Time series analysis of hydraulic data for automated productivity monitoring of pilot tube microtunneling

Monitoring and controlling construction productivity of pilot tube microtunneling (PTMT) are important in reducing delays of tunneling projects and in decreasing project costs. Collecting reliable and detailed productivity data in the field for effective PTMT productivity analysis, however, is challenging. Sensors attached to hydraulic devices of PTMT machines can automatically record time series of a boring machine's hydraulic forces during operations. These time series show cyclic patterns corresponding to cyclic PTMT operations in three stages of PTMT: (1) pilot tube installation, (2) casing installation, and (3) product pipe installation. Analyzing these time series manually for detailed productivity analysis is possible, but such manual analysis becomes tedious and error-prone. This paper presents a knowledge-based and adaptive time series analysis approach that can automatically detect cycles of construction activities from time series data and thus achieve real-time PTMT productivity analyses. This approach can tolerate noises in time series data collected in real PTMT projects and thus can adaptively adjust its parameters according to the characteristics of input data. Such adaptive capability enables engineers to apply this method to various time series collected in different PTMT sessions. The testing results in a PTMT project in Wisconsin showed that the proposed approach achieves 95% or better precision and recall on data collected during seven different sessions of PTMT construction. These data covered three phases of PTMT that use three different machines and pipeline sections under different environmental conditions in order to validate the developed algorithms. Productivity analyses results revealed that productivities on some sections almost doubled those on others and that eliminating anomalous cycles could result in up to 40% improvement in overall productivity.