Vous êtes sur la page 1sur 7

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

85

Evaluation of the ASCE 41 Linear Elastic Procedure for Seismic Retrofit of Existing Structures: Pros and Cons of the Method Luis A. Toranzo-Dianderas1
1

KPFF Consulting Engineers, 6080 Center Drive, Suite 300, Los Angeles, California 90045; PH (310) 665-1536; FAX (310) 665-9070; email: ltoranzo@kpff-la.com

ABSTRACT The implementation of linear elastic procedures for the seismic evaluation and retrofit design of existing structures has been developed with great detail in ASCE 41. Although the procedures are intended to cover the most important aspects of the evaluation and design process, some specific aspects may not be as transparent as needed for adequate results, and may require modifications or comment clarifications in upcoming editions of the standard. This paper is intended to highlight the pros and challenges of implementing linear static procedures for seismic evaluation and retrofit design of structures, particularly when compared with the use of more advanced nonlinear based analysis techniques.

INTRODUCTION ASCE 41 was published in 2006 culminating the effort initiated with FEMA 273/274 in 1997 to create a set of consensus recommendations for performance based seismic rehabilitation of existing buildings. As its predecessors (FEMA 273 and FEMA 356), in general, ASCE 41 presents two approaches to accomplish this task: linear elastic based and nonlinear based methods. The linear elastic methods are rather a more sophisticated version of traditional practice, and most designers would feel they have the knowledge and tools to implement it. The nonlinear based approaches are relatively new and more demanding, both in resources and background knowledge, but have been gaining wider acceptance in the design community thanks to the availability of faster computers with improved storage capacity, as well as greater awareness of the benefits of the use of advanced analysis in the determination of retrofit designs. Even though there is broad agreement that the nonlinear based procedures are a better tool for implementation of performance based seismic engineering, the linear elastic methods are and will continue to be used due to the relative simplicity in its implementation. It is important therefore, to maintain a critical view on the method, understand its limitations and keep searching for ways to improve the details of the process. The intent of this paper is to convey related views and findings and to clarify and suggest improvements on particular aspects.

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

86

BRIEF DESCRIPTION OF THE ASCE 41 LINEAR ELASTIC BASED RETROFIT METHODS The ASCE 41 linear elastic based method for seismic rehabilitation of existing buildings can be briefly summarized as follows: 1. Define a retrofit scheme based on ASCE 41 guidelines or other criteria and create a linear elastic model of the retrofitted structure; 2. Determine which structural components are deformation controlled and which are force controlled; 3. Apply an unreduced seismic demand to the linear elastic model; 4. On deformation controlled members, calculate the unreduced seismic component demand and compare it with expected capacities modified by specific factors assigned according to the expected inherent member ductility (m factors); 5. On force controlled members, calculate a reduced seismic component demand and compared it with the lower bound capacity of the component; 6. If the component demand is greater than the component capacity, further retrofit actions are required until the component capacity is equal or greater than the component demand. The intent of the method is to control the post-elastic demand on ductile structural members (deformation controlled components) and prevent the failure of any critical or brittle members (force controlled components) of the structure. PROS OF THE LINEAR ELASTIC APPROACH Simple Modeling Preparing a linear elastic model is significantly less demanding than creating a nonlinear model. One needs only to define elements that capture the initial effective stiffness of the components of the lateral system. Besides the geometry, it usually suffices to know the different materials elastic modulus to have a complete working model of the structure. It is expected that a model defined in this way should capture the initial distribution of the forces which will be used for the evaluation and/or design of the individual structural members. Aspects that are disregarded in a linear elastic model include stiffness and strength degradation; determination of locations of nonlinear regions; determination of cyclic degradation if a nonlinear time-history is conducted; etc. One should be able to create an acceptable linear elastic model using only the information found on a set of structural drawings. Expeditious Computing Effort Compared to a nonlinear model, the time required to run a load case on a linear elastic model could be minuscule. In the experience of the author, a linear elastic model that runs in a matter of a few minutes to complete a set of load cases,

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

87

could take hours, or even days to complete a single load case if an equivalent nonlinear model is used. Unlimited Flexibility for Load Combinations We tend to take for granted the fact that linear elastic results can be amplified, reduced, reversed and combined, to obtained results for alternative load scenarios. The results of a few basic runs can be later modified to account for any load combinations of interest. When using a linear elastic model, one only needs to run a few basic load cases in order to obtain any seismic related load combination: Dead Load (D) Life Load (L) Seismic Demand in Direction X (SeismicX) Seismic Demand in Direction Y (SeismicY)

These four basic load cases can later be combined not only in the main orthogonal directions (X and Y) but also used for multidirectional load combinations like the ones below: (D + 0.25L) + 1.00 SeismicX + 0.30 SeismicY (D + 0.25L) + 0.30 SeismicX + 1.00 SeismicY

In contrast, nonlinear elastic models require that every single load combination be run individually. One can not use basic load cases to determine alternative combinations as with linear elastic models. The task could be daunting if the number of load combinations is not determined sensibly. As a reference, attempting to implement all load combinations required by ASCE 41 for nonlinear static procedures leads to hundreds of individual runs, a task completely impractical that should be managed at the beginning of the process by reducing the load combinations only to the more relevant cases. CHALLENGES OF THE LINEAR ELASTIC APPROACH Strong Reliance on Nonlinear Response The linear elastic method on ASCE 41 relies on certain level of ductility on individual members of the lateral system to sustain the expected seismic demand. The m factors listed for different members and load conditions reflect the ductility that is acceptable for a given performance level. The values listed for m, as expected, can vary widely depending on the component and type of action. The intent is as described above is sensible, however, to assume that those individual ductility demands can be calculated using the results of a linear elastic model of the subject building could be risky. As it can be observed when conducting any nonlinear based analysis, the occurrence of a nonlinear event in a structure leads to a redistribution of the internal actions which only in simple cases can be estimated by inspection. In most structures,

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

88

the redistribution can be captured to a large extent with an explicit nonlinear analysis, and even then, one needs to acknowledge the uncertainty present on the process. The use of linear elastic methods therefore, should be limited to structures where no major redistribution actions can be expected. Two checks in ASCE 41 are apparently intended to avoid the major redistribution scenario. One is to limit the maximum demand capacity ratio observed in any individual component. This is expected to avoid an early and sustained nonlinear event that may lead to major redistribution processes in the structure. The second is to limit the complexity of structures that can be analyzed using the linear elastic methods. Structures with critical deficiencies such as in plane discontinuity, out of plane irregularity, weak story irregularity and torsional strength irregularity are directed to a nonlinear based procedure. Although ASCE 41 as well as it predecessors FEMA 273 and 356 include commentaries to different sections of the standard, this important aspect lacks a comprehensive commentary and should be included in subsequent editions. Miss to Address Overall Post-Elastic Mechanism The linear elastic procedure lends itself to focus on fixing only components where the analysis shows unacceptable demand capacity ratios. Approaching the problem in this fashion, ignoring the overall seismic response of the structure, may lead to a retrofit scheme that is lacking an adequate overall collapse mechanism. Even though the verification of the collapse mechanism is not embedded in the ASCE 41 linear elastic process, it is important to address this aspect to have a reasonably efficient energy dissipation capacity of the overall system. This aspect is particularly important on existing structures with limited ductility and potentially poor detailing. One should remember that the system ductility assigned to different structural systems in new buildings is based on the premise that a code conforming baseline design should have all the detailing that induces certain collapse mechanisms over others. This inherent condition on new structures is not necessarily present on existing structures in need of retrofit. Evaluation of Members Subject to Interacting Actions The ductility m factors are used in conjunction with the element demand and the element nominal or reduced capacity to control the level of component ductility in individual members. The structural demand of the individual elements can be obtained from the linear elastic analysis performed for the structure. For elements which are subject to actions that are not affecting each other, the determination of the nominal capacities should be a straight forward calculation based on any of the reference design codes (ACI, AISC, etc.). However, on elements that sustain actions that are affecting each other, the determination of the nominal capacity requires a definition of the scenario at which the component would reach this nominal capacity. A common condition would be the evaluation of columns where the axial/flexure capacity is determined by an interaction diagram. Axial actions are treated as force controlled whereas flexure actions are deformation controlled. The

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

89

flexure capacity, therefore, can be amplified by an m factor to meet Eq. 3-20 of ASCE 41, reproduced below. mkQCE >= QUD Where the Q action in this case is the flexure moment M on the column. The CE subscript stands for expected capacity and the UD subscript indicates unreduced demand used for evaluation of deformation controlled components. The parameter k is the knowledge factor which affects both deformation controlled and force controlled members. The equation above therefore would become: mkMCE >= MUD The axial action, or P axial force, on the other hand has to be evaluated using Eq. 3-19 of ASCE 41 for force controlled members, which is reproduced below. kQCL >= QUF In this case the Q action would be the axial force P. The CL subscript stands for lower bound capacity and the UF subscript indicates reduced demand used for evaluation of forced controlled components. kFCL >= FUF What we have then are two different methods to define demand, and two different ways of determining capacity, for the evaluation of an interactive axial/flexure condition. Needless to say, this approach creates confusion when determining what the appropriate axial/flexure pair one should be using when evaluating them on an axial/flexure interaction diagram, and what interaction diagram should one be using, the one with expected values or the one with lower bound values? A good understanding of the implications of the components reaching their capacities is necessary to make a sensible decision. A commentary on ASCE 41 discussing this aspect would be beneficial. Diaphragm Evaluation and Retrofit Design ASCE 41 has a narrow definition of the elements that can be used as chord and collectors on the evaluation of diaphragms. This definition should be expanded to include any elements of the structure that can contribute to the in-plane capacity of the diaphragm. Typically, in new structures, chords and collectors in diaphragms are supplementary elements explicitly designed for in plane diaphragm demand. However, in existing structures that may not have been engineered in that way, the diaphragm should be evaluated as a unit, including all available capacity, as the gravity reinforcement may have enough residual capacity to sustain the in-plane demand.

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

90

This problem is not addressed either on the nonlinear based procedure of ASCE41 but is exacerbated in the linear elastic method since the demand is not limited by the actual capacity of the lateral system but is set arbitrarily by an m factor that should be used in the analysis. Subsequent editions of ASCE41 should address this aspect since diaphragm repair may be a very invasive retrofit action that could be avoided if most of the available existing capacity is accounted for. Potential Design Inconsistency in Transition Regions Ductility factors m are different for every component and load condition. This approach leads to inconsistency in transition regions where the expected demand may be different at both sides of a border region. The approach makes sense when the intent is to protect one critical or brittle component at expenses of forcing yielding or other type of post-elastic mechanism in a more ductile adjacent component. But this not always results on a balanced design. Take a shear wall under life-safe evaluation, which has been designed for an m factor of 2, sitting on a foundation where the demand can be reduced by a factor R equal to 8. The end result will be a wall with a capacity about 4 times greater than the capacity that can be sustained by the foundation. Obviously, this creates an unbalanced design where a wall is designed for a force that can not be transferred by the foundation. This aspect could be addressed limiting the demand on any element to the maximum force (including overstrength) that can actually be delivered to/by the adjacent structural component. This is one example of inconsistencies that can be found, and should be expected, when trying to control the post-elastic response of a structure using a linearelastic model. A comment in the code allowing flexibility in extreme cases like the mentioned above would be helpful. Drift Estimation There is not explicit procedure to estimate drift when using the linear elastic methods in ASCE 41. One could follow the equal displacements rule and assume that the expected drift should be that calculated using the unreduced seismic demand applied to the linear elastic model. The calculated drift is expected to be the same for any range of overall building ductility. Following this approach may be acceptable for medium to long period structures, however, short period structures may need a modification similar to that implemented on the coefficient method on the nonlinear based procedure. A commentary addressing this aspect would be beneficial. CONCLUSIONS Benefits and challenges of the linear elastic based method proposed in ASCE 41 have been discussed in this paper. Although the method is significantly simpler to implement and manage than the nonlinear based methods in ASCE 41, it presents shortcomings that should addressed by the design community. In general, this method should be limited to simple, regular structures, especially since the opportunity to

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

ATC & SEI 2009 Conference on Improving the Seismic Performance of Existing Buildings and Other Structures

91

implement nonlinear based analysis is now within reach of most of the design engineering professionals. REFERENCES ASCE-41 (2006). Seismic Rehabilitation of Existing Buildings.

Copyright ASCE 2009


Downloaded 28 Mar 2012 to 144.122.102.163. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org

Vous aimerez peut-être aussi