LESSONS LEARNED FROM PAST EARTHQUAKES

Transcription

1 LESSONS LEARNED FROM PAST EARTHQUAKES Shunsuke Otani ABSTRACT The development of design seismic forces is reviewed. The earthquake engineering was shown to have developed in the last 60 years. The characteristics of ground motion and structural response were gradually understood over the last five decades. The damage statistics from recent major earthquakes indicated that the severe damage was observed in taller buildings. The ratios of buildings with operational damage were as high as 90 percent because predominant buildings were low-rise buildings and such low-rise buildings suffered much less damage. Contrary to the statistics, the nonlinear response analysis results showed higher ductility demand (damage) to low-rise buildings. Causes of earthquake damage are briefly discussed for reinforced concrete buildings. A large number of problems have been solved in research community. It is essential for the disaster reduction that the information in research domain should be transferred to the engineering community and implemented in engineering practice. The future direction of earthquake engineering is discussed in terms of structural engineering. INTRODUCTION Engineering appears similar to Science, but there is an obvious difference between the two. The objective of engineering is to produce a product to meet people's demand, whereas the objective of science is to understand the natural phenomena. The theoretical soundness and basis are essential in science. However, they may be desired in engineering, but are not essential. A large error could be tolerated if sound engineering judgement can be made. Earthquake engineering is aimed to build human environment safe from an earthquake disaster. Therefore, our predecessors studied each case of earthquake disasters carefully immediately after the event, and investigated possible causes and their solutions. Counter measures have been developed against thousands of bitter experiences to mankind from earthquake disasters. As this engineering has been built on the experience, it is not theoretically consistent. Although the earthquake engineering has improved the built environment in the world, there remains a lot to be improved to protect human life from future earthquake disasters. This paper concentrates on lessons learned from past earthquakes about earthquake resistance of buildings with an emphasis on reinforced concrete structures. Professor, University of Tokyo, Department of Architecture, Tokyo, Japan 1

2 EQUIVALENT STATIC SEISMIC DESIGN FORCES Dr. G. W. Housner presented a conference lecture, entitled "Historical View of Earthquake Engineering [1]," at the opening of Eighth World Conference on Earthquake Engineering (1984) in San Francisco. He pointed out that "Earthquake engineering is a 20th Century development, so recent that it is yet premature to attempt to write its history." As an example, the development of seismic design forces is reviewed below. J. Milne ( ), an invited British professor in mining and geology at the College of Engineering, Imperial University of Japan (University of Tokyo), recorded the disaster of the 1891 Nohbi (Mino-Owari) Earthquake in Japan [2]. The earthquake (M 7.4) killed more than 7,000 people and caused a significant damage to then modern (western) brick construction in Nagoya area. Milne [2] reported that "buildings on soft ground suffer more than those on the hard ground." and that " we must construct, not simply to resist vertically applied stresses, but carefully consider effects due to movements applied more or less in horizontal directions." The resistance against horizontal forces was already recognized as necessary for structures, but the quantification of lateral forces had not been formulated in the last century. After this earthquake the study for earthquake resistant construction started in Japan, and the improvement of structural detailing was recommended for timber construction. Quantitative definition of design seismic forces was proposed after the 1908 Messina Earthquake (M 7.5), Italy, which killed more than 83,000 people. Dr. Housner states [1] that "The government of Italy responded to the Messina earthquake by appointing a special committee composed of nine practicing engineers and five professors of engineering to study the earthquake and to make recommendations. M. Panetti, Professor of Applied Mechanics in Turin... recommended that the first story be designed for a horizontal force equal to 1/12 the weight above and the second and third stories to be designed for 1/8 of the building weight above." Although this was the first attempt to quantify the design earthquake force, this recommendation was made without the knowledge about the earthquake ground motion nor the understanding of the structural resistance. The 1923 Kanto (Tokyo) Earthquake (M 7.9) caused significant damage in Tokyo and Yokohama areas, Japan, with a loss of life more than 140,000. More than 90 percent of the dead and lost were caused by fire. The Urban Building Law Enforcement Regulations introduced a seismic coefficient of 0.10 in the seismic design requirements in 1924; i.e., maximum ground acceleration of 0.3 g estimated in downtown Tokyo during the Kanto Earthquake was divided by the safety factor of 3 used in determining allowable stress level. No response amplification by a structure was considered. In the U.S., the 1925 Santa Barbara Earthquake, California, triggered the adoption of earthquake resistant building code in several Californian communities. The first edition of the Uniform Building Code in 1927 suggested a lateral base shear coefficient of 0.1 for buildings in high seismic risk areas. Although design seismic forces were specified in the code, no practical methods of stress analysis were made available to engineers. A structural analysis method available at the time was, for example, Castigliano's theorem, but such were tools only for academic researchers. An inaccurate portal method was used by engineers when Dr. H. Cross [3] introduced the moment-distribution method in Dr. K. Muto [4] proposed the D-value method for a frame analysis (a practical method for design using tables and figures) in

3 Indeed, it is not a long time ago that our predecessors were provided with minimum tools to calculate the stresses in a structure under a simulated earthquake effect. EARTHQUAKE MOTION AND STRUCTURAL RESPONSE Dr. Housner pointed out [1] that "The first accelerographs were installed by the Seismological Field Survey of the U.S. Coast and Geodetic Survey in late 1932, just in time to record the strong ground shaking of the destructive March 10, 1933 Long Beach earthquake. This was a most important step in the development of earthquake engineering." Since then, many accelerographs have been installed in the ground and on the structure. These records have provided us with precious information about the characteristics of earthquake motions and structural responses. A strong ground motion, recorded at El Centro station during the 1940 Imperial Valley Earthquake, has been treated as a standard strong motion and has been used in research as well as in design practice. The impact of theoretical analysis was felt for the first time in earthquake engineering when the earthquake response was calculated by a mechanical response analyzer [5]. The fact that more flexible buildings would typically be subjected to lower seismic forces and that the lateral force distribution would not be uniform over the structural height was recognized in the 1943 building code provisions of the City of Los Angeles. The seismic probability map was published by the U.S. Coast and Geodetic Survey in 1948 and was adopted in the 1952 Uniform Building Code. In this manner, the concept of probability was introduced in determining seismic risks. The development of analog as well as digital computers accelerated the understanding of structural response under earthquake excitations. Nonlinear earthquake response studies of simple systems by A. S. Veletsos and N. M. Newmark [6] provided guidelines to estimate lateral force resistance required for a system having a specified deformation capability; i.e., (a) equal energy rule for short- to medium-period systems, and (b) equal displacement rule for long-period systems. This concept was used to rationalize the design seismic forces being much smaller than the elastic response spectral acceleration. The 1959 SEAOC Code [7] introduced the type of construction factor to recognize the ductility of a structural system and the importance factor for the use of a building. The accumulation of experimental data on the behavior of structural models and members under simulated earthquake forces made it possible to analyze a realistic inelastic response of structures during an earthquake in the late 1960s. J. Milne observed the difference in building response on firm and soft soils at the end of the last century [2]. The effect of surface geology on ground motion amplification was recognized in the 1976 ATC-03 recommendation [8]. EARTHQUAKE DAMAGE STATISTICS Reliable statistics about damaged and undamaged buildings are important to establish disaster reduction measures against future earthquakes. Some comprehensive 3

4 statistics were obtained after the 1985 Mexico Earthquake, the 1990 Luzon (Philippines) Earthquake, the 1992 Erzincan (Turkey) Earthquake and the 1995 Hyogo-ken Nanbu (Kobe) Earthquake. The damage of all buildings in a selected area was surveyed by external observation. The damage was grouped into three levels (six levels in the original investigations); i.e., (a) operational damage: columns or structural walls were slightly damaged in bending, and some shear cracks might be observed in non-structural walls, (b) heavy damage: spalling and crushing of concrete, buckling of reinforcement, or shear failure in columns were observed, and lateral resistance of shear walls might be reduced by heavy shear cracking, and (c) collapse, which also included those buildings demolished at the time of investigation. The 1985 Mexico Earthquake An M8.1 earthquake occurred on the Mexican west coast on September 19, 1985, followed by an M7.5 after-shock on September 21. The two successive events caused a significant damage in Mexico City approximately 400 km away from the epicenter. The structural damage was concentrated in a region of old lake bed which was gradually filled with the development of the city. The damage was attributed to the amplification of long-period components of the ground motion by deep and soft soil deposit underlain in the Mexico Valley. The damage statistics in Table 1 were obtained one to two months after the earthquake in the old lake bed zone, including heavily damaged residential as well as commercial districts, covering slightly more than 20 percent of the central city [9]. Approximately 73.4 percent of the 4,532 buildings surveyed were three-stories or lower; 94.0 percent of buildings were six-story or lower. Ninety-four (93.8) percent of the buildings were judged operational in the surveyed area because more than 94 percent of those dominantly low-rise buildings (six stories or lower) were operational. The damage increased sharply in mid- to high-rise buildings (seven stories and higher). Nearly or more than half of nine-story and taller buildings suffered heavy damage or collapsed; 37.2 percent out of 129 buildings suffered heavy damage and 15.5 percent collapsed. The ground motion, dominated by long period components in the lake bed zone, definitely increased the damage of taller buildings. The 19-story building with operational damage is the Latin Americana Tower located at the center of a commercial district. The damage was light in a area where stiff low-rise buildings were dominant. It should be noted that the percentage of severe damage was extremely small up to six-story buildings in comparison with those in the other statistics. The heavy damage and collapse of tall buildings gave a significant impact to the society. The flat slab construction collapsed in a form of piled pancakes, leaving no space for life survival between adjacent slabs. The pounding of adjacent buildings was observed partially due to the rocking motion of the structures. Tables 2 and 3 show the damage of reinforced concrete and masonry buildings in the city of Lazaro Cardenas, in the epicenter region. The damage in one- and two-story reinforced concrete buildings was relatively small, but severe damage increased with the number of stories, especially in four-story or taller buildings. The damage to oneand two-story masonry construction was also light. 4

5 Table 1: Damage statistics of buildings in lake bed zone of Mexico City [9] No. of Stories Operational Damage Heavy Damage Collapse Total 1 689(97.5) 17(2.4) 1(0.1) 707(100) 2 1,756(96.0) 57(3.1) 17(0.9) 1,830(100) 3 752(95.3) 26(3.3) 11(1.4) 789(100) 4 412(95.2) 16(3.7) 5(1.2) 433(100) 5 332(94.1) 9(2.5) 12(3.4) 353(100) 6 140(94.0) 2(1.3) 7(4.7) 149(100) 7 74(75.5) 11(11.2) 13(13.3) 98(100) 8 34(77.3) 9(20.5) 1(2.3) 44(100) 9 23(54.8) 8(19.0) 11(26.2) 42(100) 10 15(62.5) 6(25.0) 3(12.5) 24(100) 11 11(57.9) 8(42.1) 0(0.0) 19(100) 12 3(18.8) 10(62.5) 3(18.8) 16(100) 13 1(14.2) 6(85.7) 0(0.0) 7(100) 14 2(50.0) 2(50.0) 0(0.0) 4(100) 15 4(44.4) 5(55.5) 0(0.0) 9(100) 16 0(0.0) 1(50.0) 1(50.0) 2(100) 17 2(40.0) 1(20.0) 2(40.0) 5(100) 18 1(100) 0(0.0) 0(0.0) 1(100) Total 4,251(93.8) 194(4.3) 87(1.9) 4,532(100) ( ) : percentage to the total of the same story height. Table 2: Damage statistics of RC buildings in Lazaro Cardenas [9] No. of Stories Operational Damage Heavy Damage Collapse Total 1 45(90.0) 5(10.0) 0(0.0) 50(100) 2 71(89.9) 7(8.9) 1(1.3) 79(100) 3 13(72.2) 4(22.2) 1(5.6) 18(100) 4 6(50.0) 6(50.0) 0(0.0) 12(100) 5 2(50.0) 2(50.0) 0(0.0) 4(100) 9 0(0.0) 1(100) 0(0.0) 1(100) Total 137(83.5) 25(15.2) 2(1.2) 164(100) ( ) : percentage to the total of the same story height. Table 3: Damage statistics of masonry buildings in Lazaro Cardenas [9] No. of Stories Operational Damage Heavy Damage Collapse Total 1 82(100) 0(0.0) 0(0.0) 82(100) 2 24(88.9) 2(7.4) 1(3.7) 27(100) 3 0(0.0) 1(100) 0(0.0) 1(100) Total 106(96.4) 3(2.7) 1(0.9) 110(100) ( ) : percentage to the total of the same story height. 5

6 The 1990 Luzon (Philippines) Earthquake An earthquake (M7.7) occurred approximately 100 km to the north of Manila, Philippines, on July 16, 1990, and killed approximately 2,000 people. Many high-rise hotel buildings collapsed in the resort city of Baguio. The damage statistics were collected in a major commercial district of the city of Baguio. Ninety-three (92.8) percent of the buildings in the area were of five stories or lower. Seventy-six (76.2) percent of the buildings suffered operational damage, 18.8 percent suffered heavy damage, and 5.0 percent collapsed. Severe damage was observed in taller buildings. The percentage of heavy damage and collapsed buildings was higher in the low-rise buildings (six-story or lower) compared with the other earthquake cases. Table 4: Damage statistics of RC buildings in Baguio City [10] No. of Stories Operational Damage Heavy Damage Collapse Total 1 5(71.4) 2(28.6) 0(0.0) 7(100) 2 34(75.6) 10(22.2) 1(2.2) 45(100) 3 42(84.0) 7(14.0) 1(2.0) 50(100) 4 36(85.7) 4(9.5) 2(4.8) 42(100) 5 15(62.5) 7(29.2) 2(6.9) 24(100) 6 4(80.0) 1(20.0) 0(0.0) 5(100) 7 0(0.0) 2(50.0) 2(50.0) 4(100) 8 2(66.7) 0(0.0) 1(33.3) 3(100) 9 0(0.0) 1(100) 0(0.0) 1(100) Total 138(76.2) 34(18.8) 9(5.0) 181(100) ( ) : percentage to the total of the same story height. The 1992 Erzincan Earthquake An Ms6.9 earthquake occurred on March 13, 1992, with an epicenter (focal depth of 28 km) near the City of Erzincan, in the eastern part of the Anatolia Plateau, Turkey. In the City of Erzincan, two heavily damaged residential areas were selected for the inventory survey or reinforced concrete buildings; i.e., (a) Fatih, Yunus and Aksemsettin Districts, and (b) Yavus Selim District. Ninety (90.3) percent of the buildings in the area were 2- to 4-story apartment buildings. A joint team, organized by the Architectural Institute of Japan and the Japan Society for Civil Engineers, worked with the researchers from Bogazici University, Istanbul [11]. Dr. P. Gulkan, Middle East Technical University, provided guiding information about the damage in the area. The damage of one- and two-story buildings was very light (Table 5). No three-story buildings collapsed, but 23 percent of the buildings suffered heavy damage. Seventeen (17.3) percent of four-story buildings collapsed and 30.9 percent suffered heavy damage. It is interesting to note that the building height was limited to three stories along main streets and two stories in the other area after the devastating 1939 Erzincan Earthquake (M 7.9). A recent growth in population and economy as well as strong confidence in technical development relaxed the height limitation; i.e., six-story buildings are allowed along main streets and four-story buildings in the other areas. 6

7 Damage must have been much smaller without the relaxation in building height. Table 5: Damage statistics of reinforced concrete buildings in Erzincan [11] No. of Stories Operational Damage Heavy Damage Collapse Total 1 28(100) 0(0.0) 0(0.0) 28(100) 2 125(100) 0(0.0) 0(0.0) 125(100) 3 114(77.0) 34(23.0) 0(0.0) 148(100) 4 57(51.8) 34(30.9) 19(17.3) 110(100) 5 4(100) 0(0.0) 0(0.0) 4(100) unknown 0(0.0) 0(0.0) 9(100) 9(100) Total 328(77.4) 68(16.0) 28(6.6) 424(100) ( ) : percentage to the total of the same story height. The 1995 Hyogo-ken Nanbu (Kobe) Earthquake The 1995 Hyogo-ken Nanbu Earthquake killed nearly 6,300 people (including direct and indirect causes by the earthquake), injured more than 40 thousands people. The Kinki (Osaka and Kyoto Region) Branch, Architectural Institute of Japan, investigated the damage of all existing reinforced concrete buildings (3,911 buildings in total) in the region of seismic intensity VII (highest seismic intensity defined by the Japan Meteorological Agency) in Nada and Higashi-Nada wards, Kobe City [12]. Seventy-five (75) percent of the 3,911 buildings surveyed were residential buildings (including those used partially for office or shop). Forty-eight (47.5) percent of the buildings were built in conformance with the current Building Standard Law (revised in 1981). Eighty-three (82.5) percent of the buildings were lower than or equal to five stories high (73.0 percent were from 3 to 5 stories high) and 9.7 percent had soft first-story construction. Eighty-nine (88.5) percent of the buildings surveyed suffered operational damage, 5.9 percent suffered heavy damage and 5.7 percent collapsed. Among those 2,035 buildings constructed before the current Building Standard Law (1981), 7.4 percent suffered heavy damage and 8.3 percent collapsed. Among those 1,859 buildings constructed using the current Building Standard Law, 3.9 percent suffered heavy damage and 2.6 percent collapsed. The 1981 revision of the Building Standard Law enhanced significantly the performance of reinforced concrete buildings against earthquake attack. The relation between damage level and the number of stories is shown in Tables 6 and 7 for buildings constructed before and after the enforcement of the 1981 Building Standard Law. The ratio of severer damage increases with the number of stories, especially for buildings taller than 5 stories. The percentage of the pre-1981 buildings, which remained operational after the earthquake, is 87.9 percent for 1- to 5-story buildings, 66.1 percent for 6- to 8-story buildings, and 46.0 percent for 9-story and higher buildings. The corresponding percentages of the post-1981 buildings are 97.1 percent, 87.1 percent and 70.7 percent, respectively. The revision of the law improved the level of safety almost uniformly from low-rise to mid-rise buildings. Even after the revision of the law, 20 percent of buildings taller than seven stories suffered heavy damage or collapsed. 7

8 Table 6:Damage of RC buildings in Kobe constructed before 1981 [12] No. Stories Operational Damage Heavy Damage Collapse Total 1 20(90.9) 1( 4.5) 1( 4.5) 22(100) 2 215(92.7) 9( 3.9) 8( 3.4) 232(100) 3 532(93.0) 17( 3.0) 23( 4.0) 572(100) 4 524(85.8) 41( 6.7) 46( 7.5) 611(100) 5 269(79.6) 29( 8.6) 40(11.8) 338(100) 6 59(75.6) 10(12.8) 9(11.5) 78(100) 7 49(58.3) 16(19.0) 19(22.6) 84(100) 8 19(63.3) 7(23.3) 4(13.3) 30(100) 9 3(33.3) 4(44.4) 2(22.2) 9(100) 10 20(48.8) 15(36.6) 6(14.6) 41(100) ( ): Ratio to the number of buildings of the same height(%) Table 7:Damage of RC buildings in Kobe constructed after 1981 [12] No. Stories Operational Damage Heavy Damage Collapse Total 1 8( 100) 0( 0.0) 0( 0.0) 8(100) 2 85(98.8) 0( 0.0) 1( 1.2) 86(100) 3 460(98.1) 2( 0.4) 7( 1.5) 469(100) 4 508(95.7) 9( 1.7) 14( 2.6) 531(100) 5 333(97.4) 5( 1.5) 4( 1.2) 342(100) 6 135(91.8) 9( 6.1) 3( 2.0) 147(100) 7 90(86.5) 12(11.5) 2( 1.9) 104(100) 8 44(75.9) 11(19.0) 3( 5.2) 58(100) 9 19(73.1) 7(26.9) 0( 0.0) 26(100) 10 51(69.9) 18(24.7) 4( 5.5) 73(100) ( ): Ratio to the number of buildings of the same height(%) DESIGN SEISMIC FORCES AND EARTHQUAKE RESPONSE A series of single-degree-of-freedom (SDF) systems were designed using the governing building code. The nonlinear response of the SDF systems was calculated under earthquake motions recorded near the area of the inventory damage surveys after the 1985 Mexico Earthquake and the 1992 Erzincan Earthquake. The Takeda model [13] was selected to simulate the response of well-designed reinforced concrete buildings. The skeleton curve was idealized by a tri-linear relationship with stiffness changes at cracking point (D c, F c ) and yielding point (D y, F y ). Fixed relations were arbitrarily assumed for the cracking and yielding points; F c = F y / 3 and D c = D y / 12. Mass of the system was assumed to be unity. The damping coefficient was assumed to vary proportional to instantaneous stiffness with a damping factor of 0.05 at the initial elastic stage. 8

9 Response under the 1985 Mexico Earthquake Motion The earthquake resistant design before the 1985 Mexico Earthquake was governed by the 1977 Construction Regulations for the Mexico D.F. The yield base shear coefficient F y was calculated for the lake bed zone (zone III): T T Fy = [ ao + ( c ao) ]/[1 + ( Q 1) ] for T < T1 T1 T1 Fy = c / Q for T1 < T < T2 in which c=0.24, a o =0.06, r =1.0, T 1 =0.8 sec and T 2 =3.3 sec. Design ductility factors Q was selected to be 1.0, 2.0 and 6.0 for the study. The yield resistance F y was the same for Q=4.0 and 6.0. Three records (CDAF, CDAO, SCT1) were recorded in the lake bed zone in Mexico City. These records showed long period contents in the waveform. The response spectra of CDAF and SCT1 records exhibited large response at around 2.0 sec period, whereas CDAO record showed a peak at around 1.3 to 1.5 sec. The maximum acceleration of the EW component of SCT1 record was 1.68 m/sec 2, and 0.70 to 1.00 m/sec 2 in the other motions. Although the ground acceleration amplitudes were relatively small, the response amplification was large by dominance in certain long-period components. The ductility demands (maximum response displacement divided by yield displacement) are shown for Q=1.0 and Q=6.0 (or Q=4) in Figure 1. (1) For a design ductility factor of Q=1.0, the systems did not yield against CDAF and CDAO motions except at very short period. However, Record SCT1, especially the EW component, caused yielding for systems at most period range. No structures should have survived the EW motion of SCT1 unless much higher resistance was provided in the system. (2) For Q=2.0 (not shown in Figure 1), the ductility demand exceeded the target design values under CDAF and CDAO motions for a period range longer than 1.3 sec. Record SCT1 required ductility demand greater than the design value in all range of periods. (3) For Q=6.0 (or Q=4.0), the maximum response was comparable for the three records. The ductility demand exceeded the design target for a period range less than 2.4 sec. The response increased as the system period decreased. The nonlinear response analysis indicated that the systems designed with a small ductility and large yield resistance developed small ductility demand, whereas the systems designed with large ductility and small resistance developed a significantly large plastic deformation. A significant difference existed between the damage statistics and the calculated ductility demand. If the structures were to possess the lateral resistance required by the 1977 building code, and if the structures were to fail at the design ductility Q, most of the structures must have failed in the lake bed zone. On the contrary, the damage was small in low-rise buildings, which indicates that the low-rise buildings must have been designed for higher lateral resistance using lower design ductility factor Q. Furthermore, the actual resistance of buildings is normally greater than the code required value attributable to inherent additional resistance of the structural as well as non-structural elements. 9

10 Ductility Factor NS EW CDAF CDA0 SCT1 Q=1.0 Zone Period, sec (a) Design Ductility Factor Q=1.0 (Elastic) Ductility Factor Q=6.0 Zone 3 NS EW CDAF CDA0 SCT Period, sec (b) Design Ductility Factor Q=4.0 0r 6.0 (Very Ductile) Figure 1: Response to the 1985 Mexico Earthquake in the lake bed zone 10

11 Response under the 1992 Erzincan Earthquake Motion The governing design code was the 1975 Specifications for Structures in Disaster Area (Part 1), the Ministry of Reconstruction and Settlement. The yield resistance F y was determined as follows: Fy = CoK S I where, C o : seismic zone coefficient (=0.10), K: structural type coefficient (=0.6, 1.0 and 1.5), S: spectral coefficient, and I: building importance coefficient (=1.0). 60 K=0.6 Co=0.1 Soil I I=1.0 Ductility Factor K=1.5 K=1.0 Fc/Fy=1/3 Ky/Kc=1/3 h=0.05 Ductility Factor Period, sec (a) North-south motion K-1.5 K=1.0 K=0.6 Co=0.1 I=1.0 Soil I Fc/Fy=1/3 Ky/Kc=1/3 h= Period, sec (b) East-west motion Figure 2: Response to the 1992 Erzincan Motion (K = 0.6, 1.0, 1.5) 11

12 The spectral coefficient S is given below: 1 S = 0. 8 T T o in which T o : soil natural period (=0.25 sec for hard soil), T: building natural period. The soil was judged good from the micro-tremor measurement in the Erzincan. A strong motion was recorded at the Meteorological Services Buildings, located in the city approximately 5 km from the epicenter. Strong motion lasted approximately 20 sec with maximum acceleration of 0.5 g (g: acceleration of gravity) in the north-south direction and 0.4 g in the east-west direction. Ductility demand increased with a decreasing elastic period (Figure 2). For a six-story building with structural type coefficient K of 1.0, ductility demand was approximately 10 under the EW motion, and 25 under the NS motion; the ductility demand was much beyond the deformation capacity of RC buildings. All RC buildings must have collapsed if they were provided with resistance equal to the code required value. However, most low-rise buildings, especially less than 3 stories, survived the earthquake, indicating that the low-rise buildings in the area must have been provided with lateral resistance much higher than the code specified value. LESSONS LEARNED FROM RECENT EARTHQUAKES Amenity of Life, Retrofitting and Building Law The Ministry of Health and Welfare, Japan, investigated the causes of death in the Kobe disaster on 5,488 cases; 4,816 (77 %) were killed by pressure and suffocation under collapsed buildings. The fatal collapse of buildings was predominantly observed in deteriorated traditional timber houses with heavy roofs. The heavy weight in a roof was necessary for the heat insulation during hot summer days in Japan, and for the protection of the roof from blow-up during annual typhoons (storms with strong winds). The report also revealed that elderly people were affected worse than the young people. The elderly could not afford to maintain the structure of their houses nor did they wish to change their houses filled with past family memories and personal attachments. Most of structural members in collapsed houses were eaten by termites or rotten from weathering. The amenity and safety of daily living as well as the belief in low probability of strong earthquake occurrence did not motivate the people to prepare for a rare earthquake disaster. The government should have taken legal actions to update the existing structure to the minimum safety level for the sake of public welfare. Financial assistance should be considered to the elderly. In recent timber construction, newly developed light heat insulation materials are used in the roof, and the roof is attached rigidly to the structure to improve earthquake resistance. The damage statistics demonstrated the poor performance of old buildings designed using out-dated technology. The retrofit of deficient public buildings is an urgent task of the government. The owner is responsible for maintaining his building to the existing code level. The damage statistics clearly indicated that the number of buildings suffering heavy damage or collapse was small even in the most heavily damaged area. Therefore, an efficient and reliable screening procedure should be 12

13 employed to identify these probably deficient buildings from an entire existing building stock. The upgrading of taller buildings should be emphasized. The building law (or code) is a legal document that specifies a minimum design level of structure of a building. Note that the law limits the constitutional freedom of people for the sake of the public welfare. The legal requirements must be clear and prescriptive so that the building official can easily check the conformance. The minimum design level needs to be revised with the development of technology as well as the change in socio-economic demand. Legal actions were necessary to update the existing structure when the Building Law was revised. The owner of a building should select the performance level higher than that required by the code. Transfer of Technology from Research to Practice Earthquake damage is caused by the lack of proper engineering practice. Some examples are shown from the experience in Kobe and Erzincan. (1) Construction materials and practices A building must be constructed in accordance with design specifications about the use of materials, the arrangement of reinforcement, the concrete work, etc. Poor quality and placement of concrete was often observed in damaged buildings due to the lack of technology. Poor concrete work is sometimes attributed to the error in structural design; e.g., reinforcement may be congested in a section due to (a) the use of small cross sectional area, (b) the use of lap splicing, and (c) the anchorage of beam reinforcement in the already-congested beam-column joints. Construction work must be closely inspected and guided by structural engineers during construction processes. Construction workers must be trained as well as educated to handle right materials and to execute the work properly. (2) Importance of structural detailing in shear resistance After the Kobe earthquake, the writer and his associates investigated the bend at the end of hoop reinforcement in columns. Out of 64 heavily damaged or collapsed buildings studied, forty-two buildings used 90-degree hooks at both hoop ends; twenty buildings used combination of 90-degree hook at one end and 135-degree hook at the other end. Only two buildings used either 135-degree hook at both hoop ends or welded closed-shape hoops. Note that the 1933 Architectural Institute of Japan RC standard [4] recommended the use of 135-degree hooks at the ends of a hoop. However, the difficulty to bend a hoop end to 135 degree at the construction site prevented the worker from using the 135-degree hook until two decades ago. Columns of rectangular cross sections (e.g., roughly 200x500 mm or 300 x 600 mm) were often observed in Erzincan. These rectangular columns failed in shear in the major principal direction (long dimension) because the amount of lateral reinforcement did not meet shear input at flexural yielding in the major principal direction. In Baguio City, Philippines, ties were closely spaced at the top and bottom of a column probably to increase confining effect after flexural yielding. However, the spacing and amount of shear reinforcement in the middle part of columns were not sufficient to prevent shear failure before flexural yielding at the column ends. The collapse of reinforced concrete buildings was often caused by brittle shear failure of attributable to the use of (a) 90-degree hook at hoop ends, (b) wide spacing of hoop reinforcement, (c) thin hoop reinforcement, and (d) pain bars as longitudinal reinforcement. 13

14 (3) Beam-column connections No lateral reinforcement was found within beam-column connections failing in shear in Erzincan and Kobe. After spalling of concrete cover from the joint, the column longitudinal reinforcement was observed to buckle. The bottom reinforcement of a beam was often anchored straight in the beam-column connection in Erzincan. This is a common detailing practice in non-seismic regions, but will not allow flexural yielding under reversed loading. If a rectangular column is used, the beam reinforcement cannot be anchored in the narrow side of the beam-column connection. Structural Planning Structural engineers should consider the overall performance of a structure during medium as well as high intensity earthquake motions; (1) Resistance or ductility A structure can be designed earthquake resistant (a) by providing large lateral resistance but with limited ductility (strong structure) or (b) by providing large ductility but with relatively small lateral resistance (ductile structure). Naturally, a ductile structure suffers extensive structural as well as non-structural damages by a strong earthquake, or suffers minor damages even by frequent medium-intensity earthquakes. Consequently, the repair cost of the ductile structure must be significant. A strong structure will not be damaged until the resistance is reached during an earthquake. For this reason, the use of structural walls has been recommended for a long time. Only 2 percent of box wall systems, normally designed following simple prescriptive guidelines, suffered heavy damage (often in the foundation) or collapsed in Kobe [12]. (2) Weak-beam strong-column mechanism vs. Soft first-story mechanism The damage of soft first-story buildings and the other buildings is compared in Table 8 [12]. Less than three-quarters of the soft first-story buildings were operational compared to ninety (90.2) percent of the other buildings. The collapse of soft first-story structures was caused by brittle shear failure in the first story columns. The resistance and ductility must be improved in this type of construction. Table 8: Damage of soft first-story buildings during the Kobe Earthquake [12] Structures Operational Damage Heavy Damage Collapse Total Soft First-story Buildings (72.3) (14.2) (13.4) (100) Other Buildings 3,186 (90.2) 174 (4.9) 171 (4.8) 3,531 (100) The weak-beam strong-column mechanism has been preferred by many structural engineers because it dissipates earthquake-input energy at many distributed localities and reduces ductility demand at each planned yield hinge region. However, it is expensive to repair the damage at many scattered locations in the building after earthquakes of high intensity as well as medium intensity. Contrary to common belief, the damage in the soft first-story structure can be repaired using the present state of construction technology as long as the first story does not collapse to the ground. A 14

15 structural engineer should consider the necessary repair cost in design. (3) Non-structural elements and building content For the life safety, the structure should naturally be prevented from collapsing. At the same time, the response (acceleration or velocity) of a structure must be controlled to prevent heavy furniture and equipment from overturning on the floor or to prevent heavy equipment from falling from shelves; otherwise the contents of a building should be properly fastened to the structure. The Ministry of Health and Welfare, Japan, reported that 65 persons were killed under overturning heavy furniture. Non-engineer residents were greatly scared by the damage of non-structural elements, such as partitions, windows and doors. Stiff, weak and brittle brick walls, filled in a flexible moment-resisting frame, fail at an early stage even during medium-intensity earthquakes. Providing some gap on both side of a column could reduce such damage. Non-structural elements must be protected from damage because the fall of failed elements is dangerous for people escaping from the building, because the failed elements may block doors from opening, and because the building may not be occupied until the damaged elements are replaced. Furthermore, the cost of repair work is often governed by the replacement of the damaged non-structural elements rather than the repair work on structural elements. Fire Prevention and Resistance Large numbers of people were killed by fire in the 1906 San Francisco Earthquake, U.S.A., and the 1923 Kanto (Tokyo) Earthquake. Fire is a major disaster after a strong earthquake in a large urban area. In Kobe, the mortar or plaster cover protecting timber construction fire fell off during a severe vibration, and the exposed timber structure caught fire after the earthquake motion. FUTURE OF EARTHQUAKE ENGINEERING Contrary to our expectation, the damage statistics demonstrated that the percentage of severely damaged buildings was extremely low for reinforced concrete and masonry buildings even in the disastrous region. However, the impact of building collapse and heavy damage to the society was quite large as evidenced by many mass media reports. Building codes is intended to require minimum standard necessary for the life safety and public welfare. Therefore, the building owner should select a design level higher than the code specified level for the better function of the building and for the protection of properties and investment. A significant gap was observed between the expectations of structural performance by the building owner (people) and the structural engineer. During the design stage, the structural engineer should explain to the owner the expected performance of the building under various intensity earthquake motions including the expected damage of nonstructural elements. The cost of repair and occupancy after medium- and high-intensity ground motions should be taken into consideration. With the advice of the engineer, the owner should select the performance level of his building suitable for its function and within the available funds, but the performance level must be selected higher than the code minimum. The design engineer should design a building to satisfy the owner's demand making best use of available materials 15

16 and structural engineering in the community. The methods to achieve the performance level should not be limited, but the structural engineer should be able to choose most suited procedure for a structure under design. The engineer should pay more attention to safe design of tall buildings. The construction engineer should construct the building in conformance with the specifications of the structural engineer through close inspection and supervision by the structural engineer. REFERENCES 1. Housner, G.W. (1984), Historical View of Earthquake Engineering, Conference Lecture, Proceedings, Eighth World Conference on Earthquake Engineering, San Francisco, Post-Conference Volume, pp Milne, J. and Burton, W.K. (1891) The Great Earthquake of Japan, 1891, Lane, Crawford & Co., Yokohama, Japan. 3. Cross, H. (1930), Analysis of Continuous Frames by Distributing Fixed End Moments, Proceedings, ASCE. 4. Architectural Institute of Japan (1933), AIJ Standard for Structural Calculation of Reinforced Concrete Structures. 5. Biot, M.A. (1941) A Mechanical Analyzer for the Prediction of Earthquake Stresses, Bulletin, Seismological Society of America, Vol. 31, No. 2, April 1941, pp Veletsos, A. S. and Newmark, N. M. (1960), Effect of Inelastic Behavior on the Response of Simple Systems to Earthquake Motions, Proceedings, Second World Conference on Earthquake Engineering, Tokyo-Kyoto, Volume II, pp Seismological Committee (1959), Recommended Lateral Force Requirements, SEAOC Code, Structural Engineering Association of California. 8. Applied Technology Council (1976), Tentative Provisions for the Development of Seismic Regulations for Buildings, National Science Foundation and National Bureau of Standards. 9. Architectural Institute of Japan (1987), Reports on the Damage Investigation of the 1985 Mexico Earthquake (in Japanese), 599 pp. 10. Architectural Institute of Japan (1992), Reports on the Damage Investigation of the 1990 Luzon Earthquake (in Japanese), 396 pp. 11. Architectural Institute of Japan (1993), Report on the Damage Investigation of the 1992 Turkey Earthquake (in Japanese), 221 pp. 12. Reinforced Concrete Committee (1996), Damage Investigation Report on Concrete Buildings, the 1995 Hyogo-ken Nanbu Earthquake (in Japanese), The Kinki Branch, Architectural Institute of Japan, 245 pp. 13. Takeda, T., Sozen, M. A., Nielsen, N. N. (1970), Reinforced Concrete Response to Simulated Earthquakes, Journal, Structural Division, ASCE, Vol. 96, No. ST 12, pp