking process, these features would have a significant impact on the jacking forces.