Some Recent CorrosionEmbrittlement Failures

of Prestressing Systemsin the United States

Morris SchupackSchupack Suarez Engineers, Inc.South Norwalk, Connecticut

Mario G. SuarezSchupack Suarez Engineers, inc.

South Norwalk, Connecticut

FOREWORDThis paper is based on a paper pre-

sented by the authors at the FIP Sym-posium on Stress Corrosion Cracking ofPrestressing Steel in Madrid, Spain,September, 1981.

Having been asked to report on theincidence of embrittlement failures ofprestressing steels in the United Statesin the recent past, the authors havehighlighted this type of failure in thispaper. Evidently, any failure of a pre-stressing steel tendon due to corrosionof any kind is undesirable. The need toexercise extreme care to protect pre-stressing steels in any application can-not be overemphasized.

For the purposes of this paper, corro-sion, or ordinary corrosion, is an elec-trochemical phenomenon in which thesteel is affected by a particular envi-ronment, resulting in a measureableloss of metal. The failure of a givensteel element occurs when the loss ofmaterial is so great that the remainingcross section is no longer capable ofcarrying the applied load. In failure,however, the metal generally still ex-hibits its normal ductility.

On the other hand, stress corrosioncracking and hydrogen embrittlementare distinctly different manifestations ofcorrosion of prestressing steel, stressedgenerally to over 50 percent of its ulti-mate strength. Without attempting to

38

belabor the metallurgical definition ofand distinction between the twophenomena, the noticeable effect ineither case is a drastic loss of ductilityon the part of the affected steel. Fail-ures, when they do occur, are typicallybrittle, often with little or no loss ofmetal in the steel element, with possi-bly no structural warnings.

INTRODUCTIONCorrosion and, in particular, stress

corrosion cracking and hydrogen em-brittlement of prestressing steels, hasalways been a concern of engineers inthe United States and indeed through-out the world. As a contribution to theunderstanding of the problem of corro-sion of prestressing steel, and in orderto put the magnitude of the problem inproper perspective, particularly fromthe viewpoint of the structural en-gineer, the authors have attempted togather information on cases of corrosionof prestressing steel which may haveoccurred in the United States in thepast 5 years or so.

To that effect, a survey was made inan effort to reach as many individuals aspossible, throughout the country, whomight have knowledge of such cases.This paper presents the results of thesurvey, a general discussion of the in-formation received, and a description offour cases of embrittlement of prestres-sing tendons on which sufficiently de-tailed information was available. Thisreport is also an update of a previouspaper prepared by Schupack in 1978.1

SURVEY FORMAT

The limited survey consisted insending more than 250 letters to se-lected individuals throughout theUnited States who, in the opinion of theauthors, should be aware, collectively,of most cases of corrosion affecting pre-stressing steel. These individuals in-

SynopsisA limited survey was made tc

gather as much data as possible onincidents of corrosion of prestress-ing steel which may have occurredin the United States during the past5 years. From the authors experi-ence and information gathered fromclose to 100 responses to the sur-vey, 50 structures were found inwhich tendon corrosion has oc-curred.

Corrosion incidents were usuallyfound to be associated with poordetails and/or execution, generallyin the presence of an aggressiveenvironment.

Considering the fact that about200,000 tons (approx. 180,000 met-ric tons) of prestressing steel peryear are used in the United States,the number of incidents reported issmall, particularly since about 40percent of them have occurred inpost-tensioned parking structuresexposed to high concentrations ofchlorides from deicing salts.

Of the 50 corrosion incidents re-ported, 10 cases of probable brittlefailures related to stress corrosionor hydrogen embrittlement werecited. It is likely that a few more mayhave occurred.

In most cases, the exposure ofthe prestressing steel to an aggres-sive environment, particularly withsulphides present, has triggered theloss in ductility. Those incidents in-volving embrittlement, in whichmore details were available, are de-scribed herein.

Recommendations are put forthwhereby corrosion embrittlementfailures of prestressing steels canbe avoided.

PCf JOURNAL/March-April 1982 39

eluded practicing structural engineersand representatives of state highwaydepartments and other governmentagencies, prestressing steel and post-tensioning materials suppliers, con-tractors, pipe manufacturers and tankconstructors. Enclosed with the letter ofexplanation and request for informationwas a detailed questionnaire whichcould be returned with the pertinentdata.

Also, through the valuable coopera-tion of the Prestressed Concrete Insti-tute, which included it in one of itsNewsletters, the questionnaire reachedall members of the PCI, includingCompany, Associate and ProfessionalMembers.

The limitations of the survey must berecognized, Even though about 50mentions of corrosion were included inthe nearly 100 responses received, it isunlikely that every case of embrittle-ment was reported. Unfortunately, notevery key , individual contacted nor, inparticular, every prestressing materialssupplier, responded.

SURVEY FINDINGS

General

The types of prestressing steels mostprevalent in the United States havebeen: seven-wire stress-relievedstrands (ASTM A416), used exclusivelyin pretensioning; the same strands, colddrawn and stress-relieved wires (ASTMA421), and cold stretched and stress-relieved bars (ASTM A722), making uppost-tensioning tendons, both bondedand unbonded; and cold drawn wire,used to prestress cylindrical wirewound pipes and tanks.

All types of prestressing steels havebeen affected by corrosion at one timeor another. Actual incidents of corrosion.can generally be related to ill con-ceived details and/or poor execution,usually associated with an aggressiveenvironment. In a few instances of steel

failures, these could be related to man-ufacturing defects in the prestressingsteel and not to corrosion.

The majority of reported cases of cor-rosion involved unbonded tendons. Inthe early days, these tendons weremostly grease coated and paperwrapped. The grease was sometimesnot applied continuously nor uniformly.The paper sheath was also susceptibleto possible tears which could contributeto further loss of the protective grease.For many years now, plastic sheathsenclosing the greased tendons are uni-versally used. This has significantlyimproved the behavior of unbondedtendons, although improper detailing ofan adequate protection system in thearea of the anchorages and/or deficientmethods of installation have beenfound to result in some corrosion prob-lems.

No cases of corrosion of properlygrouted tendons have been reported. Inbonded construction, corrosion of ten-dons has been observed only whentendons have been found to be poorlygrouted or not grouted at all. Caseshave been reported where grout subsi-dence or sedimentation caused voids athigh points. 2 When this happens, it ispossible that a corrosive substance mayfind its way into these unfilled spaces.On the other hand, the actual behaviorof properly grouted tendons in chlorideenvironments has been surprisinglygood. This parallels the findings onpost-tensioned test beams which havebeen exposed in the tidal zone of thenortheast coast of the United, States forup to 21 years.3

Prestressed Pipes and TanksSeveral cases of corrosion of prestres-

sing steel in wire-wound pipes andtanks were reported in general terms.The principal factors causing these fail-ures were deficiencies in manufactureor in construction, resulting in insuffi-cient or damaged mortar coat over the

40

wires. The actual number of installa-tions affected is not available although,from the general information received,the percentage of corrosion problemsappears to be much lower for newerstructures as workmanship and con-struction techniques have improvedthroughout the years.

Parking StructuresThe most widely spread corrosion

problem today 4 appears to exist inpost-tensioned parking garage decks,particularly in areas of the countrywhere chloride salts are used for winterdeicing. This problem is basically thesame as that of corrosion of reinforcingsteel so prevalent in normal reinforcedconcrete parking decks and bridgedecks. On all the approximately 15post-tensioned concrete parking struc-tures known to have problems with cor-roded tendons, the concrete had achloride ion content, by weight of con-crete, as high as 0.9 percent.

Generally, the chloride content variesfrom a high percentage at the top of theslab to a lower percentage at the hot-tom. This gradation is believed to becaused by salts being deposited exter-nally on the surface as differentiatedfrom chloride placed integrally with theconcrete. Based on what has been re-ported and on the authors' own experi-ence, no stress corrosion cracking ofprestressing steel, but rather ordinarycorrosion with significant Ioss of metal,has occurred in parking decks subjectedto chloride.

Nuclear Containment VesselsA rare opportunity of observing ten-

don performance exists in the nuclearpower industry. Regulatory Guide 1.35,issued by the United States NuclearRegulatory Commission, requires ten-don inspection to take place 1, 3 and 5years after the initial integrity test, andevery 5 years thereafter, On vesselswith unbonded tendons (protected by

corrosion inhibiting grease), each in-spection is performed on 2 to 4 percentof the tendons and includes:

— Removal of protective grease andvisual examination of end anchor-age hardware and surroundingconcrete.

— Liftoff reading, detensioning andreloading tendons.

—Removal of one wire or strandfrom two or three tendons for vis-ual examination for corrosion andto provide samples for tensiletesting.

— Visual examination and chemicalanalysis of grease.

No unacceptable conditions havebeen reported on any of the approxi-mately 30 prestressed containment ves-sels built to date in the United States.There have been a few instances ofbroken wires which were traced tomaterial flaws or improperly formedbuttonheads. None was attributed tocorrosion. Potentially adverse condi-tions have been found in some tendonsduring surveillance, where water hadseeped into the conduits. However, inno case (the oldest tendon was 10years) has detrimental corrosion or em-brittlement been found.

The two existing grouted contain-ment vessels have been monitored byvarious means and nothing has beenrevealed to indicate any degradation ofthe tendons.

BRITTLE FAILURES

Of the total of approximately 50structures in which corrosion of pre-stressing steel is known to have oc-curred, 10 cases involved brittle failureof one or more tendons. These affectedonly isolated areas in each structureand have not been of a catastrophicnature.

Most of these brittle failures have oc-curred in post-tensioned unbondedconstruction in commercial structures.

PCI JOURNAUMarch-April 1982 41

Five of these incidents can be defi-nitely identified as some form of stresscorrosion. The others are poorly re-ported and probably have been mini-mally researched. No known case ofbrittle corrosion has been revealed inpretensioned construction. The few in-cidents of corrosion in pretensioninghave been generally related to inade-quate concrete cover and/or poor con-crete quality.

Four incidents of embrittlement fail-ure, where more detailed informationwas available, are described below.The metallurgical information givenwas extracted from private reports byvarious consultants who are not iden-tified to preserve the anonymity of thestructures involved.

Case No. 1—Elevated BuildingSlab over a Parking Area,Post-Tensioned in Two Directionswith Unbonded Greased and PaperWrapped 0.6-in. (15 mm) Strands

Each one of two similar structures,part of a building complex, consists of a230 x 330-ft (70 x 100 m) cast-in-placeflat plate, 9.5 in. (240 mm) thick, post-tensioned in two directions. Con-structed about 10 years ago, each slab isused as a building platform over aparking area, to support low-rise multi-ple dwellings. After the prestressedslab was constructed, wooden framedbuildings were erected above. Theslab, made of lightweight concrete, issupported on concrete columns which,in turn, are supported by concrete pilefoundations.

The 0.6-in. (15 mm) monostrand ten-dons were protected by a coating of cor-rosion inhibiting grease, inside a sheathformed by spirally wound kraft fiberpaper.

The 0.6-in. (15 min) strand had aminimum ultimate strength of 270 ksi(1860 MPa) and met all the require-ments of ASTM A416. The strands wereinitially stressed to 80 percent of ulti-

mate to overcome friction and then re-laxed and locked off at 70 percent. Withtime, the stress level in the strands,after losses, was expected to be about60 percent of ultimate. Bundles of twoto four tendons were spaced 18 to 42, in.(450 to 1060 mm) apart and were tied tosupporting metal chairs and to rein-forcing steel and other tendons in thetransverse direction.

Some tendons failed about 40 daysafter stressing and additional sporadicfailures continued to occur at reducingfrequencies. Of the approximately 1200tendons in the most affected slab, about4 percent have failed to date. The failed0.6-in. (15 mm) tendons have generallybeen replaced with '/2-in. (13 mm)strands.

Failures of a distinct brittle naturetended to occur more frequently overcolumns, in areas of high negative mo-ments and where the geometry of thetendons resulted in the sharpest cur-vatures. Samples, taken from variouspoints along the length of a failed ten-don removed from the structure, weretested in tension. Test results rangedfrom 100 percent of specified ultimatestrength and a typically normal elonga-tion at ultimate in excess of 6 percent toless than 85 percent of ultimate andunder 1 percent of elongation. Thesedifferences sometimes occurred be-tween samples removed from within 15ft (5 m) of each other.

Concrete core samples removed fromthe slab were shown to contain 180parts per million of sulphide sulphur.(Analyses of 500 to 900 parts per mil-lion were also obtained, but were sub-ject to question.) Some differences inpotential were observed between ten-dons in the slabs where failures oc-curred, Possibly related to this, about90 percent of the failures also occurredin tendons arranged in the east-west di-rection, in preference to the north-southdirection.

Samples of lightweight aggregatewere found to be quite variable. Some

42

1Fig. 1. Typical failure mode without surface attack.

aggregate particles were reddish incolor, others brown with a darker oreven black interior. The aggregate var-ied in hardness and in the size of itspores. Some samples of aggregate, withdark interiors, gave an odor of hydrogensulphide when moistened with hydlro-chloric acid solution_ The aggregategenerally had a hard glazed exterior,but in bulk it showed dusting, andthere were broken pieces of aggregatepresent.

Paper-wrapped tendons present acorrugated exterior to the concrete sur-rounding them, When tendons are ten-sioned, the movement of the strand,particularly with a small radius of cur-vature, may cause it to cut through thepaper tape and crush aggregate pro-tniding into the corrugation. This wouldexpose high strength steel directly topulverized aggregate. A concrete test

beam was made and draped tendons,stressed to 80 percent of ultimatestrength, were moved about 9 in. (230mm) under sustained load. The beamwas then sectioned and it was foundthat some of the aggregate had beencrushed.

Metallurgical examinations, by manydifferent laboratories, failed to give de-finitive answers. Some samples of bro-ken wires had enough evidence of cor-rosion to warrant labeling the failuresstress corrosion cracking. Other sam-ples did not show evidence of corrosionpits or surface attack. Fig. I shows atypical failure mode without visiblesurface attack, Many cracks, primarilyoriented in the transverse direction,formed near the point of failure, as canbe seen in Fig. 2.

An interesting feature of some of thebrittle failures was the fracture occur-

PCI JOURNAUMarch-April 1982 43

Fig. 2. Etching in acid disclosed many surface cracks.

ring at an angle of about 45 deg. to thelongitudinal axis of the wire. It hasbeen shown in the laboratory" that highstrength steel wire stressed in H 2S so-lutions forms cracks which tend to bealigned at 45 deg to the direction of thetensile stress. The preferred crystallo-graphic orientation of the steel in thewire causes this unusual crack orienta-tion. It was, therefore, assumed thathydrogen sulphide embrittlement ofthe high strength steel most likelycaused these delayed failures.

With or without a better explanationfor these tendon failures, it is evidentfrom this experience and the one citedlater (Case 2 below), that H2S must bekept away from the prestressing steel. Itis obvious that a sheath that will notpermit the sulphide penetration wouldhe a solution. However, with theexigencies of construction, it is neces-sary to avoid altogether the presence ofsulphides in concrete containing pre-stressing steel.

As a matter of interest, isolated fail-ures due to embrittlement under verysimilar conditions to those describedabove are also known to have occurredon at least one, possibly two otherstructures (information is not suffi-

ciently precise). Samples taken fromdifferent Iocations along the length ofthe same strand, after removal from thestructure, again showed widely differ-ent behavior under standard tensiletests.

It must be realized that most, if notall, unhonded monostrand tendons in-stalled in the United States today areencased in plastic tubing. Experimentsand experience have confirmed thatsuch a sheath is less likely to be dam-aged during construction than paperwrapping. Its smooth interior and ex-terior surfaces make it less susceptibleto be cut by the strand moving undertension. This does not insure, however,that a plastic sheathed monostrand ten-don is trouble free. The area near theend anchorage remains vulnerable, asits metal components remain exposedand must he effectively protectedagainst corrosion. It is not unusual thatthe plastic sheath is cut short and thestrand itself may lack adequate protec-tion.

Several cases of corrosion, generallywith heavy pitting and loss of metal,have been -reported, where end anchor-age pockets have been improperlyfilled or not at all, allowing water to gain

44

BRICK i \\

AIR SPACESHOTCRETE

A PRESTRESSING

T STEEL. p^

9 in. TANK WALL F BRICK ANCHOR' 6 p & DOVETAIL

_ D

6

Fig. 3. Horizontal section through tank wall

access and accumulate inside the plas-tic sheathing. Even cases of failuresdue to embrittlement (see Case 4below) have been known to occur. Thissuggests the possible coincidence of alocally exposed strand because of adamaged sheath, together with electri-cal potential differentials betweenthese exposed areas and the exposedend anchorages or other tendons,

Case No. 2 --- Sewage DigestersFour prestressed concrete digesters,

which are 80 ft (24 m) in diameter and32 ft (10 m) high, were built in 1950,using the wire wound system in whichthe wire is pulled through a die to in-duce the prestress. These tanks man-ifested two distinctly different corro-sion problems, ordinary corrosion andstress corrosion cracking, both resultingfrom inadequate shotcrete protectioncaused by poor details. Similar condi-tions have been observed before inother sewage tanks, including the Owl'sHead tank in Brooklyn, New York,which suffered a catastrophic failure in1961. 8 This information is presentedbecause of the embrittlement corrosionfound in 1980.

The brick anchor detail of these tanks(see Fig. 3) was the same one whichhad been used in the Owl's Head tank.Apparently to facilitate construction,the galvanized brick dovetail anchorwas wired directly onto the wrappedwire and then the shoterete was placed.This whole problem would have beenavoided if 1/4 in. (6 mm) of shotcreteseparated the dovetail from the pre-stressing wire.

These tanks were recently re-evaluated by the authors' firm. Somerandom openings were made in thebrick to examine the wire under thegalvanized dovetail. From this cursoryexamination, it appeared that somewires exhibited only slight surface cor-rosion. Since this sampling was consid-ered insufficient and the past experi-ence with this brick anchor detail waspoor, the brick facade on one tank wascompletely removed so that a detailedinspection of the wires could be made.After removal of the brick in order toinspect the wire, it was necessary toremove the galvanized steel dovetailand some minimal shotcrete that man-aged to get between the dovetail andthe prestressing wire. A close inspec-tion of the wires revealed several major

PCI JOURNAUMarch-April 1982 45

Fig. 4. Corroded wires bundled around a pipe. Shotcrete protection did not penetratebundle and was not bonded to tank wall.

vertical lines of corrosion extending thefull 16-ft (5 m) height of the dovetail. Ofthe 103 0.212-in. (5.4 mm) wires in-stalled in that height, anywhere from 20to 55 wires had failed or were so farcorroded that they could not he countedon to provide any dependable ultimatestrength. Frequently the top 6 El (2 m)of wall had no effective reinforcementremaining.

In a number of dovetail locations,where virtually no effective prestres-sing was left, vertical cracks up to 10 ft(3 m) in length had formed in the con-crete wall. In effect, the upper 6 to 10 ft(2 to 3 m) of the cylinder had failed.Fortunately, because of the earlier con-cern about the condition of the tank(some evidence of corrosion had beennoticed in 1970), it had been restrictedto a low operating level. Thus, theupper part of the tank had not beenused for some time.

The nature of the corrosion was

mixed. In places obvious brittle failuresoccurred as manifested by transversecracks with no necking. In other areas,longitudinal splitting was observed butalways associated with corrosion prod-ucts. In general, the corrosion productswere black.

A horizontal crack existed throughthe wall approximately 5 ft (1.5 m)above the bottom of the dovetail. It ispostulated that sewage gas found itsway through the crack and mixed withmoisture. Since the black corrosion oc-curred both above and below the crack,it is assumed that sulphides had accessto the wires.

The only corrosion protection avail-able where the prestressing wirecrossed under the dovetail slot was, income rare cases, a maximum of %a in. (5mm) of cement mortar. In many places,however, no protection was present.Where the mortar was dense, even asthin as Yia in. (2 mm), protection was

46

Fig. 5. Failed end of a longitudinally cracked wire with a bend test. This wire actuallyhas a longitudinal crack going through the bend and for most of its length.

offered. Since these tanks were floatingcover digesters, roof surface drainage(lid not spill over the wall. The wires,where exposed at the dovetail, were notin as aggressive an environment as theone that existed at the Owl's Head tank.This is a possible explanation for its30-year survival.

A detailed inspection was also madeof the lower portion of the tank walls,not covered with a brick facade, whereaccessible from inside the controlhouse. Areas of delarninated, dnimmysounding shoterete were identified andremoved. Here it was found that, wherethe wires were bundled around a pen-etration (see Fig. 4) and where theshotcrete was loose, extensive corrosionexisted. It was also found that somelongitudinal cracking of the wires hadoccurred. A short distance away, wherethey had been covered by good, soundshotcrete, the same wires did not ap-pear to suffer any significant amount ofcorrosion.

The corrosion of the wires was in-duced by leakage of sewage material,liberating hydrogen sulphide, whichcame through the pipe openings. It wasevident that corrosion occurred wherethe sludge material had access to thewire because it was not protected bysound shotcrete.

This illustrates the fact that, if thewire is properly protected and there isno direct access of the corrosivemedium to it, prestressing steel doesnot tend to corrode. By avoiding thebundled wires, by making sure that theconstruction procedures were such thatbond could he obtained between theshotcrete and the core wall, and byhaving the wires spaced adequately sothat each one could be properly encap-sulated in dense shotcrete, this problemcould have been prevented.

Fig. 5 shows a piece of a wire whichhad failed in the structure. The top endof the wire exemplifies a typical brittlefailure associated with a longitudinalsplit. The bend test performed on thewire indicates that no embrittlementoccurred away from the splitting fail-ure. However, by examining the wireclosely with a 20-power magnification,it was found that the wire had a con-tinuous longitudinal crack extendingpractically from one end to the other.The tape near the end of the wire ofFig. 5 indicates how far the crack wasvisible.

Similar cracked wire samples, alsotaken from next to the point of failure,when tested in tension, have exhibitedgood ultimate strength and elongation.It was also evident from these and other

PCI JOURNALJMarch-April 1982 47

wires with longitudinal cracks that cor-rosion tends to concentrate along thecrack, causing crevice type corrosion.

It is not known whether the lon-gitudinal cracks found in a number ofwires existed in the original wire, wereformed during or after their drawingthrough the die in the wire windingprocess, or propagated from a localstress corrosion incident. Microscopicstudies were not made on this wire.However, from another tank incident, itseems that incipient seams or folds willtrigger radial cracking, resulting in sub-stantial longitudinal splitting of thewire. This has been reproduced in thelaboratory, in the presence of sul-phides, but not with nitrates orchlorides.

According to several wire manufac-turers, longitudinal seams or folds inwires cannot be avoided altogether and,if not more than about 0.003 in. (0.075mm) in depth, they should not be det-rimental.

The authors' experience indicatesthat wire which suffers normal corro-sion or even chloride corrosion does notappear to exhibit this type of splitting.Cases of longitudinal cracks in wirehave been reported by Phillips.'Schupack, who has inspected more than70 prestressed tanks, has not observed,by the naked eye, this type of splittingin corroded wire in prestressed watertanks but has observed this in sewageand industrial tanks.

In summary, this digester tank, some30 years old, shows that where theshotcrete protection is adequate, thewires are in good condition. The keyconsideration for a designer is that ifthe structure will be possibly subject toan aggressive environment, the de-signer has to be certain that the conta-minants cannot reach the prestressingwires. Certainly, the detail of thedovetail located at the upper part of thetank is a classic example of an irrever-ent disrespect for the behavior of pre-stressing steel.

Fig. 6. Corroded end anchorage for 0.6-in.(15 mm) strand in a formed pocket.

Case No. 3 — Corrosion Failure ofa Monostrand Tendon Anchoragein a Parking Structure

This multistory parking garage con-sists of cast-in-place, post-tensionedconcrete slabs supported by a beam andcolumn steel frame. The first indicationof any problem with this structure wasthe very obvious protrusion of a strandabout 15 ft (5 m) beyond the edge of theconcrete slab. This occurred about 4years after construction. The totallength of the tendon was 165 ft (50 m).The anchorage holding the strand was awedge type anchorage contained in abarrel housing. The barrel housing is amachined cylinder with a conical holewhich takes the bursting force exertedby the wedges, putting the barrelhousing in permanent circumferentialtension. These anchorages have beenused by the hundreds of thousands andthere appears to be no prior incident ofa time delayed sudden failure of thebarrel.

In this particular incident, no signifi-cant corrosion was found on the pre-stressing strand after 5 years since thestructure had been built, except wherethe strand protruded about 34 in. (20

48

Fig. 7. Failed barrel of 0.6-in. (15 mm)strand anchorage showing longitudinalfailure surface.

mm) beyond the anchorages. Strandtaken from the dislocated tendon hadmore than the specified ultimatestrength and elongation. The strandtendons were of the unbonded typeprotected with a corrosion inhibitinggrease-Iike material and encapsulatedin a plastic sheath.

The slabs in which the prestressingsystem was installed were of normalweight concrete with a strength of over4000 psi (27.6 MPa). The concrete has afair amount of surface scaling indicatinga possibility of excessive water and/orover finishing. The concrete, analyzedafter years in service, had an averagechloride ion content by weight of con-crete of 0.34 percent with a low of 0.09percent and a high of 0.98 percent. Allthese values are higher than the nowgenerally recognized limitation in theUnited States," namely a chloride ioncontent of 0.014 to 0.025 percent (de-pending on the cement factor), byweight of concrete.

This structure is in the northern partof the United States where deicing isaccomplished by using chloride salts.From other experiences, it is believedthat the high chloride content is causedby saturation from surface depositedsnow and ice, laden with salt, broughtin by vehicles. This has become a majorproblem for many parking structures in

northern climates in North America.The anchorage pocket (see Fig. 6)

was packed with a mortar which may ormay not have contained calciumchloride. Analysis of the pocket fillingmaterial indicated chloride contents ofup to 0.22 percent by weight of con-crete. These pockets were filled mostlyin the winter, and it is unknownwhether any winter protection was pro-vided. Some of the anchorage pocketmortar plugs appeared to have shrinkaway and were loose. In some cases,the plug could be pulled out as a cone.Products of corrosion could be found atthe surface of a number of anchoragepockets.

The barrels suffered longitudinalcracking (see Fig. 7). The appearance ofthe failure surface is also shown. It wasreported by one metallurgical inves-tigator that the failure was due to stresscorrosion. Another investigator foundno stress corrosion, but only a decreasein barrel area with the resulting higherbursting stresses causing failure.

Based on an analysis ut the barrel, itis difficult to see bow the amount ofcorrosion shown could have reducedthe section to the point of causing afailure. The failure may have emanatedfrom a corrosion stress raiser which mayhave propagated into a longitudinalcrack. This may or may not have been astress corrosion failure.

The reader may question the reasonfor including this case in a discussion ofstress corrosion failures of prestressingsteel. It is of particular interest becausewith all the reports on prestressingsteel problems under elevated stress,practically none mentions failure of theanchorage. At the anchorage, the pre-stressing steel is subjected to thegreatest abuse and the most complexstress patterns because of the grippingeffects on the prestressing steel ele-ments.

In this particular case, the environ-ment was most conducive to the corro-sion of the prestressing steel, but the

PCI JOURNAL/March-April 1982 49

Fig. 8. Surface of wire sample adjacent to fracture showing elongated corrosion pitswhich intersect the fracture.

prestressing steel itself did not sufferany significant damage. Whether thebarrel in this case was sacrificial in thecorrosion cell is not known. If chlorideinduced stress corrosion was the causeof the failure of the barrel, this againreinforces the opinion most prevalentin the literature that prestressing steelin the presence of chlorides is generallynot subject to stress corrosion,

There is a difference of opinion inthis regard, but the experience reportedand the authors' own experience appearto indicate this. A case of particularinterest in this regard is the controlledtest of 20 beams in the tidal zone atTreat Island, Maine, in whichSchupack3 has been involved for over20 years. Post-tensioned beams havenot indicated any forms of stress corro-sion though there has been some verysmall amount of general corrosion.'

Case No. 4 — Unbonded Tendonsin Roof of Hotel Structure

Two failed tendons in the structurewere discovered when one of the

strands projected about 3 It (1 m) intoan adjacent room. The failed 0.6-in. (15mm) monostrand, greased and plasticsheathed, had been in service for about5 years. It could not be determined byvisual inspection, whether the plasticsheath was damaged prior to the tendonfailing. The failure occurred a shortdistance away from a vertical openingin the concrete slab through which thetwo tendons crossed.

It is quite possible that the plasticsheath may have been damaged at thecontact point with the sides of the slabopening or with the reinforcement of acolumn adjacent to the opening. In anycase, having a tendon passing throughan opening is not a desirable detail,since it is conducive to improper corro-sion and fire protection, as well as topotential physical damage because ofits accessibility.

A visual examination of the fracturesdisclosed that each wire in the strandhad fractured transversely in essentiallya brittle manner. Two of the samplesexhibited some slight necking adjacentto the fractures. It is likely that these

50

Fig. 9. Photo micrographic of longitudinal cross section taken through corrosion pit onsurface of sample. Typical branching and stress corrosion cracks emanating from the,corrosion pit (top) of the wire are clearly evident.

two wires were the last wires in thestrand to fail. Close examination of thestrand wire fracture surfaces disclosedcoarse, irregular, jagged profiles.

Irregularly shaped patches oflocalized corrosion were observed onthe wire surfaces both at and remotefrom the fractures, Some of these cor-roded areas contained a corrosion prod-uct on the surface while others ap-peared essentially free of any corrosionproduct, appearing as sharp edged,shallow elongated pits.

Examination of a microspecimen inthe as-polished, unetched conditiondisclosed that near the wire surface, thejagged fracture profile contained a cor-rosion product of appreciable thickness[approximately 0.006 in. (0.15 mm)],indicating that this portion of the frac-

ture was exposed to a corrosive envi-ronment for an extended time period.

Fig. 8 shows typical corrosion ob-served on the surfaces of the wire sam-ples.

The general appearance of the brittlefracture profile was characteristic ofstress corrosion cracking. This wasconfirmed by examination of the mi-crospecirnen taken through pitted areasadjacent to the fracture which clearlyexhibited branching cracks typical ofstress corrosion cracking. These cracksare shown in Fig. 9.

The general microstructure of thestrand wire as observed in transversemicrospecimens appeared to be com-prised of severely cold-worked pearlitewith possibly some bainitic transforuia-tion products present. The wire sur-

PCI JOUFiNAUMarch•April 1982 51

faces showed no evidence of decarburi-zation or intergranular oxidation.

Energy dispersive X-ray of the corro-sion product on the wire surfaces re-vealed the presence of calcium, potas-sium, sulphur, and silicon; the siliconand sulphur in Iarger amounts thanwould normally be present in the steelstrand. Additionally, chlorides werealso detected,

With the exception of the normal andexpected oxidation products of iron, theconstituents observed in the corrosionproducts strongly suggest that thestrand wire surfaces had either rubbedagainst concrete or accumulated a res-idue of concrete dust, since these ele-ments are commonly observed inaggregate, and also implies that the en-capsulating plastic conduit had beenpenetrated or perhaps fretted through atsome point along the length of thestrand. This scenario is compatible withthe possibility of damage to the plasticsheath at or near the slab opening, asindicated above.

Examination of the submitted wiresamples indicates that at least five ofthe seven strand wires failed trans-versely by stress corrosion crackingwhich cracking initiated at localizedcorrosion pits on the wire surfaces. Twoof the submitted strand wires probablycontained small stress corrosion cracks,but final failure occurred by tensileoverload.

None of the wire samples showedany evidence of manufacturing defectsin the nature of seams, laps or lamina-tions which could have contributed tothe failures. The metallurgical investi-gation also confirmed that the strandwas not corroded prior to installationnor did the grease on the wire contain acorrodent.

Had these conditions existed at thetime of the installation, failure wouldhave occurred much earlier in the ser-vice life of the strands.

Although the strand failure occurredin an area containing heavy electrical

equipment, the corrosion observed andthe mode of failure were not charac-teristic of that which results from straycurrents. Were stray current pitting theprimary cause of failure, then the frac-tures would have been essentially duc-tile in nature and not brittle as that ob-served in the failed wires.

CONCLUSION

As structural engineers, it is the au-thors' opinion that the corrosion prob-lems that have occurred are more a re-sult of improper choice of details andpoor execution than of any characteris-tics of the steels used. In all the corro-sion cases cited, any type of prestres-sing steel would eventually have suf-fered distress. It is up to the engineerand constructor to select systems andconstruction methods which can morepositively assure proper corrosion pro-tection.

RECOMMENDATIONSThe following are a few important

considerations that, if taken into ac-count, will contribute towards buildingsafe, problem-free structures.

Precast PretensionedConcrete Members

Historically, there have been aminimum number of corrosion relatedproblems with these structural mem-bers. The common use of high strength,dense, well compacted and properlycured concrete evidently provides thebest corrosion protection to the pre-stressing steel. Any significant voidsand honeycombs which are noticed,particularly in the vicinity of the pre-stressing strands should be filled in atan early date with a cement mortarusing, preferably, materials similar tothose used in the parent concrete.

52

Wire Wound PrestressedCylindrical Structures(Pipes and Tanks)

Design details and construction pro-cedures should he directed towardsobtaining a complete encapsulation ofeach individual wire in sound, dense,portland cement mortar. Wires shouldbe adequately spaced from each other,within the same layer and betweenlayers. Bundling of wires should heavoided. If pipes, manholes, or otherpenetrations are present, placing cir-cumferential bands on both sides of thepenetration should be preferred overforcing a change of direction of thewires around the penetration, with theresulting bundling.

The application of the cover coat(typically shotcrete in the case of tanks)must be done with extreme care, by ex-perienced personnel, to completely en-capsulate each wire and to obtain amonolithic, well bonded, compositestructure with the corewall. Auxiliaryanchors or inserts of any kind should heembedded in the cover coat. Metal tometal contacts with the prestressingwires should he avoided.

Post-TensioningTendons — Bonded Construction

The most critical factors in this typeof application are the quality of thegrout and of the grouting procedures.Where these have been good, thestructures have been trouble free. Thechoice of grouting materials, includingadmixtures, must be done carefully torelate to the particular application.Vertical tendons or any tendons whosegeometry produces significant differ-ences in elevation throughout itslength, must be dealt with special care,as bleeding and water separation be-come more significant. Location ofdrains and vents, particularly those at ornear the anchorage areas (grout caps,trumpets, etc.) is of the utmost impor-

tance, to prevent the formation of airpockets which may result in permanentvoids within the grout volume.

The grouting operation itself, the se-quence of opening and closing valvesand vents, the choice of locking off alltendon outlets at the end of the grout-ing operation or, as is the authors' pref-erence, of leaving the high point ventsopen, to allow for free expansion, mustbe carried out with care. Any of theseefforts, in pursuit of the highest qualitygrouted tendon which is possible, willbear fruit in producing permanent, de-pendable corrosion protection to theprestressing steel at a minimum, if any,increase in cost.

Post-TensioningTendons — UnbondedConstruction

The most predominant example ofthis type of construction is the mono-strand tendon market, which has pro-duced millions of square feet of cast-in-place post-tensioned concrete slabsand beams in all kinds of buildingstructures. Because of its popularity,this market has always been highlycompetitive. This has resulted in manycases in compromising quality for a re-duced cost.

The philosophy on the part of thepost-tensioning supplier of providingthe least costly installation which doesnot create any visible, immediateproblems during construction has, un-fortunately, prevailed in many a struc-ture. Better quality control over presentconstruction methods and slight im-provement of some of the materialsused would, at a minimal increase incost, avoid many of the problems of cor-rosion which have occurred. The fol-lowing are some of the things that canbe done:

1. Better control of grease applica-tion during manufacture to obtain com-plete and uniform coverage of thestrand.

PCI JOURNALJMarch-April 1982 53

2. Heavier wall thickness of plasticsheath to increase the resistance todamage and to crimping by tie wires.

3. Proper detailing of sheath to an-chorage connection to insure that aneffective seal exists at tendon ends. Thecommon practice of allowing the plasticsheath to end several inches away fromthe anchorage should be forbidden.

4. Proper detailing and execution ofthe filling of the anchorage pockets, toinsure an adequate sealed cover to theanchor so as to prevent the intrusion ofany corrosive substances from the out-side.

5. Control over concrete and mortarmixes to avoid any possibility ofchlorides, sulphides or other potentiallydamaging materials being present.

6. Adequate field inspection.

Suggested Scheme to Provide aHigher Level of CorrosionProtection for Bonded andUnbonded Post-TensioningTendons

In order to cope with the realities ofconstruction and the environment towhich the structure may be exposed,the authors have developed the conceptof an electrically isolated tendon (pat-ent applied for). According to this con-cept, the complete tendon (includingthe prestressing steel elements), theanchorages and the corrosive protectivemedium (whether grease or cement

grout), is completely encapsulated fromend to end within a tough plastic enclo-sure.

The sheath itself can be smoothwalled or corrugated, depending onwhether it is an unbonded or a bondedapplication. It must be adequatelysealed into the plastic cover enclosingthe whole anchorage system. In thisway, the tendon is protected, not onlyfrom the penetration of any aggressivematerials such as chlorides or H2S, butalso from stray currents and the forma-tion of long corrosion cells.

The use of this system will necessar-ily be more costly than what is commonpractice today. The authors estimatethat it may amount, for example, to anincrease of less than 1 percent of thetotal cost of a parking structure using'/a-in. (13 mm) monostrand tendons.This additional cost is insignificantwhen one considers the tremendouscost of repair and disruption of thestructure's use that result when tendonscorrode.

ACKNOWLEDGMENTThe authors would like to thank those

who so rapidly responded to theirquestionnaire on corrosion incidents.They particularly wish to thank Dr.John Hanson of Wiss Janney Elstner &Associates for supplying information ontheir investigations of parking garages.

54

.-iEFERENCES1. Schupack, M., "A Survey of the Durabil-

ity Performance of Post-Tensioning Ten-dons," ACI Journal, V. 75, No. 10, Oc-tober, 1978, pp. 501,511.

2. Schupack, M., "Grouting Aid for Control-ling the Separation of Water for CementGrout fur Grouting Vertical Tendons inNuclear Concrete Pressure Vessels," pre-sented at the International Conference onExperience in the Design, Constructionand Operation of Prestressed Concrete,Pressure Vessels and Containments forNuclear Reactors, University of York, En-gland, September, 1975.

3. Schupack, M., "Behavior of 20 Post-Ten-sioned Test Beams Subject to up to 2200Cycles of Freezing and Thawing in theTidal Zone at Treat Island, Maine," ACISP-65, Performance of Concrete inMarine Environment, American ConcreteInstitute, Detroit, Michigan, 1980, pp.133-152.

4. Peterson, Carl A., "Survey of ParkingStructure Deterioration and Distress,"Concrete International, V. 2, No. 3,March, 1980, pp. 53-61.

5. Townsend, Jr., H. E., "Hydrogen SulfideStress Corrosion Cracking of HighStrength Steel Wire," Corrosion –NACE,V. 28, 1972, pp. 39-46.

6. Schupack, M., "Prestressed ConcreteTank Performance," ACI SP-8, ConcreteConstruction in Aqueous Environments,American Concrete Institute, Detroit,Michigan, 1964, pp. 55-65.

7. Phillips, E., "Survey of Corrosion of Pre-stressing Steel in Concrete Water-Re-taining Structures," Australian Water Re-sources Council Technical Paper No. 8.

8. ACI Committee 201, "Guide to DurableConcrete," ACI Journal, V. 74. No. 12.December, 1977, pp. 573-594.

9. O'Neil, E. F., "Durability and Behaviorof Prestressed Concrete Beams; Postten-sioned Concrete Beam Investigation withLaboratory Tests from June 1961 to Sep-tember 1974," Technical Report No.6-570, Report 4, U. S. Army EngineerWaterways Experiment Station, Corps ofEngineers, Vicksburg, Mississippi (NTISAD A038 444), February, 1977.

NOTE: Discussion of this paper is invited. Please submityour discussion to PCI Headquarters by November 1, 1982.

PCE JOURNAL/March-April 1982 55