Estimating Bridge Deterioration Age Using Artificial Neural Networks
| dc.Location | 2017 TA 633 H87 | |
| dc.Supervisor | Professor Abid Abu- Tair | |
| dc.contributor.author | AL HUSSEIN, ASEEL | |
| dc.date.accessioned | 2018-04-25T06:35:57Z | |
| dc.date.available | 2018-04-25T06:35:57Z | |
| dc.date.issued | 2017-09 | |
| dc.description | DISSERTATION WITH DISTINCTION | |
| dc.description.abstract | Deterioration of reinforced concrete bridges is major issue in structural engineering due to the difficulty of estimating or predicting the service life of the bridge. Two types of models were developed to estimate the service life, the deterministic and probabilistic models. Nevertheless, the reliability of these models is questioned since they do not account for the many factors involved. Therefore, for this research artificial neural network is used to estimate the deterioration age for RC bridges based on deterioration data. Historical records of bridges located in London is used to train and test ANN. Feedforward neural network is designed to be able to estimate the deterioration age. ANN inputs are bridge type, member type, exposure, and defects while the target is the defects age. Since there are no standard neural network deterioration models, Design of experiment is conducted to select and monitor the most important parameters that would affect ANN performance. Learning algorithm, Number of hidden layers, number of hidden neurons and Transfer function are the four parameters selected for factorial design. Each factor has low and high-level options making 16 different combinations of neural networks. ANN analysis is run on MATLAB and Mean Square Error (MSE), regression and error histogram results are used to evaluate the performance of ANN. The results were mediocre reflecting the type of data provided in neural network training. ANN models could successfully train more than half of the data to achieve the target, However, the rest of the data were not able to achieve the desired output. Furthermore, Analysis of Variance (ANOVA) is used on MSE to determine which parameter influenced the outcome. Hidden neurons are significant factor were MSE of 10 neurons is smaller than MSE of 20 neurons, indicating a better performance for ten neurons models. Then, the deterioration scenarios are compared with ANN output age. Footbridge bridge and compression members had the longest average for service life. | en_US |
| dc.identifier.other | 2014143021 | |
| dc.identifier.uri | http://bspace.buid.ac.ae/handle/1234/1128 | |
| dc.language.iso | en | en_US |
| dc.publisher | The British University in Dubai | en_US |
| dc.subject | bridge deterioration | en_US |
| dc.subject | artificial neural networks | en_US |
| dc.subject | structural engineering | en_US |
| dc.title | Estimating Bridge Deterioration Age Using Artificial Neural Networks | en_US |
| dc.type | Dissertation | en_US |