Seismic Resilience Evaluation of Urban Water Distribution Network Using Machine Learning
Subject Areas : Farm water management with the aim of improving irrigation management indicators
Mahnaz Haghighi
1
(Ph.D. Candidate in Construction Engineering and Management, Department of Civil Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran.)
Ali Delnavaz
2
(Assistant professor 'Department of Civil Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran.)
Majid Safehian
3
(Assistant professor, Department of Civil Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran.)
Mohammad Delnavaz
4
(Associate professor "Faculty of Engineering, Civil Engineering Department, Kharazmi University, Tehran, Iran)
Keywords: Seismic vulnerability, Seismic resilience, Machine learning, Water distribution network, Water pipeline system,
Abstract :
Background and Aim: Seismic resilience is one of the new concepts in risk and crisis management, which refers to the investigation of seismic performance, recovery, and re-service (after an earthquake) of urban structures and infrastructures. In this regard, the water distribution network, which is one of the vital arteries for providing drinking water to the citizens after the earthquake, as well as providing water for (possible) fire extinguishing purposes, is highly concerned; Therefore, creating a resilient water supply network against earthquakes has been very important and necessary in urban areas. Therefore, evaluating and improving the seismic resilience of the urban water supply network and providing accurate computational models to predict the seismic behavior of this infrastructure has been one of the main challenges facing researchers. This research is focused on evaluating the seismic resilience of the urban water distribution network, which was conducted as a case study on the water supply network of Tehran (District 1-4 of the Tehran Municipality). Method: In this research, 5 main machine learning algorithms have been used, each with different computational branches. The results of the performance of the machine learning model based on different algorithms have shown the amount of error in predicting the strain of the buried pipe under seismic loading compared to the exact values of the strain based on the initial data set. Based on the results, it is clear that GPR models have better performance than other computational models of machine learning. Therefore, among the algorithms of the GPR method, the Squared Exponential GPR algorithm has the lowest error rate, so the machine learning model developed based on this algorithm has an MSE error equal to 3.0564 and a correlation value R equal to 0.94. To evaluate the seismic resilience of the water distribution network for districts 1 to 4 of Tehran, the amount of damage to the water supply network under 5 seismic loading scenarios was carried out using the ML seismic vulnerability prediction model. Also, by examining the Bridge index, the relative amount of damage in the pipeline network for districts 1 to 4 of Tehran was investigated separately for each of the 5 seismic scenarios. In the assessment of seismic resilience, the index of the minimum amount of water needed by the community has been mentioned as one of the main social dimensions of infrastructure resilience to provide vital water demand for citizens in the event of a disruption in the water supply network. Therefore, according to the population of each region and the amount of water needed Based on the provided standards, it has been calculated and evaluated. The recovery time index for each region and separately for each of the 5 seismic scenarios has been calculated and reviewed. The recovery speed of the water distribution and transmission network after the earthquake is the main indicator. Seismic resilience evaluation for urban infrastructure has been discussed. Based on the results of the recovery index evaluation, it is determined that the conditions of access to the water supply network until the accident (t=0) were normal and the water supply network had 100% performance and no disruption. Earthquake (t>0) the system has been disrupted and access to the water supply network has decreased significantly, then with the start of repair and replacement of the first damaged line, the access and utilization of the water supply network increased until after The recovery time (t=tr) of the network has returned to normal conditions. According to the results, it is clear that the water supply network under scenarios with greater earthquake magnitude had a longer recovery time than under earthquake scenarios with a smaller magnitude. Conclusion: The proposed model was found to be accurate enough to evaluate seismic damage to urban water distribution networks. Moreover, the mechanism proposed based on the used indices to evaluate the seismic resilience of water distribution provides comprehensive insights in order to improve the performance of the urban water infrastructure under earthquakes. The proposed model can be used as a comprehensive framework for different regions under different seismic loads in order to not only estimate the seismic vulnerability of the water pipeline system but also accurately evaluate seismic resilience and enhance seismic resilience in these critical infrastructures.