Advanced deep learning approaches and potentials in the construction sector

RWTH Publications (RWTH Aachen)

Abstract

Artificial intelligence (AI) has reached important milestones in recent years and is continuing its course in unknown directions. A broad spectrum benefits from this trend, including the construction industry. However, it should be noted that AI cannot be considered a universal means of solving all problems. Instead, it encompasses a wide range of algorithms and methods, approaches, and paradigms that can be used to solve specific problems in a targeted manner. This thesis examines the potential of artificial intelligence through the use of various AI models and algorithms in the areas of automation, resource estimation, and non-destructive processes. Based on four practice-relevant use cases, all related to the construction industry, different AI approaches are examined and evaluated. Two of these use cases were evaluated and tested on the Reference Construction Site Campus Melaten in Aachen. Another use case is based on a simulation environment, and the fourth use case is dedicated to the classification of radar imagery. Due to the heterogeneity of the investigated use cases, various metrics are used for assessment and evaluation, enabling a comprehensive analysis of the respective advantages and disadvantages of the AI models used. The results obtained show that approaches from the areas of: Machine Learning, Deep Learning and Reinforcement Learning are powerful tools that are able to solve complex problems in a targeted and efficient manner and can therefore be optimally integrated into various construction processes.

Authors 1

1. Davide Picchi corresponding Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen