L. Teresa Guevara and Luis Enrique García
PUBLISHED IN THE PROFESSIONAL JOURNAL OF THE EARTHQUAKE ENGINEERING RESEARCH INSTITUTE, EARTHQUAKE SPECTRA, VOLUME 21, NUMBER 1, FEBRUARY 2005
The accidental modification to the original structural configuration leading to a captive column by restricting its freedom to deform laterally due to the presence of nonstructural elements that partially confine it is presented. The case of short columns subjected to earthquake effects is also discussed. Examples of damage due to these effects in numerous earthquakes are presented, and the architectural decisions leading to captive and short columns are reviewed. The structural explanation of the behavior is discussed. Experimental research related to short and captive columns is presented. Recommendations to handle this type of problem are given.
1. INTRODUCTION
Earthquake damage reports, with few exceptions worldwide, present numerous cases of captive-column effect. Although the problem shows itself as damage to the column, the cause usually rests with nonstructural elements imposing a pattern of response to the earthquake motions different from the expected behavior of the column by itself without the nonstructural elements. The root of the problem of this mode of response is associated with each of the professionals involved in defining the location, dimensions, and structural properties of a column, each of them looking at it from their own professional perspective. Generally, the problem originates in the architectural design of the building. The best solution to the problem is to ensure that architectural designers and construction contractors understand the problem and avoid creating the condition.
Contractors need to understand the problem because frequently the condition is created after the building is occupied, when contractors add partial height walls between columns at the request of a building owner without the input of an architect or engineer. For these reasons this paper begins with a nontechnical explanation of the effects of these conditions and shows many examples of consequent damage. Engineers can use these examples to explain to architects, contractors, and owners the dangers of creating such conditions. The remainder of the paper provides the results of analytical and experimental research that provides a solid basis for why these conditions should be avoided.
Captive- and short-column conditions can also be caused by initial design errors in the detailed configuration of structural elements. The conditions are more frequently caused by interaction of structural and nonstructural elements that has not been taken into account in the analysis and design of the structure, or by later insertion of nonstructural elements between columns that create the conditions. The term ‘‘non-intentionalstructural element’’ (Guevara 1989) was coined precisely to describe this situation. Harmful response can only be avoided by separating the nonstructural elements from the structure or by taking into account the interaction in the analysis and design of the building.
2. CAPTIVE-COLUMN EFFECT AND ARCHITECTURAL FEATURES THAT CAUSE IT
The captive-column effect is caused by a non-intended modification to the original structural configuration of the column that restricts the ability of the column to deform laterally by partially confining it with building components. The column is kept ‘‘captive’’ by these components and only a fraction of its height can deform laterally, corresponding to the ‘‘free’’ portion; thus the term captive column.
Figure 1 shows this situation. Frequently during analysis, the structural engineer does not take into consideration the effect of the nonstructural elements in the response of structure. In most cases the analysis and design of the column is performed using the total clear height of the column because it is common to act on the assumption that the structure is free to sway laterally without interacting with the nonstructural elements.
The apparently inoffensive character of the nonstructural elements causes this common error in judgment, leading to unexpected and undesirable effects, as illustrated in Figures 2 to 3. Architectural decisions based on functional or aesthetic aspects are the most common reasons for the creation of captive columns.
The need for incorporating openings to the walls of a building in order to provide natural lighting and ventilation leads to partial lateral confinement along the height of the column by rigid elements, such as internal partitions, facades, retaining walls, and other elements.
The column ends up having adjoining walls in all its height, except in the upper part where the opening is located. The length of the column that would be free to deform laterally is reduced from the vertical floor to ceiling distance to just barely the height of the opening, as shown in Figures 2 and 3.
Location of High, Narrow Windows on Tall Windowsills
Openings above the line of sight are used when there is a perceived need to provide lighting and ventilation while restricting visibility from one space to the other. This type of configuration is often found in school classrooms, storerooms, rest rooms, doctors’ consulting rooms, and so on. In these cases, the nonstructural walls are higher than the height generally allowed for normal windowsills, and in order to comply with ventilation and lighting regulations the high windows extend from column to column (see Figures 4 and 5).
Although the strength of the nonstructural masonry walls may be lower than the strength of the column, in many cases, under lateral deformations, the resulting ‘‘nonstructural’’ walls are sufficiently stiff to affect the column behavior. The confinement provided by the nonstructural walls to the lower part of the column is so effective that usually damage to the short upper section of the column occurs before the confining wall fails.
Figure 6 shows how, when the captive column fails, the beams and the slab displace laterally in relation to their original position, while the windows have disappeared with the slab now resting on the ‘‘non-intentional structural walls.’’ Figure 7 shows how, in the same school, the outer column of the corner frame deformed but did not collapse because it was not laterally restrained by the walls.
Figures 8 to 10 show captive-column effect in the building of ‘‘Empresas Pu´blicas de Pereira,’’ Colombia, produced by the 23 November 1979 earthquake. Figure 10 illustrates an inner view of an undamaged captive column shown at the left edge of Figure 8. This column was subjected to the same lateral displacements but did not fail. In this specific case, the reason was that columns shown in Figures 8 and 9 were restrained by solid brick walls while the column in Figure 10 was restrained by walls built using much weaker clay tile block that failed at the interface without affecting the column. The captive-column effect was present only where the glass block failed.
Open Corridors in Building Complexes
A partial confinement of the clear height of the column is also common in housing complexes built during the 1950s in numerous countries. In this type of building, the configuration that followed the architectural trend of the Modern Movement (International Style), the corridors are left open to the fac¸ade. Instead of employing light, transparent handrails, heavy and stiff partial-height parapets were used, thus forming a captive-column configuration (see Figures 11 and 12). Figure 12 shows column failures due to captive-column effect in an external corridor of a housing complex with the same configuration as the building in the previous figure.
Buildings on Sloping Grounds
Partially Buried Basements
The captive-column effect is present also when partially buried basements are employed. The common practice is to use retaining walls with the columns embedded in them up to the street level, with only a small part of the height of the column continuing up to the next story slab. These openings provide ventilation and natural lighting to the basement, as shown in Figures 15 and 16.
3. SHORT-COLUMN EFFECT AND ARCHITECTURAL FEATURES THAT CAUSE IT
Framing at mid-height of the column of horizontal structural elements such as slabs, beams, and girders divides the column in two segments, thus producing the shortcolumn case. Although the terms captive column and short column have been used interchangeably in the literature, the reasons that cause them are completely different. In the former case, as explained before, the column is affected by the presence of adjoining nonstructural elements. In the latter case the column is made shorter than neighboring columns by horizontal structural elements, such as beams, girders, stairway landing slabs, and ramps, that frame at mid-height of the column, as shown in Figures 17 and 18.
Other possibilities of short-column adverse configuration appear in frames located at the transition in split-level structures where the frame has beams located at half the usual vertical clear height in order to support slabs at alternating sides. Short columns are also caused by having a story of the structure with all columns much shorter than neighboring stories. This situation appears at the foundation, where foundation grade beams interconnect columns above the footing, leaving a gap between the grade beam and the footing.
A variant of this case of short columns is present when atypical clear height floors are employed for sanitary or mechanical reasons. Figure 19 shows the vulnerability of this configuration in the 10 October 1980 El-Asnam, Algeria, earthquake. Numerous buildings had a one-meter crawl space under the first floor called a ‘‘sanitary story’’ to install plumbing and provide ventilation under the first-floor slab. This configuration converts into short columns all columns of the frame at the sanitary story level, resulting in numerous buildings’ losing this story by failure of all columns.
4. ASSOCIATED STRUCTURAL BEHAVIOR
In the most common case of captive-column effect when the column is restrained by adjoined nonstructural walls, column and walls interact, restricting the lateral deformation of the column. The upper free column segment is then responsible for accepting the deformation that the full height of the columns was designed to sustain. To acquire some insight on the captive-column effect from the structural perspective, we can divide the effects on the frame—and its columns—into those caused by gravity loads and those caused by lateral forces.
The deformations of the frame elements vary in shape and magnitude, as shown schematically in Figure 20 for gravity loads and lateral forces. The ends of a column in a frame subjected only to gravity load (Figure 20a) remain basically plumb, unless the frame is extremely irregular. In the case of lateral load (Figure 20b), the upper end of the column displaces horizontally with respect to the lower end a distance denominated story drift (D). The order of magnitude of the column lateral deformations is significantly less for gravity loads than for lateral load. In a schematic and simplified manner.
Figure 21 shows the deformations (d) with respect to the original position of the undeformed column, the internal forces of the element—flexural moment M, axial force P, and shear force V—and the moment diagrams for both cases of gravity loads and lateral forces. In the column that is part of a frame subjected only to gravity loads, Figure 21a, the lateral deformations of the column depend only on the magnitude of the applied moments and the flexural stiffness of the column (d 5de). For the column belonging to a frame subjected to lateral forces (Figure 21b), the lateral deformations of the column depend on the sum of two factors (d 5de1dd): the first one (de) relates in the same fashion to the flexural moments and the stiffness of the column, while the second one (dd) depends directly on the story drift (D).
The story drift is a function of the stiffness of the story and the structure, the geometry of the frame, the mass of the structure, and the earthquake motion. The individual column flexural stiffness plays a minor role in the order of magnitude of the story drift. Technical literature is rich in procedures for obtaining the internal forces and the general deformations of the frame for both cases. The relationship between the internal flexural moments that act at the column ends and the shear force associated with them can easily be obtained through the application of equilibrium principle and disregarding the P-Delta effect that may be significant for large lateral deformations:
Thus, the shear force V corresponds to the algebraic sum of the moments at the ends of the column (Ma1Mb), divided by its clear height h. In the captive column, due to the presence of a restraining element external to it, the clear height is significantly reduced, increasing the shear force in inverse proportion. To give an idea of the order of magnitude involved, for a typical 2.5-m-story clear height the presence of 2-m-tall nonstructural walls creating a 0.5 m opening in the upper part will form a lateral restriction that will increase fivefold (2.5/0.555) the shear force that the column has to resist as compared to the shear force computed for the column without the nonstructural wall restriction.
The sad lesson, learned again and again in every earthquake, is that the relatively rigid nonstructural element has the power to control the shear force the column must resist!
If this is the case, the first question that comes to mind is why doesn’t the captivecolumn effect cause problems in the gravity load cases? In reality the problem is there, but since the magnitude of the flexural moments is small, the lateral deflections of the column are also small, and only in extreme cases does the column feel the restriction imposed by the nonstructural wall; the problem is minor, if it exists at all.
The sad lesson, learned again and again in every earthquake, is that the relatively rigid nonstructural element has the power to control the shear force the column must resist!
If this is the case, the first question that comes to mind is why doesn’t the captivecolumn effect cause problems in the gravity load cases? In reality the problem is there, but since the magnitude of the flexural moments is small, the lateral deflections of the column are also small, and only in extreme cases does the column feel the restriction imposed by the nonstructural wall; the problem is minor, if it exists at all.
This is not the case under lateral load where the flexural moments are large, the corresponding lateral deflections are also large—mainly because they are controlled by the story drift—and the presence of the nonstructural wall is felt by the column from the onset of the lateral deformation inducing the extremely large shear forces that produce the observed column failures. With regard to reinforced concrete structures subjected to strong earthquake motion effects, one of the premises of modern earthquake resistant design philosophy is to let the structure respond in the nonlinear range at levels well beyond the deformations that will cause yielding of the longitudinal reinforcement of the elements. This response in the inelastic range produces an energy dissipation through flexure that diverts part of the energy that induces vibration in the structure, thus allowing the structure to survive the earthquake motions without having the full strength that would be required if the energy dissipation did not occur. For this dissipation of energy to occur, two fundamental premises must be observed, among others:
• The concrete must be able to accept strains well beyond the values that will cause failure of the material under normal circumstances. This is achieved by ample use of confining transverse reinforcement at critical locations within the structural elements, and
• The structural element should not fail in shear before the energy dissipation takes place.
This is achieved by the adequate use of transverse shear reinforcement along the full length of the element. In order to warrant compliance of the second premise, modern earthquake resistant design codes require that the design shear force (Ve) of the element should be obtained from the use of the probable flexural moment strengths (Mpr) at the ends of the element. The probable moment strengths must be obtained, in turn, employing the actual longitudinal reinforcement area at the faces of the element, a yield strength for the reinforcing steel equivalent to 1.25 times the nominal yield strength, and a strength reduction factor (f) equal to one. The design shear force (Ve) is then obtained using Equation 1 and the appropriate values, as described in Equation 2:
The above-described procedure tries to avoid the existence of structural elements that would fail in shear before reaching the capacity to dissipate energy in flexure. Unfortunately, the existence of captive columns caused by nonstructural elements not taken into account by the structural engineer when applying Equation 2 defeats the whole purpose of this procedure. The value of h to be employed in Equation 2 must be consistent with the actual deformation restraints applied by the structural and nonstructural elements. Now, turning our attention to the captive-column cases described previously where horizontal structural elements frame into the column at mid-height or columns are made shorter by sloping ground, the same principles just described can be employed.
The flexural stiffness of a column is inversely proportional to its clear height. As the clear height is halved (in reality it is made even less than half due to the depth of the horizontal framing element), the lateral stiffness of the column increases in inverse proportion. When the lateral force story shear is distributed to all columns in the same floor, shorter columns will be called upon to resist a larger portion of the story shear than normal height columns. Although most frame-analysis procedures detect this shear difference, the shear strength required for the column responding in the nonlinear range must be determined from the probable flexural strength at the ends of the column as described by Equation 2. This means that the shear forces obtained from analysis have no relation whatsoever to the required shear strength, and if the shear design of the column is made using the analysis results, the shorter portion of the captive column will be extremely vulnerable under earthquake motions.
This vulnerability exists in all captive columns designed under codes enacted before the shear-related-to-flexural-strength requirement was introduced. The fact that the shear strength of the element should be computed from the flexural strength at its ends was first proposed by Blume et al. (1961). It was introduced in Appendix A of the ACI-318 code in the 1971 issue (ACI Comm. 318 1971) and was adopted by the Uniform Building Code in 1973 (ICBO 1973). Based on this, any captive column designed and built before the mid 1970s would be suspect of being vulnerable in shear, and this explains the disproportionate number of cases of captive-column failures observed during earthquakes in old buildings and structures not built following modern seismic codes. This situation can also be present even in more recent structures built following modern codes, where interaction with nonstructural elements was not taken into account in design or when stiff nonstructural elements are introduced by the owner or occupant without assessing its potential harmful effect.
5. EXPERIMENTAL RESEARCH ON SHORTAND CAPTIVE COLUMNS
The number of experimental research projects on short and captive columns does not correspond to the number of times the problem is observed in the field during earthquakes and to how widespread the problem is in general. The reason behind this is related to the fact that actual solutions from the structural point of view are of dubious effectiveness and the general approach has always been to term it as an architectural problem that can be solved through proper education of the architectural community at large. Nevertheless, before trying to propose solutions that only highlight the ‘‘interface’’ root of the matter, it is important to have a feeling of the behavior of the short and captive column as observed in experimental tests. Two research programs are worth describing within this context. The authors make no claim of giving an exhaustive listing of all the published research on the problem, and describe the following experimental programs solely as paths toward finding an understanding of the problem.
Tests at the University of Texas at Austin
During the 1980s at the University of Texas at Austin, a series of experimental research programs designed to establish the parameters that control the behavior of structural short columns were conducted. The aim of the research project was to establish design parameters that could warrant an appropriate behavior of the column when subjected to earthquake imposed loads. The experimental program and the main conclusions obtained are described in Mayurama et al. (1984), Umehara and Jirsa (1984), and Woodward and Jirsa (1984). Several short columns were tested under imposed lateral cyclic deformations. The main conclusions that can be derived from this experimental program are as follows:
• When the behavior of short columns with no axial load were compared with short columns with axial loads below the balanced axial load, it was found that the axial load increased the stiffness and the lateral load strength of the column. On the other hand, this axial load increased the strength degradation for cyclic loading once the maximum lateral load resistance was reached.
• When comparing cyclic test with monotonic tests, the strength of the column under cyclic loads decreased once the maximum lateral load resistance was reached.
• When comparing columns with different transverse reinforcing spacing (65 to 300 mm), shear strength was found to be insensitive to the transverse reinforcement spacing and depended solely on the shear strength of the concrete alone. This factor, by itself, could explain the brittleness of these elements during earthquakes.
• It was practically impossible to obtain a stable flexural inelastic response of the columns, and all of them exhibited hysteretic unstable responses. Columns with less longitudinal reinforcements exhibited a better response than those with greater longitudinal reinforcing ratios.
A confirmation of the findings at the University of Texas had been observed during the 12 June 1978 Miyagi-ken Oki earthquake in Japan. Nonstructural but monolithically cast concrete walls restrained the deformation of a slender reinforced concrete column and caused the lateral deformation to concentrate in a short length as shown in Figure 22. The column had transverse reinforcement as required by the then current code.
Test at the Universidad de los Andes at Bogota, Colombia
In 1994 J. C. Pineda performed experimental tests, at the Structures Laboratory of the Universidad de los Andes in Bogotá, Colombia, trying to reproduce captive-column failures in 1:3 scale models (Pineda 1994). In total, three models were tested. Figure 23 shows the captive-column failure observed in Model 1. The other two models were used to test captive-column solution schemes.
The solution schemes were aimed at finding solutions from the nonstructural element point of view by adding wall elements that would close the opening near the column, as shown in Figure 24. Figure 25 shows the final state of cracking for model 2.
The conclusions of the project indicated a way to avoid captive-column failures only for those cases where the frame containing the captive column is able to control the lateral deformations of the whole structure. The proposed solution was simply defending the column with the addition of masonry inserts at both sides of the column, closing the gap that causes the captive-column effect and allowing the compression strut in the masonry wall to travel along the masonry wall plane, thus diverting away the critical shear force from the reinforced concrete column.
The proposed recommendation derived from these concepts was to add masonry inserts with a horizontal length of twice the gap opening that produces the captive-column problem, with the masonry covering the whole height of the column. It was also recommended that the wall should be checked for the imposed loads to guarantee that it will not fail along the compression strut, and the columns should be able to resist the forces imposed by the strut at the corners of the panel.
6. CONCLUSIONS AND RECOMMENDATIONS
Historically, the more damaging effect of interaction of reinforced concrete frames with nonstructural elements has been the captive-column effect. This type of wall arrangement is common in educational buildings or other buildings where the window is provided for lighting/ventilation purposes only. This situation introduces large shear stresses, when the structure is subjected to lateral forces that are not accounted for in the standard frame design procedures.
The best solution for captive columns is to avoid the problem. Any type of nonstructural element that could hinder the free deformation capacity of a vertical structural element should be located in a different acting plane than the structural element, or must be separated from the structural element by appropriate joints. When the isolation option is adopted, the designer should guarantee the out-of-plane lateral stability of the wall. For split-level buildings, in order to circumvent the short-column effect, the architect should avoid locating a frame at the vertical plane where the transition between levels occurs. For buildings on slopes, special care should be exercised to locate the sloping retaining walls in such a way that no captive-column effects are induced.
Where stiff nonstructural walls are still employed, these walls should be separated from the structure, and in no case can they be interrupted before reaching the full height of the adjoining columns. The architect must study carefully the use of nonstructural components in order to avoid unwelcome interaction with the structure of the building. The structural designer should address the influence of masonry infill walls in the lateral force behavior of the structure, either by taking them into account in the design process or by a separation gap from the column. If a separation gap is provided, then appropriate measures should be taken to warrant the out-of-plane stability of the masonry when subjected to lateral forces from wind or earthquake.
The requirements for obtaining a remodeling permit from a construction authority should provide prescriptions for the owner, the architect, and the contractor on what to do when a captive-column effect is present in a project for remodeling a building. The ‘‘Essential Requirements for Reinforced Concrete Buildings’’ document (ACI et al. 2002) developed under an agreement between the American Concrete Institute, the Colombian Institute for Technical Standards and Certification, and the Colombian Association for Earthquake Engineering provides some recommendations regarding the two alternative corrective measures discussed.
L. Teresa Guevara. M. EERI. Director, Proyectos V&G Consulting Architects, Caracas, Venezuela. Guest Associate Profesor, Facultad de Arquitectura y Urbanismo, Universidad Central de Venezuela (FAU-UCV). email: tguevarap@hotmail.com
Luis E. García. M. EERI. Professor of Civil Engineering, Universidad de los Andes, Bogotá, Colombia. Partner, Proyectos y Diseños Ltda. Consulting Engineers, Bogotá, Colombia. email: legarcia@cablenet.co
7. REFERENCES
[1] Guevara, L. T., (1989), Architectural considerations in the design of earthquake-resistant buildings: influence of floor-plan shape on the response of medium-rise housing to earthquake. Ph.D. in Architecture Dissertation. Berkeley: Graduate Division, University of California, Berkeley, p. 38.
[2] García, Luis E., (1991), Columnas de concreto reforzado, (in Spanish), Universidad de los Andes, Serie Selecta de Asocreto, Bogotá, Colombia, p. 192. (In Spanish.)
[3] Asociación Colombiana de Ingeniería Sísmica (AIS), (1998), Normas colombianas de diseño y construcción sismo resistente - NSR-98 (Ley 400 de 1997 y Decreto 33 de 1998) , (in Spanish), Bogotá, Colombia, 2 Vol. (In Spanish.)
[4] Blume, J. A., N. M. Newmark, and L. H. Corning, (1961), Design of Multistory Reinforced Concrete Buildings for Earthquake Motions, Portland Cement Association, Skokie, IL, 318 p.
[5] Committee 318, (1971), Building Code Requirements for Reinforced Concrete (ACI 318-71), American Concrete Institute, Detroit, MI, 102 p.
[6] International Conference of Building Officials, (1973), Uniform Building Code (UBC-73), ICBO, Whittier, CA, 704 p.
[7] Mayurama, K., H. Ramírez, and J. Jirsa, (1984), Short reinforced concrete columns under bilateral load histories, Journal of Structural Engineering, Vol. 110, 1, American Society of Civil Engineers, January, pp. 120-137.
[8] Umehara, H., and J. Jirsa, (1984), Short rectangular reinforced concrete columns under bi-directional loading, Journal of Structural Engineering, Vol. 110, 3, American Society of Civil Engineers, March, pp. 605-618.
[9] Woodward, K., and J. Jirsa, (1984), Influence of reinforcement on reinforced concrete short column lateral resistance, Journal of Structural Engineering, Vol. 110, 1, American Society of Civil Engineers, January, pp. 90-104.
[10] Pineda, J. C., (1995), Ensayos experimentales sobre control de columnas cortas, (in Spanish), Proyecto de Grado IC-94-II-26, Advisor: L. E. García, Departamento de Ingeniería Civil, Universidad de los Andes, Bogotá, 43 p. (In Spanish.)
[11] American Concrete Institute - ACI, Instituto Colombiano de Normas Técnicas y Certificación - Icontec, and Asociación Colombiana de Ingeniería Sísmica - AIS, (2002), Essential Requirements for Reinforced Concrete Buildings, International Publication Series 1, American Concrete Institute, Farmington Hills, MI, 246 p.
[12] Guevara, L. T. and L. E. García, (1999), La columna corta o columna cautiva, (in Spanish), Revista Noticreto, 52, Asociación Colombiana de Productores de Concreto – Asocreto, julio-septiembre, Bogotá, pp. 46-54. (In Spanish.)
