Professional Documents
Culture Documents
Viorel POPA1
1
Tudor POSTELNICU1
1 INTRODUCTION
The rehabilitation of existing seismic vulnerable buildings represents one of the most difficult tasks that the design professionals have to face in seismic areas. The number of existing buildings is by far much larger than the number of the new ones and the knowledge progress in the earthquakeengineering, imposes that an increasing number of buildings to be considered vulnerable. In these circumstances the development of new methods for the structural performance assessment of existing buildings becomes very important. The paper presents the application of a full probabilistic procedure, developed by Cornell & al. (2001) for determining the structural performance of a widely spread category of buildings in Romania, namely the industrial single level precast halls. The procedure which is a performance based assessment method takes into account the variability of the main parameters that characterize the ground motion and structural behavior.
2 THEORETICAL FUNDAMENTALS
Likewise in other assessment methods the performance checking consists in the comparison between a parameter that characterize the structural capacity and the corresponding parameter that defines the seismic demand. The procedure operates with three main random elements. First of them is the intensity measure (IM) of the ground motions. The spectral acceleration at the fundamental vibration period of the structure (determined for 5% damping), Sa, was chosen to describe the ground motion intensity. The Sa values can be associated with different hazard levels using site specific hazard functions, that provide the mean annual probability that a random value Sa equals or exceeds a certain level sa. The second element is the damage measure (D). Usually the lateral drift can be used to describe the structural and nonstructural damage level. Given that the analyzed structure is a single level one the top displacement was chosen as damage measure. The third element is the structural capacity (C). To be able to directly compare the capacity with the demand the capacity was expressed in the same form as the damage measure. As mentioned above Sa can be related to the hazard level using hazard functions. These functions, usually available for different sites, are developed by seismologists. To establish a relation between the intensity measure, Sa, and the damage measure, D, is the duty of the structural engineer. This structure specific damage hazard function provides the annual probability that the damage measure, D, will exceed a certain level, d: HD (d) = P D d S a = x dH( x )
(1)
If the damage hazard function is combined with the capacity, C, one can determine the annual probability of the performance level not being met, PPL: PPL = P[C d] dHD(d)
(2)
The above relations are not convenient for common engineering practice. To simplify them, one can assume that, in the vicinity of the value that makes the capacity to be reached, the median demand D can be approximated by a power function:
b D = a (S a )
(3)
Incremental dynamic analysis (IDA) was considered to relate the damage measure with the spectral acceleration and to determine the a and b values. Also, for Sa values situated in the vicinity of value corresponding to the capacity, the hazard curve can be represented by:
H(S a ) = k o (S a ) k
(4)
where, ko constant depending on the site seismicity k logarithmic slope of the hazard curve Considering that the damage measure, D, and the capacity, C, are both lognormal distributed about their median, D and C , and following the assumptions made above, the mean value of PPL can be found:
C = H (S a ) e 2 1 h
2
PPL or,
(5)
1 k2 2 2 2 2 ( DR + DU + CR + CU ) 2 b2
(6)
h
b
DR
DU
CR
CU
standard deviation in hazard coefficient that represents the demand increase as a function of hazard (b=1 is consistent with the equal displacement rule) standard deviation in demand due to randomness in response. This value was determined by performing several nonlinear dynamic analyses using different site-specific ground motion records. These records were scaled to result in the same spectral acceleration at the fundamental period of the structure. Then DR was computed by making statistics of the demand values assuming them to have a lognormal distribution. standard deviation in demand due to uncertainties in the model. The procedure used to determine it is similar to the one used for DR only that, this time, a single ground motion have been used for the investigation of several analytical models of the real structure through nonlinear dynamic analysis. These models were generated using sample data of the uncertain parameters considered to have a decisive influence on the structural response. standard deviation in capacity due to randomness in the response. To compute CR several values of the capacity, C, associated with different ground motion records should be determined for a single analytical model. standard deviation in capacity due to uncertainties in the model. To determine CU, the capacity, C, was computed using several sets of sample data of the uncertain parameters considered to have a decisive influence on the structural capacity. Statistics of C values were performed assuming them to have a lognormal distribution around their median.
Sa
H(Sa ) = k o (Sa )k
D = a (Sa )b
If the PPL is set to be equal to the performance objective, Po, the relation above yields to a more convenient load resistance design factor (LRDF) checking format as follows:
1 k2 2 ( CR 2 + CU 2 ) 2 b
e or,
C e
1 k2 ( DR 2 + DU 2 ) 2 b2
DS a
Po
(7)
o C DS a .
(8)
Sa
500
450
450
400 650
700
600 N = 899 kN b)
N = 337 kN a)
N = 203 kN
N = 1256 kN
Fig. 4. Columns dimensions and reinforcement details: a) 12.00 m span b) 24.00 m span
600
10 9
2
10 9
8 7 6 5 4 3 2 1 0 0.000
8 7 6 5 4 3 2 1 0 0.000
0.250
0.500
0.250
0.500
a)
10 9
b)
50%
STDEV
0.5000
0 0.000
0.500
a)
b)
10 9 8 7 6 5 4 3 2 1 0 0.000
2. 11 0 D = 0 .0035 S a
10 9 8 7 6 5 4 3 2 1 0 0.000
1 . 647 D = 0 . 0091 S a 1 . 773 D = 0 . 0091 S a
2
D = 0. 022 S a
1. 023
D = 0 . 023 S a 1 . 012
0 . 909 D = 0. 0304 S a
0.500
0.500
a) b) Fig. 7. Power functions fitted on the median IDA curve a) 12.00 m span b) 24.00 m span
(ii)
Statistics of these curves have been performed, considering a lognormal distribution of the response values. The mean and the median curve together with the standard deviation in demand due to the randomness in response, DR, are presented in Figure 6. The power function which serves for the approximation of the median demand, D , given in Eq. 3, was determined for each level of interest of the ground motion intensity. The results are presented in Figure 7. To compute DU and CU, the concrete compressive strength and the yield strength of the longitudinal reinforcement have been selected as the uncertain parameters having a decisive influence on the structural response. Their cumulative distribution functions were used to generate 30 sample sets. To compute DU, for each set the analytical model characteristics such as the yielding moment, the initial stiffness and the strain hardening ratio have been computed. This resulted in 30 analytical models for each analyzed structure. For each model, the nonlinear dynamic analysis provided the peak values of the top displacement. Then, assuming a lognormal distribution of these values, the standard deviation in demand due to uncertainty in the model, DU has been computed. The next step is the calculation of the dispersion in the capacity due to uncertainty in the model, CU. For each combination (fc, fy) the yielding and ultimate bending moments and the corresponding curvatures u and y at the bottom sections of the columns have been computed. The ratios u/y for the central column of each structure computed for different samples (fc,fy) are displayed in Figure 8. Assuming a common, simplified, curvature distribution over the height of the columns, the top displacement capacity has been computed for each model. CU has been computed by performing statistics of the computed capacity values assuming them to have a lognormal distribution around their median. The standard deviation in capacity due do randomness in response, CR, had to be determined also. The displacement criterion (DM based approach) was selected to define the attainment of the desired performance level. This means that the capacity is considered to be attained if the DM equals (or exceeds) the value associated with the considered performance level: DM CDM . As a result CDM does not depend on the ground motion and CR is equal to 0. The hazard level is represented by the hazard curves for the city of Bucharest. (Lungu & al. 2002) To apply the procedure spectral accelerations associated with different mean return intervals had to be determined together with the logarithmic slope of the hazard curve, k. Assuming that the hazard representation can be approximated, at least in the vicinity of the Sa value that makes the capacity to be reached, by a power function (Eq. 4) k can be determined by considering two convenient hazard levels around the desired Sa and their corresponding spectral acceleration values. H ln 1 H 2 k= S ln 2 S 1 where: H1 and H2 hazard levels S1 and S2 spectral acceleration associated with H1, H2
(iii) (iv)
(v)
(vi)
(vii)
(9)
(viii)
o The final step consists in the computation of the C / D Sa ratios for each ground motion intensity level. This express the confidence factor associated with the ground motions characterized by the mean return intervals mentioned above. The results are displayed in Table 1. In Figure 9 is represented the variation of the confidence factor corresponding to a continuous variation of the mean return interval. One can see that the capacity of the two analysed structures corresponds to the spectral acceleration values associated ground motion with mean return intervals of 110 and 100 years, respectively.
24-32 16-24 8-16 0-8 32 24 16 8 26.13 23.13 20.13 313 17.13 fc (Mpa) 353 393 433 14.13 473 513 553 593 11.13 633 20
15 10 5
26.13 23.13 20.13 17.13 313 353 393 433 14.13 473 513 553 593 11.13 633
u / y
u /y
fc (Mpa)
fy (Mpa)
fy (Mpa)
a)
b) Fig. 8. u/y ratios for the central columns. a) 12.00 m span b) 24.00 m span
500 450 400 Mean Recurence Interval 350 300 250 200 150 100 50 0 0.000 0.500 1.000 1.500 2.000 2.500 Mean Recurence Interval
500 450 400 350 300 250 200 150 100 50 0 0 0.5 1 1.5 2 2.5 3 Confidence factor Confidence factor ,
a) b) Fig. 9. Confidence factors for different recurrence intervals a) 12.00 m span b) 24.00 m span Table 1: Analysis results
12.00 m span structure Hazard Mean 1k 2 SA (from D C ( CR + 2CU ) Level Return e median) (median) (median) e 2 b (Median) Period 20 0.0500 3.195 0.076 0.165 0.987 50 475 0.0200 0.0021 4.486 7.646 0.108 0.255 0.165 0.165 0.981 0.984
1k 2 DR + 2 DU 2b
24.00 m span structure Hazard Mean SA (from D C 1k 2 Level Return ( CR + 2CU ) median) (median) (median) e 2b (Median) Period 20 0.0500 3.033 0.072 0.131 0.994 50 475 0.0200 0.0021 4.342 7.558 0.107 0.250 0.131 0.131 0.995 0.991
1k
e 2b
2 DR
+2 DU