Applying radial basis function neural networks to estimate next-cycle production rates in tunnelling construction

Sze Chun Lau, Ming Lu, Samuel Ariaratnam

Research output: Contribution to journalArticlepeer-review

24 Scopus citations


To better cater to tunnel construction productivity studies, the present research extends time series analysis by accounting for additional critical state variables of the tunnelling construction system which represent geological factors and operation delay factors. Those state variables are readily assessed at the end of each tunnelling cycle or can be easily obtained from the actual data recorded in current data collection systems. Radial basis function (RBF) neural networks (NN) provide the accuracy, flexibility and efficiency in mapping complex non-linear relationships between system states and system outputs at consecutive time events. Using data obtained from a tunnel project in Hong Kong, a case study of applying the RBF-based time series analysis for estimating next-cycle production rates was conducted. Identification of those additional state variables for rock tunnel construction by the " drill and blast" method is elaborated. The RBF NN model is retrained at the end of each cycle with the most recent data added to the NN training set. The updated RBF NN model is then used to assist tunnel engineers in estimating the production rate on the immediately following cycle.

Original languageEnglish (US)
Pages (from-to)357-365
Number of pages9
JournalTunnelling and Underground Space Technology
Issue number4
StatePublished - Jul 2010


  • Artificial Intelligence
  • Construction
  • Forecast
  • Productivity
  • Radial basis function neural networks
  • Tunnelling

ASJC Scopus subject areas

  • Building and Construction
  • Geotechnical Engineering and Engineering Geology


Dive into the research topics of 'Applying radial basis function neural networks to estimate next-cycle production rates in tunnelling construction'. Together they form a unique fingerprint.

Cite this