Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


DissertationRafiei_Finali.pdf (5.02 MB)

ETD Abstract Container

Abstract Header

ADVANCED NEURAL NETWORK AND MACHINE LEARNING MODELS FOR CONSTRUCTION, MATERIALS AND STRUCTURAL ENGINEERING

Rafiei, Mohammad Hossein
http://rave.ohiolink.edu/etdc/view?acc_num=osu1480630474535471

Abstract Details

2016, Doctor of Philosophy, Ohio State University, Civil Engineering.
Machine learning (ML) is a core technology in development of intelligent systems. The goal of this research is twofold: first to explore the applications of advanced ML models such as deep belief restricted Boltzmann machine (DRBM) and the enhanced probabilistic neural network (EPNN) in infrastructure engineering (IE); second to develop a new classification algorithm (CA) with applications in IE. The findings of dissertation are as follows: 1) An ML model was developed by integrating a non-mating genetic algorithm and DRBM for estimation of sale prices of real estate units in any given city at the design phase or beginning of the construction. The model can be used by construction companies to gage the sale market before they start a new construction. An effective data structure was presented that takes into account a large number of economic variable/indices. 2) A supervised DRBM was developed for estimation of concrete material properties. Its effectiveness was compared with back propagation neural networks (BPNN) and support vector machine (SVM). The model was tested using 103 concrete test data from the ML repository of the University of California-Irvine. The proposed model provided more accurate results and was computationally more efficient than two other models, BPNN and SVM. 3) An ML-optimization model was developed for concrete mix design through adroit integration of an optimization algorithm (OA) and a CA used a virtual lab. The proposed model can take into account any variations in the properties of the various ingredients. The model was tested using previously collected data and three OAs, and three CAs. The most cost effective solutions are achieved by the combination of neural dynamics classification of Adeli and Park (NDAP) and EPNN. 4) A new supervised CA, neural dynamic classification (NDC), was developed with the goal of a) discovering the most effective feature spaces and b) finding the optimum number of features required for accurate classification using NDAP. The model was compared with three other CAs: probabilistic neural network (PNN), EPNN and SVM using two sets of classification problems. In general, NDC yielded the most accurate classification results followed by EPNN. 5) An earthquake early warning system model was developed for forecasting the earthquake magnitude and its location weeks before occurrence using a combination of a CA and an OA. The role of the CA is to find whether there is an earthquake in a given time period greater than a predefined magnitude threshold and the role of the OA is to find the location of that earthquake. The most accurate forecasts were achieved by the combination of NDC and NDAP. 6) An ML model was developed for health monitoring of large structures through integration of two signal processing techniques, fast Fourier transform and synchrosqueezed wavelet transform, and two ML algorithms, restricted Boltzmann machine and the NDC. The model was validated using experimental data. The NDC algorithm was replaced by PNN, EPNN, and k-nearest neighbor for the sake of comparison. The model with NDC found to be the most accurate model.
Hojjat Adeli, Professor (Advisor)
227 p.
Civil Engineering; Computer Science

Recommended Citations

Citations

  • Rafiei, M. H. (2016). ADVANCED NEURAL NETWORK AND MACHINE LEARNING MODELS FOR CONSTRUCTION, MATERIALS AND STRUCTURAL ENGINEERING [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1480630474535471

    APA Style (7th edition)

  • Rafiei, Mohammad Hossein. ADVANCED NEURAL NETWORK AND MACHINE LEARNING MODELS FOR CONSTRUCTION, MATERIALS AND STRUCTURAL ENGINEERING. 2016. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1480630474535471.

    MLA Style (8th edition)

  • Rafiei, Mohammad Hossein. "ADVANCED NEURAL NETWORK AND MACHINE LEARNING MODELS FOR CONSTRUCTION, MATERIALS AND STRUCTURAL ENGINEERING." Doctoral dissertation, Ohio State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=osu1480630474535471

    Chicago Manual of Style (17th edition)

osu1480630474535471
2,713
© 2016, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.