Influence of mix proportions on rheology of cement grouts co(6)

742L.Svermovaetal./Cement&ConcreteComposites25(2003)737–749

Table4

Parameterestimatesofsevenstatisticalmodels

Mini-slump(mm)R2¼0:99EstimateProb:>jtj

a1a2a3a4a5a6a7a8

TransformInterceptW/BSPLSPVAW/BSPW/BLSPW/BVA

none111.50

21.370.0031.250.0012.750.00)17.630.0011.880.00––––

Platecohe-sionmeter

(mm)R2¼0:97EstimateProb:>jtjnaturallog)0.77)0.370.01)0.930.00)0.320.020.560.00––0.270.04––

Yieldvalue

(Pa)R2¼0:97EstimateProb:>jtjnone6.98

)0.250.37)1.840.00)1.060.010.280.31)1.260.012.680.00––

Plasticvis-cosity(Pas)R2¼0:97Estimate

Prob:>jtjnaturallog)1.18)0.560.00)0.280.01)0.280.010.420.00––––)0.160.06

Induced

bleeding(Lmin)R2¼0:99EstimateProb:>jtjnone1.08

0.0240.08)0.0480.010.140.000.0380.02)0.10.000.10.000.170.00

fc07days

(MPa)R2¼0:91EstimateProb:>jtjnone27.28)4.500.00––)6.480.00––––––––

fc028days

(MPa)R2¼0:93EstimateProb:>jtjnone32.10)4.130.00)2.230.03)6.630.00––––––––

wherea0denotestheoverallmean;coe cientsanrep-resentmodelconstants(contributionofindependent

variablesontheresponse),andeistherandomerrortermrepresentingthee ectsofuncontrolledvariables.Thethirdorderinteractionisusuallyneglected.

Forexample,mini-slump,platecohesionmeter,yieldvalue,plasticviscosity,inducedbleedingandfc0at28daysaregiveninEqs.(1)–(6),respectively.Mini-slumpðmmÞ¼111:5þ31:3SPþ21:4W=B

À17:6VAþ12:75LSPþ11:9W=BSP

lnPlatecohesionmeterðmmÞ

¼À0:77À0:93SPþ0:56VAÀ0:37W=BÀ0:32LSPþ0:27W=BLSP

YieldvalueðPaÞ¼7þ2:7W=BLSPÀ1:8SP

À1:3W=BSPÀ1:1LSPþ0:3VAÀ0:3W=B

lnPlasticviscosityðPasÞ¼À1:2À0:56W=B

þ0:42VAÀ0:28SPÀ0:28LSPÀ0:16W=BVA

InducedbleedingðLminÞ

¼1:1þ0:17W=BVAþ0:14LSPÀ0:1W=BSPþ0:1W=BLSPþ0:05SPÀ0:04VAþ0:03W=B

0fC

ð1Þ

ð2Þ

ð3Þ

ð4Þ

plasticviscosity,inducedbleedingandcompressivestrengthat7and28daysare0.99,0.97,0.97,0.97,0.99,0.91and0.93,respectively.Thehighcorrelationcoe -cientofmostresponsesdemonstratesexcellentcorrela-tionwhereitcanbeconsideredthatatleast95%ofthemeasuredvaluescanbeaccountedforwiththeproposedmodels.

Table5indicatestheaveragemeasuredresponseofthefourreplicategrouts,coe cientofvariation,estimatederrorwith95%con dencelimit,aswellasrelativeerrorforeachofthemeasuredproperties.Theestimatederrorofcementgroutformini-slump,platecohesionmeter,yieldvalue,plasticviscosity,inducedbleeding,andcompressivestrengthsat7and28dayswereÆ6.6mm,Æ0.08mm,Æ0.80Pa,Æ0.04Pas,Æ0.026Lmin,Æ3.3,andÆ2.9MPa,respectively.

Therelativeexperimentalerrorsformini-slump,yieldvalue,plasticviscosityandcompressivestrengthareshowntobelimitedto2–12%.Ontheotherhand,therelativeerrorfortheplatecohesionmeterwas22%.Thisvalueisexpectedtodecreasewiththeincreaseinrelativecohesionplatevalue,sincethemeanofcohesionplatevalueofthegroutcorrespondingtothecentralpointswasslightlylow.

The owtimemodelisnotgivenastheresultsoftheMarshconecouldnotbeusedfor nalevalua-tionbecausethegroutswithhighlevelofviscosityagent(0.07%)hadsuchhighviscositythattheywereunableto owthroughthe5-mm-outletofthefunnel.Inthiscase,anoutletbiggerthan5mmisrecom-mended.

4.2.Accuracyoftheproposedmodels

Theaccuracyofthestatisticalmodelswasdeterminedbycomparingaveragevaluesofpredicted/measuredratioformixesofgroutwhichwereproducedforveri -cationoftwo-levelfractionalfactorialdesign.Theaver-

ð5Þ

at28daysðMPaÞ¼32:1À6:6LSPÀ4:1W=B

À2:2SP

ð6Þ

Thecorrelationcoe cientsoftheproposedmodelsformini-slumptest,platecohesionmeter,yieldvalue,