Analyzing abnormal cycles of pilot tube microtunneling through pattern recognition in time-series data of hydraulic pressure

Abnormally long operation cycles of pilot tube microtunneling (PTMT) can cause construction delays, difficulties of project coordination, and equipment maintenances costs because of improper operations. Manually logging and analyzing abnormal operation cycles is tedious and time consuming. This paper presents an approach that enables automated identification and analysis of abnormal cycles of PTMT based on automatically logged PTMT operational data. The data collection instrument is a data logger that automatically records time-series of hydraulic pressures of boring machines from pressure transducers attached to their hydraulic lines. Different PTMT operations (e.g., boring, retracting) result in different patterns in these time-series. The authors developed an approach that automatically recognizes time-series patterns representing operation cycles of three phases of PTMT installation. This approach uses an artificial neural network (ANN) to classify a time-series as belonging to a certain PTMT phase, and then applies an adaptive anomaly detection algorithm (AADA) to clean and split the time-series into operation cycles. The results from a case study show that this automated approach enables users to analyze abnormal cycles of PTMT and gain insights about how to improve PTMT productivity.