A novel risk assessment approach for data center structures

Abstract

Structural safety includes evaluation of both structural and nonstructural components of buildings. Although structural design is completed only considering structural elements of buildings, nonstructural components are crucial in an earthquake event. Post-earthquakes areas show that structural safety may not be ensured even when the load-bearing system is undamaged. Failure of nonstructural components resulted in loss of enormous economic losses and loss of life in past earthquakes. Therefore, nonstructural components should also be included in seismic safety evaluation of structures. Researches show that cost of the nonstructural components ranges from \%70 to \%90 of the total cost of buildings. Therefore, nonstructural component failure in structures with high-tech equipment, laboratories, data centers can damage economy significantly. Additional to economic losses from downtime of these structures, repairing or replacement of equipment inside increase the cost extremely. Apart from the economic losses, damaged nonstructural components can be the cause of deaths directly by falling onto people and closing pathways. During and after an earthquake event, damaged nonstructural components can prevent escape of people inside and entry of medical staff. Moreover, operational failures caused by nonstructural components in critical facilities such as hospitals and fire stations, can lead to higher number of deaths after earthquake occurred. Nonstructural components do not participate to load-bearing systems in structures. However, they are still subjected to external loads with the load-bearing system. Therefore, it is crucial to design structures by considering the nonstructural systems inside. Nonstructural components can be classified in 3 groups by their functions: (i) architectural components such as, partition walls and lighting systems, (ii) mechanical-electrical components such as piping systems and generators, and (iii) building equipment such as, computers and file cabinets. Researches show that some of nonstructural components are sensitive to acceleration whereas the rest are sensitive to floor displacement ratio. According to the function of the structure, design should be completed to limit the defining response. This study aims to propose a new method and generate risk curves for structural design and structural evaluation of data centers in high seismic risk regions. A sample structure with base isolation system is selected from current literature in companion with standards for data centers. Structural properties are also selected in companion with standards. After the structure model is generated, probabilistic seismic hazard assessment is completed for the selected site where the main campus area of Istanbul Technical University in Maslak, Istanbul. Source-to-site distances are determined by using online map in General Directorate of Mineral Research and Exploration website. The closest point of main line of Western North Anatolian Fault is approximately 28 km away from ITU campus and the longest effective distance is selected as 65 km on the Western NAF. Probability of rupture distance is taken as uniform and 6 different values of distances between 28 km and 65 km are used in Ground Motion Prediction Equations. Characteristic earthquake method is considered and the characteristic magnitude is used as 7.2 in GMPEs. Probabilistic study is conducted on this structure by using Monte Carlo simulations with the selected structural parameters. Probabilistic distributions for different parameters are taken from various studies in literature. Random samplings are generated for each parameters according to the belonging probabilistic distributions. For comparison purpose the structure is also analyzed as a fixed-base structure. Same procedures are repeated for the fixed-base structure. Failure of nonstructural components are investigated in two different ways. The first failure criterion is overturning-sliding behavior of server racks. FEMA P58 and ASCE 7-16 is used to calculate acceleration limits for anchored nonstructural components. The second failure criterion is the acceleration limitations of servers given by producers and researchers. A special MATLAB code script is generated to run Monte Carlo simulations on OpenSees platform. Fragility curves are generated according to the predefined failure criteria. Risk curves are created for both structures with the site specific annual hazard curve and generated fragility curves. Results show that base-isolation systems reduces the accelerations significantly comparing to the fixed-base structures in higher floors. Another outcome is the isolation systems are highly sensitive to earthquake characteristics rather than structural variables in terms of accelerations. It was also understood that the critical failure mod in data centers is the overturning-sliding behavior rather than vibration failure of servers.