Automated Productivity Evaluation of Concreting Works: The Example of Concrete Pillar Production

Abstract

Site schedules are usually developed by the rule of thumb based on the experience of on-site managers. While this approach can be suitable for smaller job sites, it is challenging to make good decisions for larger projects. Planning errors can result in massive delays and increasing costs. Significant improvements in other industries showed that data-driven productivity analysis of past processes advances the planning and execution of current and future projects. However, in the Architecture, Engineering & Construction (AEC) domain, automated productivity analysis of the construction phase has barely been investigated. To overcome this deficiency, this paper presents a first approach for multi-level productivity analysis of shell constructions. We discuss several state-of-the-art vision-based technologies that serve as a foundation for large-scale evaluation of the progress on a construction site. A complete pipeline is introduced that uses different types of neural networks to extract productivity information from images at various levels of detail. The proposed workflow is demonstrated for the construction process of cast-in-place concrete pillars, implementing the first two layers. Finally, remaining challenges are discussed.