Tunnel boring machine (TBM) tunneling data have been extensively collected to utilize TBM information technology by analyzing and mining the data for achieving a safe and efficient TBM tunneling. Feature extraction of big data could reduce the complexity for problems, but conventional indexes based on feature extraction, such as field penetration index (FPI), specific penetration (SP), and boreability index (BI), have some disadvantages. Thus, we present novel boring indexes derived from tunneling data in the Yinchao TBM project. Linear thrust-penetration and torque-penetration relationships in filtered ascending sections (
p
≥ 2 mm/r) are proposed using statistical features and through physical mechanism analysis of parameters in the TBM cyclic tunneling process. Boring indexes, such as normal boring difficulty index, initial rock mass fragmentation difficulty index, and tangential boring difficulty index, are defined using the coefficients of the linear thrust-penetration and torque-penetration relationships. Subsequently, the defined boring indexes are verified using performance prediction of 291 cyclic tunneling processes. Finally, a preliminary application of support measure suggestions is conducted using the statistical features of boring indexes, where certain criteria are proposed and verified. The results showed that the criterion of boring indexes for support measure suggestions could achieve a reasonable confirmation, potentially providing quantitative quotas for support measure suggestions in the subsequent construction process.
Advanced prediction of tunnel boring machine performance based on big data
