Automating the use of learning curve models in construction task duration estimates

Standard scheduling tools for construction projects with repetitive tasks assume that labor productivity remains constant throughout the project lifetime. None of these tools accommodate the dynamics of learning throughout the project. This paper introduces a tool that uses nonlinear optimization to integrate learning curve concepts into task duration estimates for construction project scheduling. The tool, featuring a graphical user interface, mines past data to select the most appropriate learning model from a suite of existing models. Testing the tool on data obtained from five published case studies with varying sizes and locations suggests that the tool offers accurate estimates for task completion times, even when the size or quality of the input data is minimal. Directives for use of this tool and an estimate of potential savings in practice are also provided through an example based on real-world data. These savings amounted to 28% of the overall labor costs within the real-world project. The contribution of this paper is a tool to estimate construction task durations in such a way that learning is incorporated. The tool uses nonlinear optimization to select and calibrate the best learning model making the tool of value to practitioners working across a variety of linear and nonlinear repetitive projects in a range of geographical regions.