performance evaluation of structures · structural concrete (aci committee 318, 2005). this method,...

27-1

27Performance Evaluation

of Structures

Richard A. Miller, Ph.D., P.E.*

27.1 Introduction ......................................................................27-127.2 ACI 318-08 Provisions on Strength

Evaluation of Existing Structures.....................................27-227.3 Pretest Planning for Reliable Structural Evaluation .......27-427.4 Nondestructive Testing for Material

and Structural Assessment................................................27-6Ultrasonic Testing • Infrared Thermographic Testing • Modal Testing

27.5 Static/Quasi-Static Load Testing ......................................27-9Use of Dead Loads • Vehicle Testing • Testing Using Servohydraulic Cylinders

27.6 A Discussion of Instrumentation and Data Acquisition ......................................................27-13Important Issues in Instrumentation Selection •Types of Instrumentation

27.7 Case Studies in Performance Evaluation of Concrete Structures ....................................................27-21Testing of a Three-Span Slab Bridge • Truck Load Testing of Damaged Concrete Bridges: Testing for Continuity • Testing of Shear-Key Cracking in Adjacent Box-Girder Bridges • Testing of a High-Rise Building after Collapse • Testing of Concrete Panels on a Steel Stringer Bridge

References ...................................................................................27-31

27.1 Introduction

Concrete structures are difficult to evaluate using theoretical models. These models require precise defi-nitions of material properties, support conditions, and the stiffness of individual members. All of theseare difficult to define for concrete structures. Other than compressive strength, test data on materialproperties are often not available, and when such data are available, they are likely to exhibit considerablescatter. Support conditions are not easily defined, and the stiffness of a cracked concrete member can bedifficult to determine with any precision. For most structures, the fact that theoretical models cannotpredict structural exact behavior is not a problem. The use of proper safety factors and years of experience

* Professor of Civil Engineering, University of Cincinnati, Cincinnati, Ohio; expert in nondestructive testing andevaluation of existing bridges.

© 2008 by Taylor & Francis Group, LLC

27-2 Concrete Construction Engineering Handbook

in designing and building concrete structures have led to safe and economical designs; however, for certainstructures, theoretical evaluation is inadequate. Structures with suspected design or constructions flaws,damaged or deteriorated structures, or with unusual design features require a performance evaluation.

Performance evaluation can be defined as the destructive, partially destructive, or nondestructivephysical test of a structure, structural system, or structural subsystem for determining critical responsesunder one or more loading conditions. Performance testing is used for evaluating service load behavior,determining cracking load or ultimate load capacity, determining stiffness or flexibility, detecting damageor deterioration, and evaluating the effects of the damage or deterioration, the effects of support condi-tions, fatigue or cyclic load behavior, and response to impact or vibrational loads. Performance tests canbe divided into two groups: tests for quality control (including verification of load capacity) and tests toassess structural performance when existing theories are inadequate or nonexistent.

When most people think of performance evaluation, they usually think of strength evaluation—spe-cifically, the method described in the American Concrete Institute’s Building Code Requirements forStructural Concrete (ACI Committee 318, 2005). This method, discussed in Chapter 20 of the ACI Code,is intended to evaluate structures where there is some doubt that the structure meets the safety require-ments of the Code. ACI 437R-03, Strength Evaluation of Existing Concrete Buildings (ACI Committee 437,2003), presents a more in-depth discussion of structural strength evaluation. The report defines severalcases where the strength-evaluation techniques discussed in the report may be useful:

• Distressed structures that show signs of damage due to overload, fire, vibration, etc.• Deteriorated structures that show signs of damage through material degradation such as corrosion,

excessive cracking of the concrete, spalling, etc.• Structures suspected of containing understrength materials or to be substandard in design or

construction• Structures where the capacity to hold loads applied in the future is in doubt because the original

design or construction data are not available• Structures for which a change in usage would apply loads in excess of the original design loads• Assessment of structures undergoing repair, retrofit, or strengthening• Structures that require testing by order of a building official

The committee recognizes other areas where performance testing may be valuable but does not recom-mend using the strength-evaluation method presented in the report for these areas:

• Structures with unusual design concepts• Product development testing used for approval or quality control for mass-produced elements• Evaluation of foundations or soil conditions• Structural research

Other possible areas, not mentioned in the report, where strength or performance testing may be usefulinclude:

• To assess the condition of a structure, especially when hidden damage or deterioration is suspected• To verify the results of analytical modeling

This chapter attempts to present performance testing in the broadest possible terms.

27.2 ACI 318-08 Provisions on Strength Evaluation of Existing Structures

The method described in Chapter 20 of the ACI Code (ACI Committee 318, 2008) is the one mostcommonly used for performance evaluation of concrete structures. It makes sense to begin any discussionof performance evaluation with this method. Provisions on the strength evaluation of existing structures


Performance Evaluation of Structures 27-3

are intended to evaluate the safety of a structure for which the load-carrying capacity is in doubt (Section20.1). If the effects of the strength deficiency are well understood and the structural dimensions andmaterial strengths are easily verified, analytical evaluation of the structure is permitted (Section 20.1.2);otherwise, load testing is required (Section 20.1.3). If the structural safety is suspect due to deterioration,Section 20.1.4 seems to imply that load testing is also required, as it says that deteriorated structuresmeeting the load test requirements may remain in service. Section 20.1.4 allows the engineer to requireperiodic reevaluation if deemed necessary.

If load testing is to be undertaken, Section 20.3 requires the structure to be loaded such that anysuspect areas are subjected to maximum stress and deflection. In some cases, a single load placementmay not simultaneously maximize all of the critical responses and multiple load placements must beused to demonstrate the adequacy of the structure. The ACI 318-08 Code stipulates that the total loadtest including dead load already in place should not be less than the larger of the U value determined byEquation 27.1:

U = 1.15D + 1.5L + 0.4(Lr or S or R) (27.1a)

U = 1.15D + 0.9L + 1.5(Lr or S or R) (27.1b)

U = 1.3D (27.1c)

where D = dead load, L = live load, Lr = roof live load, S = snow load, and R = rain load. The load factoron live load in Equation 27.1b can be reduced to 0.45 for garages, areas occupied as places of publicassembly, and all areas where L is greater than 100 lb/ft2.

When conducting a load test, the testing sequence is as follows:

• An initial measurement is taken of the important structural responses: deflections, strains, crackpatterns, crack openings, rotations, etc. Of all these responses, only deflection and cracking areused for acceptance, but the provisions recognize that other responses, although not required,may be of interest. These responses must be measured not less than 1 hour before the test.

• The load is applied in four equal increments. Between each increment, structural response ismeasured. Section 20.4.4 requires the structural responses to be measured after each load incre-ment is applied, and the Commentary (R20.4.2) suggests inspecting the structure after each loadincrement is applied. No maximum or minimum time limit between applications of the loadincrements is specified. Section 20.4.3 requires that uniform loads be applied in such a way thata uniform load distribution is obtained. In some cases, loads may be applied incorrectly, resultingin an uneven load distribution. The most common case, cited in the Commentary, is when concreteblocks are used for loading. As the structure deflects, the blocks may rotate and touch at the top,creating an arch. Subsequent loads are carried by these arches, and nonuniform load distributionsresult.

• When the total load is in place, it is left for 24 hr. Structural response measurements are taken atthe beginning and end of this period.

• After the 24-hr period, all structural responses are measured, and the load is immediately removed.The structure is left for 24 hr, and the structural responses are measured at the end of the 24-hrperiod.

The tested area cannot show any signs of failure. Crushing of the concrete is considered failure.After removal of the load, the structure must recover at least 75% of the maximum deflection:

(27.2)

where ∆r is the deflection 24 hr after removing the load, and ∆1 is the deflection under load after the loadis in place for 24 hr. Because very stiff structures have small deflections, it may be difficult to accuratelymeasure deflection recovery, so the code also accepts a structure if:

∆ ∆r = 14



(27.3)

where lt is the span of the member (taken as the smaller of center to center of supports or clear span + h)and h is the member thickness. For two-way slabs, lt is the shortest span.

Structures that fail the deflection part of the test may be retested after 72 hr and accepted if:

(27.4)

where ∆2 is the maximum deflection during the second test.The structure cannot exhibit excessive cracking, spalling, or crushing, which may indicate failure. The

structure fails if inclined cracks appear, which may indicate a shear failure. Any inclined cracks in areasof a member without transverse reinforcement must be evaluated if the crack is longer than the depthof the member. For members with a variable depth, the depth at the midpoint of the crack is used. Inanchorage zones or in areas where there are lap splices, short inclined cracks or horizontal cracks mustbe investigated, as these may be signs of a bond or anchorage splitting failure.

It is clear that ACI 318 requirements are narrowly tailored to evaluate structural safety against failure.If other structural responses are desired, such as performance and serviceability under working load,changes in structural response due to damage or deterioration, etc., then broader performance-testingtechniques are necessary.

27.3 Pretest Planning for Reliable Structural Evaluation

The most critical phase of performance evaluation is the pretest planning. In this phase, a detailed planor critical path should be established not only to ensure that the tests are performed correctly but alsoto be sure that all critical responses are measured. Pretest planning can be broken down into severaldistinct general tasks:

• Determine the specific purpose of the test and desired final product. This is probably the mostoverlooked part of pretest planning. Prior to beginning the test, the following items should beevaluated:• Determine the specific desired outcome. Possible outcomes could be service load behavior,

proof of ultimate load levels, or behaviors that may indicate the presence of damage, deterio-ration, possible deficiencies in materials, or construction flaws. Before beginning any test, it isimportant to have a clear picture of the final outcome as this determines the choice of testmethod, load levels, and instrumentation.

• Determine the test method that best meets the goal. When evaluating a structure, engineersoften choose load testing as a first, or only, alternative. In fact, there are many ways to evaluatea structure other than just load testing (e.g., modal testing).

• If load testing is chosen, be sure the loading method is realistic for the structure. Uniformloads are appropriate for a general structure but are not a good choice for testing bridges astraffic loads tend to be point loads. Be sure the loading method will produce the desired results;for example, due to the difficulty of applying a moving load for many cycles, stationary cyclicloads are often used to simulate repeated traffic loads, although stationary, cyclic loads maynot produce the same response as an actual moving load (Petrou et al., 1994).

• Determine the type of data required to accurately assess the structure. In many performancetests, only deflection is measured, which may not provide sufficient data. Often strains, rota-tions, differential movements, temperature changes, or accelerations may also have to bemeasured. Instrumentation should be properly chosen to provide the necessary data.

∆max,

≤ lh

t2

20 000

∆ ∆r = 25



• Gather all available design and construction information. When the purpose of the test has beenestablished, the next step is to gather all available design and construction information to providea basic background of the structure and a baseline for comparing actual material test results andactual dimensions. Necessary information on design loads, material specification and strengths,design drawings, construction records, as-built drawings, and records of any modifications to thestructure must be assembled and carefully examined to form a complete picture of the structure.• It is important to obtain as much information as possible about the as-built condition of the

structure as variations and deviations from the original plans are not infrequent. Obtainingas-built drawings and copies of inspection records is highly recommended.

• Reports on actual material strength are also of great importance, especially in destructive orpartially destructive tests. It should be kept in mind that material specifications are usuallyminimum or maximum properties (as applicable), and the actual materials may have substan-tially greater or lower properties than required. For example, ASTM A 36 steel has a minimumyield strength of 36 ksi, but in reality the yield strength can be much higher, with values of 50ksi not being unusual.

• Conduct a pretest inspection of the structure and material sampling. For many structures, theconstruction records maybe inadequate. The records have been lost or destroyed or the inspectionsmay have been insufficient, so detailed inspection and construction records do not exist. In suchcases, a pretest inspection of the structure is necessary.• The structure should first be checked to verify all dimensions. Several measurements of member

sizes should be taken at critical sections. Where possible, member support conditions shouldbe checked. At this time, a detailed survey of surface and visible defects should be made. ACI201.1R-97, Guide for Making a Condition Survey of Concrete in Service (ACI Committee 201,1997), provides a good method for making such a survey.

• Reinforcing bar size, spacing, and cover should be documented. Three common nondestructivemethods for measuring the reinforcing bar size, cover, and spacing are magnetic, radiographic,and radar (ACI Committee 437, 2003; Malhotra and Carino, 2004). Bar size and location canalso be found by removing the cover in isolated areas. This last method is often used to calibratenondestructive testing (NDT) methods. Where practical, bar samples should be removed forfurther evaluation, including strength tests and assessment of the degree of corrosion, if present.

• Concrete strengths are normally determined by coring (ASTM, 2004a), as this seems to be theonly accurate method for determining in-place concrete strength. Guidance on using cores todetermine in-place strength is found in ACI 214.4R-03 (ACI Committee 214, 2003). Nonde-structive tests for determining concrete compressive strength include the rebound hammer orprobe penetration, but these methods do not directly measure strength, and they show a highdegree of variability, especially when different types of concrete are being tested. As a result,the methods require calibration with drilled cores for each type of concrete being tested, andeven then they are of questionable accuracy. Tensile strengths can be found by performing thesplit cylinder test (ASTM, 2004b) on cores or by use of sawed beams (ASTM, 2004a), if practical.

• In addition to cores removed for strength testing, cores should also be taken for petrographic(ASTM, 2004c) or chemical testing. Petrographic and chemical testing can reveal potentialweaknesses in the concrete due to alkali–silica reaction (ASR), inhomogeneity, bleed, or seg-regation; poor air-void systems that allow freeze–thaw damage (ASTM, 2006); reinforcing steelcorrosion; abrasion; fire; D-cracking of aggregates; and weathering. It is also necessary tosample and test auxiliary materials such as overlays, bearing pads, material in attached struc-tures, etc. This is especially necessary when such materials may affect loading, boundaryconditions, and structural or material behaviors.

• When sampling the material, the question arises as to how many samples are necessary to yieldreliable results. The answer to this question is based on the practicality of sampling the materialand the amount of error that can be tolerated. In general, the number of samples is affectedby the following: (1) The amount of possible damage that would occur when samples are



removed; the attraction of nondestructive testing is that any number of samples can be takenwithout damaging the structure. (2) The cost, in money and time, to perform the test; oftentesting must be limited due to time and budget constraints. (3) The importance of the data;for example, the flexural strength of reinforced concrete is only slightly affected by the com-pressive strength so less accuracy can be tolerated in determining the compressive strength ofa flexural member. On the other hand, the strength of axial members is greatly influenced bycompressive strength, so a more accurate determination of compressive strength is necessary.To determine the number of samples required, any of a number of statistical techniques canbe applied. ASTM C 823 provides information on developing a program for material sampling(ASTM, 2000).

• Perform a pretest analytical investigation. The data collected from records and inspection can beused to create an analytical model of the structure. This model will provide necessary informationfor designing the test. In the case of load tests, the model will provide information on probablecracking, yield, and ultimate load capacities. The model can also be used to pinpoint criticalresponses so the instrumentation can be properly placed to record these responses during theactual test. In the case of dynamic tests, frequencies, accelerations, magnification (impact) factors,and vibrational mode shapes can be estimated from the model.

• Design the testing and instrumentation system. The test can be designed using the results from themodel. In the case of load testing, it is necessary to choose the type of load, load levels, loadposition, and method of loading (e.g., trucks, blocks, hydraulic). Load-reaction frames for hydrau-lic-loading cylinders also have to be designed at this point.

• Lay out instrumentation grids based on the analytical results so the instruments are in place to capturecritical responses. The model results are also used to estimate the magnitude of the response so aninstrument with a proper range and accuracy is chosen (see Section 27.6). The data acquisitionsystems should be chosen at this time.

• Perform the test and evaluate the results. With a properly designed test, performance of the testbecomes a relatively simple task. Time spent up front in planning will greatly reduce the chanceof problems occurring during the test and will lead to a successful outcome.

Unfortunately, evaluation of the results is not an easy task. Even in tests that have excellent pretestplanning, ambiguities in the data often must be reconciled. In such cases, it is useful to create post-testmodels to assist in data evaluation.

27.4 Nondestructive Testing for Material and Structural Assessment

In pretest evaluation, it is necessary to carefully evaluate the condition of the material and the structure.Some of the previously stated methods are partially destructive or inspect only a small portion of thestructure. For large-scale detection of flaws, other methods are needed. Of the many nondestructivemethods for evaluation of concrete structures, only a few are presented here. A complete treatment ofNDT methods can be found in Malhotra and Carino (2004).

27.4.1 Ultrasonic Testing

Sound waves have been used for testing concrete for many years. Many states still employ the chain-dragmethod for concrete bridge decks. A large metal chain is dragged across the concrete surface, subjectingthe surface to small impacts, and delaminations in the concrete are found when a hollow sound is heardduring the dragging of the chain. The limitations of this method are obvious: It only detects flaws nearthe surface, an experienced person who can differentiate the sounds is needed, and it cannot providenumerical data on the size and depth of the flaws. More modern methods of testing concrete utilize



ultrasonic waves transmitted through the concrete. A simple ultrasonic test, the pulse-velocity test, consistsof creating an ultrasonic pulse, detecting the same pulse at a remote receiver, and measuring the time ittakes for the pulse to travel the distance. Changes in the pulse transmission time can represent changes inthe properties of the concrete; for example, as young concrete gains stiffness, the pulse transmission timedecreases. Increases in pulse transmission time often indicate the presence of damage, as the pulse cannottransmit through a crack and instead must take a longer path around the damaged area. The impact-echotechnique provides a more quantitative measurement of flaws in the concrete (Carino and Sansalone, 1992;Carino et al., 1986). In this method, a small impactor is used to create a wave in the concrete. When thewave hits an interface, either a free surface or a crack, it is reflected back to a receiver. By processing thereturned signal, the depth and size of flaws can be determined (Carino and Sansalone, 1992).

27.4.2 Infrared Thermographic Testing

Infrared thermographic testing works on the principle that flaws in concrete will affect the way in whichheat flows through the concrete. This, in turn, affects the surface temperature distribution. The concreteis either heated up or cooled down, and surface is photographed using a camera that is sensitive toinfrared radiation. Damage to the concrete is seen as cooler or hotter areas (Delahaza, 1996; Weil, 1993).Often, simply allowing a concrete structure to heat up during the day due to sunlight or to cool downat night produces sufficient temperature variations to make infrared thermographic techniques usable.One advantage to the themographic method is that it can be used in a moving system to examinepavements without having to stop or detour traffic. Several case studies outlining the use of thermographictesting for pavements are given by Weil (Malhotra and Carino, 2003).

27.4.3 Modal Testing

When evaluating a concrete structure, it is beneficial to have information about the actual stiffness orflexibility of the structure. Changes in flexibility over time can be used as a measure of global deterioration,and local changes in flexibility can be used to pinpoint local damage. For modeling purposes, the accuracyof a model can be determined by comparing the actual flexibility of the structure with the flexibilityobtained from the model. One method of determining the actual flexibility of a structure is polyreferencemodal testing. Modal testing can be used by itself as a nondestructive method of structural evaluation.It can also be used as a pretest evaluation procedure for planning a load test of a structure, as modaltesting can identify damaged, deteriorated, or suspect areas of the structure (Aktan et al., 1992). Afterthe load test, the flexibility matrix obtained from modal testing can be used to verify the results of theload test and to calibrate finite-element models.

In the modal-testing method, structural vibrations are produced by an impact (Aktan et al., 1992;Allbright et al., 1994). (Forced vibrations have also been used, but this method is more complex so onlythe impact method will be discussed here.) The structural vibrations resulting from the impact are thenmeasured by seismic accelerometers. By knowing the time histories of the impact (measured with a loadcell) and the corresponding accelerations distributed through the structure, the modal frequencies, modalvectors, and damping can be calculated. Modal testing offers the advantage of not having to calculatethe mass and stiffness of the member; instead, vibrational characteristics are measured and used directly.From the magnitude and frequency of the vibrations, the natural mode shapes and frequencies can bedetermined, and these can be used further to calculate a flexibility matrix.

Prior to conducting a modal test, an approximate theoretical or finite-element analysis should be madeto provide a clear understanding of the structure for the purpose of establishing impact points andaccelerometer (reference) points. The results of approximate theoretical or finite-element analysis areused so the accelerometers are not placed at grid points that would correspond to the nodes of the firstfew significant modes. The analysis is also used to establish the bandwidth of expected frequencies, whichis necessary so the dynamic data acquisition system will read only within the band of frequencies ofinterest.



In theory, the test can be performed with a single accelerometer, but in practice a minimum of threenoncolinear reference points are used, although more reference points will often provide better data(Aktan et al., 1992; Allbright et al., 1994). Impact to the structure is supplied either by a sledgehammeror a falling weight. In both cases, the impactor must be instrumented with a load cell so the exact impactload time history is known. Instrumented hammers are commercially available. Each grid point (includ-ing the reference points) undergoes impact, and the resulting vibrations are measured by the accelerom-eters. The impactor load cell and the accelerometers are read using a dynamic analyzer (these are alsocommercially available). A fast Fourier transform (FFT) is used to transform both the input (impactor)data and the output (accelerometer) data from the time domain into the frequency domain. The frequencyresponse function (FRF), which is basically the ratio of the output FFT to the input FFT, is then calculated.The calculation of the FFT and FRF will normally be performed by the dynamic analyzer, although theoperator should have some degree of expertise to ensure the quality of the final data.

When impacts are applied to the structure, care must be taken to ensure that the impactor does notbounce and create multiple impacts while the acceleration data are being taken, as this would corruptthe acceleration data. Also, in some cases, the impact is too hard or too soft and the resulting data areoutside the range of the data acquisition system. Such tests should be discarded and redone. This is whyit is essential to use a dynamic analyzer capable of processing the results in real time.

Each time a point is struck and acceleration data are taken, a single dataset is created for that point;however, one dataset per impact point is not enough to yield reliable results, so each point is testedseveral times to collect multiple datasets, the results of which are then averaged. In general, it is necessaryto have at least five good datasets per point. When the dynamic data have been obtained for all impactpoints, they can be transferred to a commercial program that will then calculate the modal frequencies,modal vectors (mode shapes), and damping. Anomalies in the frequencies or mode shapes can sometimebe used to identify damage (Allbright et al., 1994), but the flexibility matrix is usually required for anaccurate evaluation:

(27.5)

where:

[F] = flexibility matrix.[ψ] = unit-mass scaled modal vectors.[1/ω2] = diagonal matrix of ascending natural frequencies.

A method for calculating the flexibility matrix from mass-scaled modal vectors can be found in Catbaset al. (1997).

Modal testing does not provide an exact flexibility matrix, as an infinite number of modes would berequired to calculate the exact flexibility; however, if a sufficient number of modes can be detected, themodal flexibility matrix can be close enough to the actual flexibility to provide highly accurate results.When the modal flexibility matrix has been obtained, it can be used to evaluate the structure:

(27.6)

where:

[P] = vector of point loads placed at impact points.[F] = modal flexibility matrix.[δ] = vector of displacements at impact points.

Any loading pattern can then be simulated by placing point loads at the corresponding impact points.(A uniform load is simulated by placing point loads at all impact points.) This shows one of the advantagesof modal testing. Often, damage to a structure is localized and can only be detected through load testingif the load combination and load placements are ideal to reveal the damage. In a real load-test situation,

FT

=

ψ ωψ1

2

P F = δ



this would require an unrealistic number of load cases; however, with the modal flexibility matrix, a largenumber of loading patterns can be efficiently checked to see if any damage is detected. For damagedetection, it is best to have a baseline modal flexibility matrix of the undamaged structure. Damage canthen be detected by comparing deflections from subsequent modal flexibility matrices to deflectionsfound from the baseline matrix. If a baseline matrix is not available, deflections found from the modalflexibility matrix can be compared to deflection generated from an idealized model of the undamagedstructure (e.g., finite-element model), but this comparison is less accurate. Finally, the modal flexibilitymatrix can be used to fine-tune the stiffnesses and boundary conditions of a finite-element model. Here,the deflections generated from the model are compared to those generated from the modal flexibilitymatrix, and the model is adjusted to obtain the best comparison for several load cases. The tuned modelcan then be used for analysis of the structure.

27.5 Static/Quasi-Static Load Testing

Static/quasi-static load testing is the most common form of structural testing. Because many structural,loads are static in nature, a static or monotonic quasi-static load test can provide valuable informationon stress, strain, load distribution, and deflection under normal loading conditions. Quasi-static cyclicloading is often used to assess fatigue behavior. For moving or dynamic loads, cyclic quasi-static loadtesting is sometimes used as a simpler or less expensive alternative to using actual moving or dynamicloads; however, the validity of using cyclic quasi-static loading to evaluate moving or dynamic loads isquestionable (Chung and Shah, 1989; Petrou et al., 1994).

27.5.1 Use of Dead Loads

One method of conducting static-load test consists of placing a dead load on the structure. Typical deadloads are concrete blocks, bricks, sandbags, or containers of gravel, sand, or water. The advantages ofusing dead loads are as follows:

• They can easily simulate a uniform load on a structure. A single layer of 200-mm (8-in.) hollow-core concrete blocks with 25-mm (1-in.) gaps in between to prevent arching can provide a uniformload of 1.4 kN/m2 (29 psf).

• The loading materials are relatively inexpensive. In many cases, bricks, blocks, crushed stone, orsand can be borrowed from building supply yards or quarries.

The disadvantages of this type of testing are numerous and often relate to safety:

• Except for exposed structures such as roofs or bridges, the dead load often cannot be placedremotely (as by crane) and must be placed by hand. This presents a danger to workers should thestructure fail while load is being applied.

• Should the structure begin to crack or collapse, the loads cannot be removed quickly or safelyenough to prevent additional damage or complete collapse.

• Should the structure completely or partially collapse, the load may fall freely and perhaps injureworkers or unintentionally damage other parts of the structure.

• The actual load on the structure can only be assessed through the cumbersome process of weighingthe individual loading blocks or containers, and there is no easy way to record the load on moderndata acquisition systems.

• This method cannot be used to simulate moving or fatigue loads.

For safety reasons, dead load testing should be limited to service load tests well below damage orcollapse limit states. ACI 437R-03 (ACI Committee 437, 2003) allows the use of dead load testing but,because of safety issues, does not recommend it. When using dead loads, several practices should beobserved. All loading blocks or containers should be of the same weight; ACI Committee 437 allows a



±5% variation between individual units. The loads should be placed with gaps between the individualloading blocks or containers. This is because the structure will deflect under load, and if the loadingblocks or containers are too close together they may touch and form an arch. The arch formations makethe actual loading pattern uncertain. Loose material (sand, gravel, water) should not be used for loadtesting as it may gather in low areas of the structure and pond, again creating an uncertain loadingpattern. Any loose material used for loading should be placed in containers and the full containers usedto apply the load.

27.5.2 Vehicle Testing

Vehicle testing is preferred for service load testing of structures such as bridges or parking decks wherevehicle loads are the predominant live loads. Vehicles can be empty or loaded with sand or gravel; thus,loaded dump trucks are very effective for applying loads (Aktan, et al., 1993; Dimmerling et al., 2005;Shahrooz et al., 1994a,b). Vehicle axle or single wheel weights can be obtained using truck scales such asthose found at concrete plants, building supply yards, or highway department weigh stations. Many statehighway patrols have portable scales that can be used. Vehicular load testing has several advantages:

• It is the type of load that would exist on bridges, parking decks, garage floors, etc. in service loadconditions.

• Because trucks can often be supplied by highway departments or local contractors and the sandor gravel to load them can be borrowed from concrete plants or building supply yards, vehicularloading is a relatively inexpensive testing method.

• Vehicular loading can be used to apply static, moving, or impact loads to a structure.

The disadvantages include:

• As with dead loads, vehicle loading requires workers (drivers, in this case) to actually place theload on the structure. Although movable, vehicular loads usually cannot be removed quicklyenough to prevent injury to the drivers in the event of sudden cracking or collapse.

• Dump trucks are often used for vehicle loadings as they are usually available from contractors orstate departments of transportation. Most structures designed to carry vehicles are usually designedto carry vehicles much heavier than a typical dump truck. In many cases, even a large number offully loaded dump trucks may not be heavy enough to cause large structural responses. It may benecessary to use large trucks loaded with very heavy loads (such as concrete blocks) to obtain alarge enough response to be meaningful (Issa et al., 1996).

• It is very difficult to apply the load in increments, as this would require slowly adding load to thevehicle.

• Due to possible clearance problems, it may be difficult to place the loads close together or nearbarriers.

• If multiple vehicles are used, it is extremely difficult to attain the same axle loads on all of thetrucks, even if similar trucks are used. In addition, the loads tend to shift during testing, affectingthe axle loads.

• Unless the vehicles are turned off, engine vibrations will affect the test results.• The load cannot be recorded electronically.

27.5.3 Testing Using Servohydraulic Cylinders

Hydraulic testing consists of applying load through one or more hydraulic cylinders or actuators. Thistype of loading is preferred (ACI Committee 437, 2003), as it is the safest method. Hydraulic loadingcan be as simple as using jacks controlled through hand-operated pumps or as complicated as usingsophisticated electronic, servohydraulic systems that use electric pumps and computerized controls. Theadvantages of hydraulic loading include:



• Hydraulic tests are safer. Loads can be placed on the structure via a remote-control mechanism,thus keeping workers safe should the structure collapse. Also, electronic hydraulic systems can beequipped with fast unloading circuits and safety shutdowns that will quickly remove all loads atthe first signs of collapse.

• In a load-control test where a structure is subjected to predetermined increasing or decreasingincrements of loads, a hydraulic-loading system can precisely control load increment, total load,and rate of load application.

• Use of servohydraulic, closed-loop control allows the use of displacement or strain-controlledtests, which are more stable than load-control tests and less likely to result in sudden collapse.This type of testing will often allow post-peak behaviors to be obtained.

• Hydraulic testing is the most efficient way to apply cyclic or fatigue loads.

The disadvantages of hydraulic testing include:

• Hydraulic systems apply point loads. Application of a line load or a uniform load over a smallarea can only be done using loading blocks or a spreader device (Azizinamini et al., 1992; Milleret al., 1994). Hydraulic systems usually cannot be easily adapted to apply a uniform load over alarge area.

• Computer-controlled servohydraulic systems and the associated control systems are expensive topurchase, maintain, operate, and transport.

• Skilled technical people are often required to handle the hydraulics and electrical work to assemblethe system.

• A reaction mechanism is necessary for the hydraulic system to push or pull against.

The need for a reaction mechanism is the biggest drawback of hydraulic systems. In some cases, anotherpart of the structure can be used as a reaction frame, but doing so introduces the possibility ofunintentional damage to other structural members. In many cases, a reaction frame tied to a foundationor a self-equilibriating testing frame can be built (Azizinamini et al., 1992). Another approach, shownin Figure 27.1, is to use post-tensioning tendons grouted into the soil or rock below the structure(Miller et al., 1994).

FIGURE 27.1 Use of post-tensioning tendons as a loading system.

12" stroke, double-ended,3000 psi, servo-control hydraulic actuators

Concrete load transfer block

Pier

WaterRockRock

Rock anchors

FillPile

Note: For clarity, effects of bridge skewness are not strictly accounted for.

Abutment

Overlay to be removed Reaction plate

8-strandrock-anchor cable350 kips each



Hydraulic-loading systems can be precisely controlled through the use of servohydraulic, closed-looptesting. In closed-loop testing, the structure is loaded such that some specific parameter (e.g., load,deflection, strain) increases or decreases at a set rate. To understand the usefulness of the closed-loopservohydraulic system and its limitations, it is first necessary to understand how such a system works. Aschematic of a typical closed-loop system is shown in Figure 27.2.

The system consists of five parts: a pump, a servohydraulic valve, a hydraulic actuator, a feedbackinstrument, and a controller. The pump supplies hydraulic fluid at a set pressure up to a maximum flowrate. The rate of flow of the hydraulic fluid is controlled by a servoelectronic (“servo”) valve, which is ahigh-speed, proportional valve capable of quickly and precisely varying the flow of hydraulic fluid throughthe system. The flow rate through the valve is directly proportional to the current input to the valve bythe controller.

The hydraulic actuator is a hollow cylinder fitted with a piston and rod. To apply load, hydraulic fluidis pumped into a chamber behind the piston. Some actuators are dual action, meaning fluid can bepumped into a chamber on either side of the piston so the actuator can exert force in either direction. Itis important to note that actuator capacities are based on the area of the piston and the maximum pressuresupplied by the pump. Quoted actuator capacities are usually based on a 21-MPa (3000-psi) pumppressure. If the pump being used in the test has a maximum pressure different than that given in theactuator manufacturer’s specifications, the quoted capacity will be different. Also, because one side of thepiston has the rod attached (thus reducing the area), actuator capacities are different in each direction.

During operation, fluid is pumped into the chamber on one side of the piston. If the actuator is dualaction, the chamber on the other side of the piston is opened to allow fluid to flow out. If the rod isattached to a flexible structure, pumping fluid into the chamber produces a combination of pressure inthe chamber (which translates to load on the structure) and movement of the piston, which translatesto deflection of the structure. Unloading is accomplished by releasing fluid and pressure from the activechamber and pumping fluid into the other chamber.

The load applied by the actuator can be accurately measured by measuring the hydraulic fluid pressureand multiplying by the piston area. This method can often be more accurate and less cumbersome thanplacing a load cell between the actuator and the structure. Pressure can be measured by pressure trans-ducers that can be directly linked into a computer data acquisition system. In dual-action actuators, thereis fluid on both sides of the piston, and both chambers have pressure; therefore, when attempting tocalculate load, the differential pressure between the two chambers must be used.

FIGURE 27.2 Closed-loop system.

Return

Pressure

Servohydraulic valve(s)

Hydraulic pump

Hydraulic lineElectrical cable

Controller

Emergencyshut-off

Servo cable

Hydraulic cylinder(s)

Feedbackinstrument

Feedbackcable

Pressure transducers



The feedback instrument is any electronic instrument that can read the desired structural response(e.g., deflection) and translate it into a voltage. The controller uses the signal from the feedback instru-ment to regulate flow through the servovalve. Normally, the controller is set to regulate two things: therate of the structural response and the maximum value of that response.

For the sake of example, assume that the system shown in Figure 27.2 is set to be a displacement-controlled test of a beam; that is, the system will control the load such that the beam deflects at a setrate up to some predetermined maximum deflection. Initially, the controller is used to pump some fluidinto the actuator, and a small load is applied to the beam. An electronic instrument on the beam readsthe deflection and converts it to a voltage that is transmitted to the servohydraulic controller. Thecontroller, which is programmed to convert the voltage back to displacement, checks the rate of deflectionagainst the target rate. If the rate of deflection is too slow, the controller will attempt to increase the flowto the hydraulic actuator to increase the loading rate and therefore the deflection rate. If the deflectionrate is too fast, the controller will attempt to decrease or, if necessary, reverse the flow to the hydraulicactuator to slow the deflection rate.

When the beam reaches the target maximum deflection, the controller stops the flow to the actuator;however, the system always has some electrical and mechanical noise, so a servohydraulic system will nothold at the specified target but will oscillate about it. This oscillation will introduce noise in the responseof the instruments attached to the structure, often giving the responses a sawtooth appearance. Oscilla-tions of the system can also be caused by backlash in the control instrumentation. Backlash is a measureof the change in response necessary to elicit a change the instrument reading when the response is reversed(see Section 27.6). Noise in the system will usually cause the controller to slightly overshoot the intendedtarget response (e.g., deflection, strain), so the controller will unload slightly to correct this condition;however, if the instrument has significant backlash, the response under unloading may not be detectedimmediately. The controller will miss the target and then attempt to correct by reloading. Thus, instru-ments with significant backlash will cause large oscillations in the system and should be avoided.

Servohydraulic systems provide an excellent means of conducting a controlled test of structures orelements. These systems, however, are extremely complicated, and a high level of expertise is required todesign, build, and operate such systems; therefore competent control engineers should be retained todesign and operate servohydraulic systems.

27.6 A Discussion of Instrumentation and Data Acquisition

The importance of instrumentation in a structural test is often overlooked. There is a tendency to assumethat the instruments and any associated equipment for data acquisition are transparent; that is, the datacoming from the instruments are not affected by the instrumentation itself or the data acquisition system.This is not true. Each instrument, whether a simple mechanical dial gauge or a sophisticated electronicdevice, has its own particular characteristics that can affect the test results. It is only when the test iscomplete and the data are found to be faulty that any problems with instrumentation are even considered.

27.6.1 Important Issues in Instrumentation Selection

When selecting instrumentation for a performance test, several issues must be addressed.

27.6.1.1 Range

It is useful to speak of the range of the instrument in terms of total range and working range. Total rangeis the maximum and minimum values of input that the instrument can read. Working range is theanticipated maximum and minimum values of response which will measured during the test. It isnormally desirable for the working range to be about 50% of the total range. If the working range is tooclose to the total range, any error in estimating the maximum or minimum responses may cause theinstrument capacity may be exceeded. Some instruments exhibit nonlinearities or other errors near thelimits of the range, so measurements near these limits should be avoided. Having a working range that



is very much smaller than the total range is also undesirable. Because the precision of the instrument isnormally a function of the total range, trying to make small measurements with an instrument with alarge range will usually provide imprecise results.

27.6.1.2 Precision

Precision is a measure of how reproducible a measurement is. Note that precision is an absolute number.A measurement can be precise, but incorrect (or inaccurate). If an instrument has the same error in eachmeasurement, the results may all be within a narrow band, and therefore precise; however, if the averagevalue is not close to the true value, the measurement is precise but inaccurate.

27.6.1.3 Accuracy

Accuracy can be defined as the difference between the instrument reading and the true quantity beingmeasured. Determination of the accuracy of an instrument is a complex and difficult process that requirescareful calibration of the instrument. The instrument accuracy is actually the sum of several instrumenteffects and errors. In general, the response of any instrument can be separated into five components,although for a given instrument some of these responses may not be present. The components are given by:

(27.7)

where:

Xtotal = total instrument response.X1 = response due to the load.X2 = response due to non-load effects on the structure.X3 = instrument error.X4 = error due to external effects on the instrument.X5 = error due to effects of instrument mounting.

In most cases, the total response of the instrument is assumed to be only the caused load, and theother responses are ignored. This can be a grievous error. It is necessary to remove, minimize, or at leastaccount for the other responses of the instrument. The response due to non-load effects (X2) includessuch things as changes in strain and deflection due to temperature. These effects can be significant, asprevious research has shown that changes in load distribution, strains, and deflections of an exposedstructure due to temperature can be greater than the response due to load (Miller et al., 2004). It ispossible to correct for non-load effects by analytical estimation of the values of these effects, by placingadditional instrumentation designed to specifically measure non-load related effects, or by performingtests under no applied load to specifically measure these effects.

Every instrument has a certain amount of error associated with its measurements, given by the X3term in the previous equation. The data acquisition system itself can cause additional sources of error.When considering the sources of error due to the instrumentation and data acquisition system, it isnecessary to consider the following:

• Resolution—Resolution is what most people are referring to when they speak of accuracy, but itis only one component of the total accuracy of the instrument. Resolution is the smallest valueof a response that can be registered by the instrument. For mechanical gauges, this is the smallestvalue, a single tick mark. It is possible to interpolate between two marks, but this not objective asit depends on the person reading the gauge. For gauges with digital displays, the resolution is thesmallest value on the display. Resolutions of analog electrical instruments are more difficult todetermine. In theory, these instruments have infinite resolution but in practice the resolution islimited by the resolution of the digital data acquisition system, electrical noise, and the behaviorof the mechanical components within the gauge. Thus, for analog electrical instruments theresolution is the smallest response that provides a reliable reading from the gauge.

X X X X X Xtotal = + + + +1 2 3 4 5



• Sensitivity—Sensitivity is the output of an instrument for a given input and is expressed as theratio of output to input (Figure 27.3). Usually, sensitivity is determined by a best-fit straight line.For electrical instrumentation, the output is often influenced by the excitation applied to theinstrument, so this is included in the sensitivity. As an example, consider an instrument thatmeasures displacement and output voltage and is excited by a voltage. The sensitivity may beexpressed as X volts/volt/mm. This means that a movement of 1 mm will cause the instrumentto outputs X volts if a 1-volt excitation is used. If a 10-volt excitation is used, a 1-mm movementwould output 10X volts. Sensitivity does not affect accuracy through the instrument itself, butrather through the data acquisition system. If an instrument has a small sensitivity, the outputwill be small, and the data acquisition system may not be able to read small quantities accurately.Also, small outputs in electrical instruments are often affected by noise.

• Linearity—Many instruments (although not all) are assumed to be linear; the output of theinstrument has a linear relationship with the response. The linearity error is the measure of howfar the data deviate from the best-fit line (Figure 27.2).

• Repeatability—Repeatability is the ability of the device to output the same value for the sameresponse over a number of trials when the response is always increasing or decreasing (e.g., alwaysextending the instrument, always heating the instrument) (Figure 27.3).

• Hysteresis and backlash—Hysteresis is the difference in a reading at a given point depending onwhether the reading was obtained by an increase or decrease in response (e.g., the difference inresponse when a given point is reached by extending an instrument as compared to the responseat the same point reached by contracting the instrument). Backlash is related to hysteresis and isa measure of the change in response required to elicit a change the instrument reading when theresponse is reversed (e.g., extension to retraction, load to unload, heating to cooling) (Figure 27.3).

Accuracy problems related to resolution and sensitivity can be avoided by proper selection of instru-ments based on reasonable estimates of the quantities to be measured. Again, this points to the needto perform proper pretest planning. The remaining errors—linearity, hysteresis/backlash, and repeat-ability—determine the maximum total error of the instrument. Linearity is found by providing a seriesof inputs over the range and plotting the input against the output (Figure 27.3). For best results, inputsshould be made in both directions (e.g., extension and retraction, loading and unloading). At leastthree trials in each direction should be made and the results plotted on a single graph. The maximumdistance from a best-fit straight line to any data point is the maximum total linearity error. Hysteresis/backlash is found by providing input to the instrument through the range in one direction and thenproviding input through the same range in the reverse direction. The maximum hysteresis/backlasherror is the maximum distance between any two points, one in each direction, corresponding to thesame input (Figure 27.2). Again, at least three trials should be used. To assess repeatability, input isprovided to the instrument through the range but always in the same direction. Maximum repeatability

FIGURE 27.3 Definition of sensitivity, linearity, hysteresis, and backlash.

Backlash

Decreasinginput

Increasinginput Hysteresis

Input

Linearity error

Out

put



error is the maximum distance between any two points corresponding to the same input but on twodifferent trials (Figure 27.3).

The X4 term relates to error in the system introduced by external forces. Two major sources of theseerrors are drift and noise. Drift is the condition where the instrument reading changes even though theresponse being measured does not. Drift is often caused by temperature. Noise is a random fluctuationin the instrument output. In electrical instrumentation, noise is typically caused by external electromag-netic sources, such as power lines or electric lights.

The final term, X5, is the error associated with the instrument mounting. These errors include mis-alignment, random environmental vibration of the mounts, creep of the epoxy or glue used to hold themounting bracket, slip of mechanical mounting, etc. Most of these errors can be eliminated by carefulchoice of mounting hardware, proper installation, and shielding the instruments from wind and vibrations.

27.6.2 Types of Instrumentation

A large number of different instruments can be used for measuring structural responses. These rangefrom surveying instruments used to measure large deflections to strain gauges accurate to 1 × 10–6. Eachtype of instrument has particular uses, accuracy, and limitations. Clearly, it is not possible in the limitedspace available here to cover all of the possible types of instruments that can be used. A more compre-hensive coverage of instruments can be found in Dunnicliff (1993). Although Dunnicliff ’s work is aimedat geotechnical applications, many of the instruments listed are usable for concrete applications. Whatfollows is a description of some instruments commonly used for concrete.

27.6.2.1 Mechanical Gauges

Mechanical gauges provide an easy, reliable, and inexpensive method for measuring displacement andstrain or distortion; however, the use of these gauges is extremely labor intensive and error prone becausethe gauges must be read manually. This exposes the workers to hazards, leads to possible errors inrecording the data, and requires a large workforce or a large amount of time to read all of the gauges.Among the many types of mechanical gauges, the two most common are the dial gauge and the portablestrain/distortion gauge (Figure 27.4). Dial gauges consist of a rod attached to a dial by a gearing system.As the rod moves, the movement is registered on the dial. These gauges have various ranges and accuracies.The accuracy of the dial gauge is somewhat difficult to assess. They can be reliably read to the nearesttick mark on the analog dial, but users frequently attempt to interpolate between marks. This adds adegree of subjectivity to the data. Also, the analog gauges are easy to misread, leading to error. Somemore modern dial gauges replace the mechanical dial with a digital display, thus reducing these errors.Finally, even the best jeweled movement in the gearing system will exhibit some stick (backlash), makingsmall measurements difficult. Although a dial gauge may respond to dynamic motions, it is impossibleto read accurately under these conditions. It is therefore usable only under static or low-frequency cyclictesting. The main advantages of these instruments is that they are easy to use, easy to read, and requireno auxiliary equipment (data acquisition systems). They are also inexpensive, especially if the gauges arepurchased through a machinery supply house rather then from a scientific supplier.

The portable distortion/strain gauge consists of a metal bar, two points, and a dial gauge (Figure 27.4).One point is fixed; the other point is attached to the dial gauge and is movable. Two targets, pieces ofmetal with drilled holes, are attached to or embedded in the concrete at the basic length of the gauge. Azero reading is obtained by placing the points of the gauge into the holes in the targets and reading thedial gauge. Subsequent measurements are taken in the same way and the differential movement is foundby subtracting the zero reading. This instrument is particularly useful for measuring over large gaugelengths. For smaller gauge lengths, Vernier calipers or micrometers can be used. The portable distortion/strain gauge is also useful in cases where a permanently mounted gauge is impractical. This method isoften used to measure strains when conducting long-term creep tests (ASTM, 2002). Bodocsi et al. (1993)used this type of gauge to measure seasonal movements of pavement joints. The portable distortion/strain gauge has the same problems as the dial gauge, but also has the additional problem that the accuracy



is dependent on how well the points sit in the holes in the targets. In general, there is always some playin the interface so the measurements are often not repeatable. For this gauge, it is best to take severalmeasurements and average them. Because the gauge must be manually placed to take measurements, itis only usable for static or quasi-static tests.

27.6.2.2 Electrical Instrumentation

Electrical instrumentation has greatly improved the ability of the engineer to safely and accuratelymonitor structural response. These instruments have the advantage that they can be remotely monitoredsuch that the workers never have to enter the structure during the test. This also allows engineers whocannot be present during the test to monitor the test results in real time from distant locations. Electricalinstruments also have the advantage of directly sending all data to computer files, thus reducing datatranscription time and errors. Many data acquisition programs allow the instrument responses or math-ematical permutations of these responses to be plotted in real time so load–deflection, load–strain, orstress–strain graphs can be plotted while the test is being conducted. One major disadvantage of electricalinstrumentation is that these instruments tend to cost much more than similar mechanical instruments.They also require expensive auxiliary equipment, such as data acquisition systems and cables. The biggestdrawback, however, is the need for electrical power, which is not always available at remote sites. Wherepower is not available, it is necessary to use a generator. Generators tend to emit a large amount ofelectrical noise, which affects instrument accuracy, and they are prone to power surges that can damagesensitive equipment. As with mechanical instruments, it is not possible to cover all the possible instru-ments here, but a few common instruments will be discussed.

27.6.2.2.1 Linear Variable Differential TransformersLinear variable differential transformers (LVDTs) (Figure 27.5) are used to measure displacements.Typical LVDTs will have ranges from 2 to 150 mm (0.1 to 6 in.). The instrument consists of a primarycoil, two secondary coils, and a magnetic core. An AC voltage excites the primary coil, which in turninduces voltage in the secondary coils. The amount of voltage in each of the secondary coils depends onthe position of the magnetic core. The secondary coils are connected series opposing so only the differ-ential voltage is output. This differential voltage is linear with respect to core position. When using longlead wires, DC is preferred. Many LVDTs contain circuitry so a DC voltage can be input and a DC voltageis output. This is done by converting the input DC voltage to AC and output AC voltage back to DC.Such LVDTs are frequently referred to as DCLVDTs or DCDTs. The main error in an LVDT is linearityerror, especially near the ends of the range or in the middle when the LVDT passes the null point wherethe differential voltage changes sign (Hrinko et al., 1994). Dunnicliff (1993) reported that these instru-ments have no hysteresis, but hysteresis, backlash, and repeatability errors have been found in carefulcalibration (Hrinko et al., 1994). The LVDT, theoretically, has infinite resolution, but the actual resolutionis limited by cable noise and the resolution of the data acquisition system. The instrument responds fastenough to use in any type of testing, including dynamic testing; however, LVDTs have some drift whichmakes them less than ideal for long-term monitoring.

FIGURE 27.4 Portable strain gauge.

Dial gaugeMovable

slide

Fixed point

Target attachedto structure



27.6.2.2.2 Wire PotentiometersWire potentiometers (Figure 27.6) are also used to measure displacement but have much bigger rangesthan the LVDT. Typical ranges for a wire potentiometer are 125 to 900 mm (5 to 30 in.). The potentiometeris a contact that moves along a resistor. The resistance in the circuit changes as the contact moves, andthis causes a change in voltage in the circuit. The voltage change is directly proportional to the movementof the contact. In the wire potentiometer, a wire is wound around the spindle of a spring-loaded rotationalpotentiometer. As the wire is pulled, the potentiometer moves and registers a voltage change proportionalto the wire movement. Spring loading keeps the wire taut at all times. As with the LVDT, the wirepotentiometer itself has infinite resolution, with the actual resolution limited by the data acquisitionsystem and electrical noise. Wire potentiometers have higher error rates than do LVDTs, but they alsohave larger ranges, so the error, as a percentage of the range, is about the same as for an LVDT (Hrinkoet al., 1994). The main errors in a wire potentiometer are backlash and hysteresis (Hrinko, et al., 1994),which introduce noise in the loading if the wire potentiometer is used as the control instrument in aclosed-loop test. The backlash and hysteresis errors also make the instrument unusable for dynamictesting and of marginal use for high-frequency cyclic tests.

27.6.2.2.3 Electrical Resistance Strain GaugesThe electrical resistance consists of several loops of wire bonded to the structure such that the loops areparallel to the direction of strain. Strain in the structure causes strain in the wires which is directlyproportional to change in the wire resistance. The change in resistance is read by a wheatstone bridgecircuit, which is well known and can be found in most standard physics or electric circuit texts. Anexcellent discussion of the bridge circuit as it relates to strain gauges can be found in Dunnicliff (1993).Resistance gauges come in several types:

• Bonded foil gauges consist of a grid of thin metal foil bonded to a plastic backing which in turnis bonded to the structure. These gauges are not particularly useful for concrete. Surface prepa-ration is critical, but it is often difficult to obtain a smooth, clean surface on concrete, especiallyin the field. The foil gauges are not very robust, so the gauges must be well protected. Even then,though, the gauges will sometimes fail under extreme field conditions or when embedded.

FIGURE 27.5 Linear variable differential transformer (DC type).



• Weldable gauges have the wire loops attached to a metal backing which can then be welded to thestructure. In concrete structures, these gauges are useful only for attachment to rebar or to metalembedments. As for foil gauges, protection of the gauge is of primary importance.

• Embedment gauges are specifically meant to be embedded in concrete. The wire loops are encasedin a sheathing that protects the gauge during the casting process.

Levi (unpublished data) found that resistance strain gauges were prone to drift, making them unsuit-able for long-term monitoring. He also found that, although modern resistance strain gauges are madeto compensate for temperature effects, the strain readings could be influenced by temperature effectsin the rest of the system—specifically, the lead wires and bridge completion circuit. Thus, resistancestrain gauges are recommended only for short-term or dynamic monitoring, and it is recommendedthat baseline zero readings be taken before the start of each test. The electrical resistance gauge presentsanother problem when used with electronic data acquisition, as the gauge requires a completion circuit.It is possible to complete the circuit externally (using three more unstrained strain gauges or threeresistors) and then attach the entire circuit to a channel in the data acquisition system, which reads thevoltage; however, making and using the external completion bridge can often be difficult and timeconsuming. To avoid this complication, many data acquisition system manufacturers make specialboards with the strain gauge completion bridge already wired in. This allows the strain gauge to bedirectly attached to the data system. Unfortunately, these boards often have eight, or multiples of eight,channels so some multiple of eight channels must be dedicated to only strain gauges. With the totalnumber of channels in a given system limited, dedicating all these channels to strain gauges may bewasteful if only a few gauges are used. This again illustrates the need for careful planning of theinstrumentation system.

27.6.2.2.4 Clip Gauge (Omega Strain Gauge)The clip gauge (or omega strain gauge) consists of a metal arch attached to two feet such that the resultingstructure resembles the capital Greek letter omega (Ω) (Figure 27.7). The arch is instrumented with fourresistance strain gauges wired in a full bridge. The feet of the gauge are attached to the structure, usuallyby gluing metal targets to the structure and then attaching the gauge to the target with bolts. As thetargets move with the structure, strain that is directly proportional to the movement of the targets (andthe strain in the structure) is induced in the arch. This strain is read by the strain gauges. Use of the archdesensitizes the strain gauges, and the clip gauge can read much larger strains than the strain gaugesalone; however, the clip gauge is less accurate than the strain gauge. Clip gauges are easy to install andread and are most useful for measuring fairly large strains (>100 microstrain) over short periods of timeunder laboratory or otherwise controlled environments. The gauges do exhibit considerable drift andnoise and are not suitable for long-term field use.

FIGURE 27.6 Wire potentiometer.



27.6.2.2.5 Vibrating Wire Strain GaugesThe vibrating wire strain gauge (VWG) is an excellent means of measuring strain in concrete under staticor quasi-static conditions. The gauges are accurate to 1 microstrain and are not subject to drift, as arebonded gauges. This means that it is possible to come back to installed VWG years after installation andstill obtain accurate strain values; thus, these gauges are well suited to long-term monitoring. VWGsconsist of a steel wire, threaded through a tube and attached to two rigid end blocks. These blocks sealthe tube but are free to move. The end blocks are either embedded into the concrete or are attached tothe concrete surface. When installed, the wire is under an initial tension. An electromagnet is placedaround the outside of the tube. On command from the data system, the electromagnet pulses, sweepingvarious frequencies. One of these frequencies will be the natural frequency of the wire. After sweepingthe frequencies, the magnet remains on, and the vibrating wire cuts the magnetic field, creating electricalpulses with the same frequency as the wire vibration. The data system measures these pulses anddetermines the frequency of vibration. From the frequency, the tension of the wire can be found. Thisin turn can be used to determine the elongation and the strain in the wire. Because the end blocks areattached to or embedded in the concrete, the change in strain in the wire is the change in strain in theconcrete. Although calculation of the strain from the frequency is quite simple, it is also not necessaryas most data systems will do the calculations and provide direct strain readings if the gauge calibrationfactor (supplied by the manufacturer) is input to the data system software.

The main drawback to the VWG is that the process of reading the frequency can take a few seconds,so the gauges are not useful for dynamic situations. The module that excites and reads the wire vibrationis expensive, so it is impractical to have a vibrating wire excitation module for each gauge. Normally,multiple gauges are read using a multiplexing unit that allows a single-channel data system to readmultiple channels by reading one channel at a time. This allows a large number of VWGs to be readusing a single excitation unit. Reading the gauges one at a time does take some time, given the slowresponse of individual gauges; for example, a 16-channel multiplexing unit reading all 16 gauges maytake 15 to 20 seconds. Because the gauges are only used in static or quasi-static tests, this is not a limitation.VWGs also require temperature correction if the gauge is attached to or embedded in something with acoefficient of thermal expansion different from steel. The problem here is that under a temperaturechange, the wire may expand or contract more than the material to which it is attached, giving a falsestrain reading. This temperature can be easily corrected with a simple calculation if the calibrationtemperature (provided by the manufacturer) and the actual temperature are known. Most VWGs havebuilt-in temperature sensors that measure the gauge temperature so the correction can be made. Thismakes VWGs particularly useful when they are embedded in concrete, as the temperature sensors canmeasure the heat of hydration and internal curing temperatures. After initial curing, these sensors canmeasure temperature profiles through the concrete, allowing for calculation of temperature effects in thestructure.

FIGURE 27.7 Omega or clip gauge.



In cases where the gauge is used on a steel structure, the wire expands or contracts at approximatelythe same rate as the structure; thus, the gauge is self-correcting for temperature strains. If the gauge isattached to concrete, the output should be corrected for the difference in the coefficient of thermalexpansion between the two materials. Because the coefficient of thermal expansion of concrete variesand is usually not known, the accuracy of any temperature correction is questionable. The coefficient ofthermal expansion for concrete is very close to that of steel, so any error in uncorrected data is likely tobe small. The vibrating wire gauges are extremely stable and rugged, making them ideal for long-termfield measurements.

27.6.2.3 Optical Fiber Strain Gauges

Fiberoptic sensors represent a different method of measuring strain in concrete structures. Unlike elec-trical sensors, strain is measured by changes in light waves transmitted through the fibers. One suchsensor is the Fabry–Perot (Measures, 1995), which consists of two mirrors, separated by a distance (gaugelength) set perpendicular to the axis of the fiber. The mirrors can be internal to the fiber or at the endsof two fibers, separated by a gap and joined in a coupler. This device works by sending a coherent (single-wavelength) light beam down the fiber. Some of the light is reflected off of the first mirror, which is onlypartially reflective. This forms a reference wave. The remaining light passes through the partially reflectivemirror, reflects off the second mirror, and passes back through the partially reflective mirror. This secondwave is now out of phase with the reference wave because it has traveled a longer distance. When thetwo waves are combined, an interference pattern results. As the fiber is strained, the distance betweenthe two mirrors changes. This changes the path length of the second wave and therefore the interferencepattern. Because the interference fringe pattern is dependent on the wavelength of the light (constant)and the path length, strain can be measured by determining the change in the interference pattern andconverting this to a change in path length. A second type of sensor uses a high birefringent optical fiber(Ansari et al., 1996). Here, a circularly polarized light wave is divided into two modes along the principalaxes of the fiber. External deformation causes interference of the two modes, and the output has asinusoidal variation, the period of which is a fringe. The change in length is related to the number offringes formed. A third type of sensor uses a Bragg grating (Chen and Nawy, 1994; Nawy and Chen,1996; Measures, 1995). The Bragg grating is imprinted on the fiber lead, making a longitudinal, periodicvariation in the refractive index of the fiber core. As light passes through the grating, a portion of thelight is reflected, but the reflected light has a spectral shift in wavelength. As the grating is strained, thereflected wavelength changes, and this change in wavelength can be directly related to strain (Figure27.8). This sensor has the unique capability of pinpointing the strain at precise locations in the structurerather than detecting average strain in a stressed zone. It was originally developed and used in concretestructural research and proved to be effective for online remote-control sensing of the deformation andcracking of structural elements at their critical locations of high stress, using sensors mounted eitherinternally on the reinforcement or externally on the concrete surface.

27.7 Case Studies in Performance Evaluation of Concrete Structures

In surveying the recent literature on performance evaluation, it was found that the great majority ofstudies concern bridges. This is not to imply that buildings are not evaluated, only that such evaluationsare rarely reported. One reason for this may be that bridges are subjected to more severe loading andenvironmental conditions and are often required by law to have periodic inspections. As a result, per-formance evaluation of bridge structures may be more frequent. The more probable reason for the lackof published material on building structures, however, may be that buildings tend to be privately heldstructures whose owners may, for legal reasons, be hesitant to publish information on building deficien-cies. Although the material presented here is heavily weighted toward bridge structures, the informationis applicable to almost any type of concrete structure.



27.7.1 Testing of a Three-Span Slab Bridge

Extensive performance evaluations of a deteriorated, three-span concrete bridge were conducted by Aktanand his team at the University of Cincinnati (Aktan et al., 1992, 1993; Miller et al., 1994). The bridge,shown in Figure 27.9, had extensive deterioration along the shoulders and sides of the slab. The deterio-ration was so severe that the top mat rebar on the shoulders was completely exposed. The testing of thisbridge had three main objectives: (1) to determine how severe deterioration affected the elastic responseof the bridge, (2) to determine the effect of deterioration on the load capacity and failure mode of thebridge, and (3) to evaluate the effectiveness of nondestructive testing techniques in bridge evaluation.

The first step in the evaluation of this structure was to determine the cause and extent of the damage.Attempts were made to remove core samples from the shoulder area, but the deterioration was so severethat cores could not be removed intact and came out inlayers. Some of these samples were used formaterial evaluation. Petrographic examination showed that the aggregate had extensively D-cracked. Thistype of cracking occurs when large pieces of semiporous aggregate are saturated and then permitted tofreeze. Some of the aggregate was also reactive, and signs of alkali–silica reaction were found. It is believedthat the aggregate cracked and the expansion of the aggregate cracked the surrounding cement paste.This provided a path for water to enter the slab, which then drove the alkali–silica reaction. There werealso signs of carbonation, which lowers the pH of the cement paste and creates an environment wherecorrosion can occur. Chloride testing revealed that the cracked shoulder areas had high chloride contents

FIGURE 27.8 Fiberoptic Bragg grating (FOBG) sensor system. (From Nawy, E.G. and Chen, B., Fiberoptic Sensingthe Behavior of Prestressed Prism-Reinforced Continuous Composite Concrete Beams for Bridge Deck Applications,Transportation Research Board, Washington, D.C., 1997.)

Fiberoptic Bragggrating strain sensors

Light source(LED)

3 dB coupler

Optical spectrumanalyzer

(a)

Spectrum

120P

40P

BA

(b)

1322.0 1332.0SWPWIIKRLIIKR

1327.0 nmSAURnmnm/ RESYS

1323.24 nm20P/



from road salt. The chloride, along with the air and water entering the slab through the cracks, causedthe steel to corrode, further cracking and damaging the shoulder areas.

It is believed that the deterioration was limited to the shoulder areas because chloride-laden ice andsnow would have been pushed to the shoulders by traffic and plows. Here, the snow melted and eitherran over the slab edge (no other drainage was provided) or soaked into the porous asphalt layer and wastrapped in between the asphalt and the slab. This would have saturated the shoulder area and providedan environment for the deterioration to occur. It is also of interest to note that chloride intrusion andmilder signs of corrosion were found over the piers. It is probable that the chlorides entered throughflexural cracking. Due to the short span of the bridge (the end span was only 30 ft long), it was foundthat only one axle would actually be at a critical point in the span as a truck passed (Figure 27.10). Thisrear tandem axle was duplicated by the use of two loading blocks fitted with hydraulic cylinders. Theloading blocks were placed in one lane, as this provided a realistic loading pattern and the researchersfelt this would also reveal any effect of nonsymmetrical loading on the slab. Estimates of bridge capacityand performance were obtained by running several finite-element analyses (Shahrooz et al., 1994a).

FIGURE 27.9 Three-span concrete slab bridge.

FIGURE 27.10 Load placement to simulate an HS20-44.

30

Pump

105.74' Bridge Limits

40'-0''32'-0''

36'-6

''

32'-0''

87'87'Roadway

20' R

oadw

ay

CL

Abutment

Abutment

AbutmentCL of pier

CL of bridge

Edge of bridgeEdge

of roadway

CL of roadway

CL of bridge

Edge of bridge

CL of roadway

C of pier

Plan View of Simulated LoadingPlan View of Loading System

12.8'12.8' 19.8'

Loading block

Loading block

Rock anchorsActuators

Rock anchorsRock anchors

Edgeof roadway

12'

8-strand rock-anchor cables

Pier

WaterRockRock

FillPile

Elevation View of Simulated LoadingElevation View of Loading Block

Anchor head

Hydraulic actuator

Top of slab

7'' Diameter core1'-3''

L


27-24C

oncrete C

onstru

ction E

ngin

eering H

and

book

FIGURE 27.11 Instrumentation grid for the slab bridge.

CL of pier CL of pier

40'32' 32'

A B C D E F G H I J K L M N O10

98

7

65

43

21

6' 6' 5' 5'4 @ 5' each 3 @ 10' each 4 @ 8' each

4 @ 4.2' each

2 @ 6.2' each

4.3'

4.3'4.4'

Vertical wire potentiometer location (qty. 38)Vertical DCDT location (qty. 18)Horizontal DCDT location (qty. 6)

Load blocks

N

Total number of instruments shown = 62



After the finite-element analyses were completed, an extensive instrumentation was instituted (Figure27.11) for 38 grid points. Two types of instruments were available for measuring deflection: 2-in.-rangeDCDTs and 10-in.-range wire potentiometers. The points in the loaded span were instrumented with the10-in.-range wire potentiometers because preliminary finite-element analysis indicated that deflections inexcess of 2 in. would be expected. Strains in the reinforcing bars were measured at 11 points on the bottomsteel (clustered around the loading blocks) and at 8 points in the top steel (4 over the piers near the loadingblocks and 4 at midspan near the loading blocks). Bonded resistance gauges were used in these applications.Concrete strains were measured using DCDTs that were attached between two angles glued 6 in. apart onthe slab. After the truck load tests, which measured deflections, the bridge was tested to failure.

Using rock anchors (Figure 27.10), 7-in.-diameter holes were cored at four points in the deck. A 6-in.auger was then passed through the core hole and was used to drill a hole 45 ft into a rock layer beneaththe bridge. A cable made of several seven-wire prestressing strands was then grouted into the rock. Twoconcrete loading blocks were then cast on the bridge deck over the core holes. Four 350-kip-capacityhydraulic cylinders were then placed into holes cast into the loading block. The shafts of cylinders werehollow and the rock anchor cables were passed through the shafts and tied off into button heads.

The loading cylinders were controlled by an electronic controller. This controller was linked to one ofthe wire potentiometers monitoring deflection below the loading blocks. The bridge was initially loadedto 64 kips (16 kips/cylinder) and then unloaded. Additional load and unload cycles were made to 112kips and 124 kips. In each case, the bridge remained linear, although some slight hysteresis was observedon unloading. The bridge was reloaded after several days to 700 kips and then unloaded. It showed somenonlinear behavior, and flexural cracking was observed. The bridge failed at a load of 720 kips when itwas reloaded the next day. Failure occurred as a circular arc around the loading blocks, resembling apunching-shear type of failure (Figure 27.12).

Subsequent analysis of the data revealed several interesting conclusions for the destructive test (Milleret al., 1994). Initially, the bridge carried the load in a direction parallel to the traffic lanes. When arotational restraint at the abutment was overcome, the bridge began to carry load perpendicular to theskew. At about 700 kips of total load, the longitudinal reinforcing began to yield, and the load path againshifted to parallel to the traffic lanes. This shifted load back to the damaged corner (as evidenced by asudden increase in top rebar strain at this location). The load was apparently too much for the damagedshoulder, and it failed in shear with the failure propagating throughout the entire slab. It is also of interestto note that the top steel in the shoulder areas was completely exposed and corroded; however, it wasstill possible to yield these bars before the failure pulled them out.

FIGURE 27.12 Final failure of the bridge.

CL of pier

A B C D E F G10

9

8

7

6

5

4

3

21

Loading blocks

Edge of roadway Appearance of failure on top surface of bridge

App

performance evaluation of structures · structural concrete (aci committee 318, 2005). this method,...

Documents