This document summarizes research on building pounding and how mass distribution affects the response. Building pounding occurs when adjacent buildings collide during an earthquake due to lack of separation. It can cause local damage at impact points and global damage by energy/momentum transfer. The study models pounding using lumped masses and distributed masses to represent floors. Different mass distributions are modeled to investigate their effects on pounding response.
Original Description:
Original Title
Building pounding and the effects of mass distribution
This document summarizes research on building pounding and how mass distribution affects the response. Building pounding occurs when adjacent buildings collide during an earthquake due to lack of separation. It can cause local damage at impact points and global damage by energy/momentum transfer. The study models pounding using lumped masses and distributed masses to represent floors. Different mass distributions are modeled to investigate their effects on pounding response.
This document summarizes research on building pounding and how mass distribution affects the response. Building pounding occurs when adjacent buildings collide during an earthquake due to lack of separation. It can cause local damage at impact points and global damage by energy/momentum transfer. The study models pounding using lumped masses and distributed masses to represent floors. Different mass distributions are modeled to investigate their effects on pounding response.
Building pounding and the effects of mass distribution
Gregory Cole, Rajesh Dhakal, Athol Carr and Des Bull contact: Gregory.Cole@pg.canterbury.ac.nz
What is pounding? How do we model pounding? What are the effects of
Building pounding occurs when two adjacent buildings collide. This study considers the simple case of collision between two buildings mass distribution? Earthquakes can cause pounding when adjacent buildings have little or with floors at the same height. The common method of modelling no gap providing separation. As the photos show, the resulting change in approximates the collision of two floors as collision between two lumped To investigate the effects of mass distribution a simple building demand loads can lead to catastrophic collapse of one or both buildings. masses. configuration is subjected to numerical analysis. m1 m2 Two types of pounding damage can occur = Building 1 Building 2 1. local damage at the point of impact Building 1 Building 2 10 m 2. global damage resulting from the energy and Floors are modelled as lumped masses under two conditions; if only one 20 m 50 m momentum transfer caused by collision. node is used to represent the floor, or if all nodes in a floor are Building Weight Interstorey Floor stiffness Building Period Yield load horizontally slaved together. By using conservation of momentum and stiffness Local damage is caused by the collision force while conservation of energy, the post collision velocity (v‟) of each floor can Building 1 1600 kN 420000 kN/m 1881300 kN/m 0.2 Sec 2400 kN global damage depends on the dynamic properties be calculated in terms of each building‟s mass (m) and initial velocity (v). Building 2 3000 kN 350000 kN/m 752520 kN/m 0.3 Sec 1400 kN of both buildings at the time of collision. However the lumped mass assumption prevents any calculation of collision force or collision duration. To model contact in the program, a spring is placed between the two Many New Zealand buildings have little or no separation with their neighbouring structures. The floors at each storey. This spring only activates when the two floors are An alternate solution may be found by modelling the mass as evenly additional risk to life safety caused by pounding is in contact, and is known as the “contact element”. Izmut earthquake, Turkey 1999 distributed over the length of the floor. The solution to this model very difficult to assess for these structures. This is Source: http://www.world-housing.net/ whereport1view.php?id=100031 requires the use of the one dimensional wave equation. Six different models were tested and compared to a model with 20 axial due to the large number of variables that may affect elements in each floor. The distributed masses were modelled using 1, 2, pounding damage. m1, k1 m2, k2 m1 m2 3, 4 and 5 elements per floor. The contact stiffness for these models used T1 2 T2 2 Experimental testing of typical building configurations could be = k1 k2 a relationship developed by Watanabe and Kawashima (2004). Three undertaken. However this form of testing is expensive, time consuming Building 1 Building 2 versions of the lumped mass model were also tested using different and requires highly specialised equipment. Furthermore the results are The resulting relationships differ when compared to the lumped mass stiffness's for the contact element. The contact element stiffness used in difficult to extrapolate to generalised cases, due to the complexity of lumped mass models varies widely, so the stiffness was set as 0.1, 1 and models. Furthermore, both collision duration and collision force can be interactions between buildings. 10 times the floor stiffness of building 1 in the three models. easily calculated. The solution introduces new parameters including the axial stiffness (k) of the each floor, and the „collision period‟ (T) of each The maximum displacement envelope error of the second storey of A much cheaper alternative is to use non-linear time history modelling in building. Note building numbering is selected so that T1 ≤ T2. building 2 is presented order to approximate the response on a computer. This form of analysis can be undertaken in a design office and can represent the specific Building 1 post Building 2 post Collision force Collision duration for ten seconds of three collision velocity collision velocity problem on hand. different earthquakes. m2 m1 This error represents the Lumped mass v1 v1 2 v1 v 2 v2 v2 2 v 2 v1 - - In order to be confident in the results from time history models, an m1 m2 m1 m2 difference in maximum appropriate representation of each structure must be created. v1 v1 2 1 v 2 v1 v 2 v2 2 1 v1 v 2 F 2 v1 v2 m1 m2 displacement between Distributed mass mT T2 m2 T1 T2 T min 2 ,2 1 1 2 m2T1 T1 m1 m1 m2 k1 k2 the 20 element model and each tested model. In time history programs, specific elements for distributed masses do not generally exist. Instead the distributed masses are approximated using multiple axial elements in each floor. However, the number of axial Conclusions elements in each floor required to provide an accurate solution is not known. Thus, a numerical analysis is performed to determine the All lumped mass models show significantly more error than their minimum number of elements that should be used when modelling distributed mass equivalents. At least one element per floor is required to distributed masses. be consistently accurate to 10 %. Thus the selection of mass modelling can cause significantly different pounding responses in numerical mA mB mC mD mW mX mY mZ analysis. ≈ L'Aquila earthquake, Italy 2009 Source: http://www.air-worldwide.com/PublicationsItem.aspx?id=17228 Mexico City earthquake, Mexico 1985 Source: http://www.smate.wwu.edu/teched/geology/eq-Mexico.html D Z References: Watanabe G and Kawashima K, Numereical Simulation of Pounding of Bridge Decks, where m1 m j and m2 mj 13th World Conference on Earthquake Engineering Conference Proceedings. 2004 j A j W