Digging deeper: A systematic literature review on data cleaning techniques in tunneling research

In recent years, research into Artificial Intelligence (AI) and machine learning (ML) applications within the tunneling industry has increased significantly. Given that tunnel boring machines (TBM) have a vast number of sensors, it is expected to use the data, for e.g., analysis of tunneling excavation, geology classification, and prediction of various TBM parameters. Despite the increasing application of ML methods to TBM sensor data, data preprocessing is often treated as a secondary or implicit step. In this paper, we conducted a systematic literature review in which we analyzed how researchers preprocessed their TBM sensor data. Our review reveals that a substantial number of studies rely on methods that are incompatible with the temporal and non-stationary nature of TBM sensor data, such as static 3 σ or boxplot-based outlier detection. By systematically analyzing existing preprocessing strategies, this review highlights methodological shortcomings, discusses their implications for prediction and classification tasks, and provides guidance for selecting preprocessing techniques that are consistent with the characteristics of TBM data.