Nature-Inspired Sensor Data Analytics Methods for Simulation Input Modeling
Abstract
The general proliferation of technology including smartphones and sensors present both an opportunity and a challenge for the construction industry. On the one hand, it creates an opportunity for improved efficiency via greater data-driven decision-making, but on the other hand, presence of noise and uncertainty in the captured data (due to the dynamic and intermittent nature of construction processes), pose significant hurdles to widespread adoption and utilization. Moreover, there is a dearth of domain-specific research concerning the systematic treatment and elimination of such noise. This can have significant impact in the output. As the chaos theory explains, initial noise (even in small portions) can prove to be detrimental to the overall efficacy of a system due to the volatility induced by propagation of such noise through the system. Most natural systems, however, maintain stability and improve over time. In particular, species have improved with evolution, and complex biological information have been preserved and transferred through DNA coding and utilized effectively across generations. Thus, the hypothesis of this research is that methodologies based on principles of natural phenomena can enable reliability of the collected sensor data. This hypothesis is validated by processing data through genetic algorithms (GA), sequence alignment (SA), and multi-dimensional sequence alignment (MSA), all rooted in nature. Processed data is then used to create key input for simulation models describing the real system. Findings of this work is sought to provide project managers and stakeholders with better insights into the nature of crew activities and interactions, and help select the most effective combination of resources while reducing the amount and frequency of rework.
Subject
Construction simulationactivity sequence
chaos theory
genetic algorithm
fuzzy data
sensor networks
sequence alignment
construction activity recognition
multi-dimensional sequence alignment
Citation
Shrestha, Prabhat (2018). Nature-Inspired Sensor Data Analytics Methods for Simulation Input Modeling. Master's thesis, Texas A & M University. Available electronically from https : / /hdl .handle .net /1969 .1 /173517.