Comparative Analysis of Learning Curve Models on Construction Productivity of Diaphragm Wall and Pile

Abstract

Abstract In the analysis of construction operations, Learning Curves (LC) are considered one of the most important factors that determines the on-site variation in the productivity, which is usually considered in the construction projects during the estimation and planning stage. This research attempts to assess the suitability of LC models for the analysis of the learning phenomenon using productivity data for fairly complicated construction operations concerning the Diaphragm Wall and Pile Construction process from large-scale construction projects. In this study, the role of different LC models (i.e., Wright or Straight Line, Quadratic, Cubic, Knecht or Combined Exponential Log-linear, Stanford B) is investigated by the comparison of their outcomes through the utilization of cumulative productivity data of the activities involved in the Diaphragm Wall and Pile Construction process. The two main research objectives are (i) the investigation of the model which is the best bit for the historical productivity data of the completed construction activities (ii) an endeavour is formed to work out which model predicts the future performance better. The best suited LC model is predicted based on the least deviation from the yielded results of each model with respect to the actual construction data. Analysis of the cumulative average productivity data predicted that the Knecht or Combined Exponential Log-linear Model best fits both the complete Diaphragm Wall and Pile Construction Process in both the cases of fitting Historical data and in predicting future performance.