1. Silvano Marchetti University of L'Aquila, Italy silvano@marchetti-dmt.it Some specific comments on DMT/SDMT Frontespizio 1

2. 2 Two primary “DMT Manuals”. Both are in the proceedings book (and USB key) Contains 95% of what is needed to know on DMT TC16 DMT Report 2001 1 A supplement / Upgrade. Clarifications and new developments Some 2015 Updates (SM paper to this Conf.) 2 Then for SDMT : operator manual & software manual (down loadable) Standards : Eurocode 7 (2005), ASTM (2007) ISO (Tech. Spec. 22476 being converted to Standard (info Powell, Arroyo…)

3. (S)DMT can be pushed by DRILL-RIG Two distinct methods – not confuse 3 pushro d   Drill rod Sciacca method or “Penetrometer mode” Cable exits laterally in air Test starts from ground surface Torpedo method or “Downhole mode” Cable exits laterally in the hole Test starts from bottomhole quickest

4. 4 Central component clean.  Electrical contact

5. Disc & small components : Clean ! In general not necessary to clean Necessary if membrane or tube broken : mud inside  Absence of B reading

6. IS THE BLADE IN SHORT CIRCUIT ? (THE SOUND NEVER STOPS) Use the continuity tester as in the picture. Insert plug behind. There has to be sound when you touch the metal washer, no sound in any other part of the blade Possible reasons could be : A piece of metal is floating somewhere below the sensing disk (remedy: clean and blow compressed air) The inner longitudinal cable inside the blade is broken and is in contact with the metal of the blade

7. Yellow flag - on back of low press gage : OPEN OPEN ! only in low pressure gages 6-8 bar If you have observed in the past a variable zero offset, such inconvenience should disappear. Avoids the “barometer effect” i.e. avoids that the zero offset depends on the ambient atmospheric pressure

8. Two diagrams obtained by instrumented blades : expansion Po  P1 approx linear Campanella et al. (1985) Proc 11 ICSMFE, S. Francisco, 2, p. 852. Fretti et al. 1992 Rig 26,4, 237-243 Hence central movement S=1.1 could be any value (two points). But : If S too small : (P1-P0) could be too small to measure it accurately If too high : Membrane “inertia” would rapidly increase (would work in tension rather in bending) 8

9. INFLATION SPEED. Time to A and to B Start inflation immediately after blade stops (say  2 sec) Keep the “time 0 to A”  15 sec (10-20 sec), same A to B. If A in >20 sec: appreciable drainage may occur, esp. in silts Seasoned operators “fast-slow” metod: inflate fast up  75% of expected reading. Then inflate slowly to the reading. Takes half the time. Recommended in silts, to possibly test in undrained conditions. 9 ADVANCING SPEED NO NEED 2 cm/s (twice or half ok)

10. 10 SOIL NEAR SURFACE MAY APPLY PRESSURE NOT ENOUGH FOR CLOSING THE MEMBRANE If soil applies  h <  A, the membrane will not close No usual sound upon penetration.  Apply depression with the syringe and simply write a negative A value (algebra will take care correctly). Then B. NOTE. While doing Seismic DMT, the membrane must be closed (contact). Hence near surface, if there is no sound, depression must be kept with the syringe, to insure contact. This lack of contact happens frequently near surface if operator has already done a DMT expansion.

11. “Complaint” : Cu field vane > Cu dmt (in very plastic clay) From Book “Soil Mechanics in Eng. Practice” (Terzaghi, Peck, Mesri) : Cu field vane needs a correction factor before it can be used in stability analysys. The Bjerrum correction is eg 0.70 when PI = 70. Cu field vane reduced by Bjerrum’s correction is often considered the best available Cu for stability analysis. The DMT 1980 correlations for Cu were developed using for calibration such “operative” Bjerrum’s-corrected Cu values. It is therefore Cu field vane uncorrected which is too high - in plastic clays. 11

12. 12 Here just few tips for semi liquid soils  paper. Offset of pressure gage Take  A,  B before &after each sounding, several times Start inflation immediately after reaching the test depth (in practice within a couple of sec), particularly in silts. Inflate Zero to A in  15 sec and A to B in  15 sec If semiliquid soil is silt, try to speed, to possibly remain in undrained conditions. Seasoned operators inflate "rapidly" until  70% of expected reading (A or B), then slow to accurately read the pressure (“brief inflation”  8 sec). See also Sect 15.4 : Semiliquid and submerged (e.g. bottom slurry of mining wastes hydraulically transported and discharged in a pond).

13. 13 Drainage conditions. “Niche silts” Ideally A & B should be taken simultaneously A B  In niche silts B is too low (hence too low B-A and B-A derived Id,Ed,M). Niche silts recognizable : Id  0 (low Ed,M) Contact Stress  h A log time Niche silt Undrained If we stop at a given z and take repeated A-readings… Clay  no  h decay In niche silts : B-A, Id,Ed,M. But a reasonable A may still be obtained (brief inflation 8 sec), and the parameters dependent only on A, notably Cu. Can still do undrained analysis E.g. in a tailings dam of niche silt : can do undrained analysis More details in the paper substantial  h decay Sand : drained Clay : undrained Silts : partial

14. 14 Displaying the DMT results Recommended graphical output TC16. Diagrams should be presented not separated, but side by side, + expressive Displaying just p o and p 1 is unexpressive The struggle is between factuality (p o,p 1 ) and expressivity (figure) The "conventional" formulae are "average" correlations in textbook soils. In non textbook soils an alternative could be to plot, superimposed, results by site specific and conventional correlations (also evidencing deviations vs "average" sites). Recommending the format in Fig. is not to suggest one set over another set of correlations, but preserve simultaneous vision of various profiles.

15. 15 The indication is : reflects essentially D r (only to minor extent stress history) reflects D r + various stress history effects such as OCR, aging, Ko, structure (cementation) Q cn KDKD

16. 16 Elusivity of Dr in situ There is no unique mapping Dr=f(Qc). Jefferies – Been ISOPT 1995 : “Hilton Mines sand at Dr=60% produces the same Qc as Monterey sands at Dr=40%” Very difficult estimate Dr in situ  unmeasurable Specs in terms of Dr problematic Schmertmann, 1986 in a paper on monitoring compaction : Since objective of compaction: limit settlements…..more rational specs on Modulus (e.g. M dmt ) than Dr

17. 1 DISPLACEMENT The fixed 1.10 mm is  H between two solids 1% (of quantity, NOT of full scale !). DMT : load-displacement test. Two ingredients. 2 PRESSURE Pressure gage (transducer) AT SURFACE : Zero offset continously checkable. Or reset – eliminated. [in tools with sensor at depth Zero offset real problem – uncontrollable -!] Can use even a v high sensitivity gage (low scale) – semiliquid soil. Operator cannot regulate, spoil, tamper. No skill. … 1 % Despite being AT DEPTH ! Repeatibility : inherent in the construction of the device  0.01 mm Question by Mike Long DMT repeatible as it is made in just one place or other…?

18. May use M = constant if  ' v large ? M = Eoed=1/mv=  ' v /  v (at  ' vo ) Except highly structured clays (sharp break), M variation across pc is moderate Error in assuming M ~ constant : often acceptable (other methods for M : not infrequent error factors 2-3)   18

19. Using the 1-D formula in 3-D problems e.g. small isolated footings Moreover : Engineer has to choose (?)  x e  y not well predicted by elasticity (?) E  0.8-0.9 M DMT (elasticity) M=M DMT Boussinesq solutions (simple, engineer independent) denominator is less sign minus Difference 1-D vs 3-D : 10-15% Poulos Istanbul : What is important is the modulus, not the formula !

20. Obviously Id is not a sieve analysis. Eg. a mixture sand- clay would probably be "wrongly" interpreted as silt. On the other hand such mixture could perhaps behave mechanically as a silt. The engineer is often interested to the grain size distribution not "per se", but just to infer mechanical properties, PERHAPS, in SOME cases, it could be better to have the Id interpretation than the sieve analysis results and to infer from them the mechanical behaviour. A “mechanical” information (a sort of Soil Type Behaviour Index) that, in design, might be even more important than the granulometric composition. Reliability of material index Id

21. Clay. Initial idea (1980) : investigate E d -E u. Problem: E u, lab too dispersed, impossible set up correlations. Hence  E d -M. Link E d -M presumably weaker, but at least can be tested. NOTE. E d -M must be a complex function of many variables, among them the Skempton p. p. parameters A & B and anisotropy (E d horizontal, M vertical), which in turn depend on soil type (~represented by I d ) and on OCR (~ represented by K d )  some basis to expect at least some degree of correlation E d -M using I d and K d as parameters. Final word : real world observation. Large number case histories favourable comparisons measured vs DMT-predicted settlements - or M dmt -M reference. Lambe et al (Jnl ASCE March 1977 “Foundation Performance of Tower of Pisa” p.246) : “Drained moduli of saturated clays are typically about one-third to one- fourth the undrained values”. Hence a broad connection drained-undrained stiffness already invoked in the past. How derive M (drained) from E d (undrained) ??? 21

22. SETTLEMENT CALCULATION M DMT + 1-D method Calculation every 20 cm, not because thicker layers are inadequate. It is just more convenient, since Mdmt available in the computer every 20 cm 22

23. 23

24. Pouso Alegre / MG Piso de galpão de logística para 5,0 tf/m 2 Ensaio DMT em aterro compactado com solo local 24

25. 25 Ladd 1971 (44 yrs ago) Short course MIT Soft Ground Construction Even good OC samples may give oedometers moduli 2 to 5 times too low (Terzaghi & Peck book 1967)

26. Ed must be corrected to obtain M M=Rm Ed with Rm=f(Kd,Id) Don’t use Ed as Young’s Modulus Rm has various correction tasks Distortion Horiz to vertical Drained Undrained Once Ed is converted to M Young’s E’  0.8-0.9 M (elasticity) 26

27. OCR and Ko in sand OCR in sands can be  evaluated from ratio M DMT /q c. M DMT /q c = 5-10 in NC sands M DMT /q c = 12-24 in OC sands. or from the continuous Monaco’s curve Ko in sand. See TC16 DMT Report A way can be (once estimated  OCR) : K o =K onc OCR m 27

28. 28 Z (m) KPa Cu at N ATIONAL S ITE FUCINO Nc= 14 to 22 ? A.G.I., 10 th ECSMFE Firenze 1991,1 : 37 0 5 10 15 20 Z (m) KP a Cu at N ATIONAL S ITE BOTHKENNAR Nash et al., Géotechnique, June 1995: 73

29. INPUT PARAMETERS FOR PLAXIS (Hard Soil model “HS”) Recommended basic input E 50 ref (Trx modulus confined at 1 bar ). Schanz found for quartz sand E 50 ref =15 to 75 MPa (loose to very dense sand). He also found E 50 ref correlated to E oed ref. In sand the range of M DMT by DMT is remarkably similar to above range (15-75 MPa). Hence as a 1st approximation may assume E 50 ref  M DMT ------------------------------- Predicting settlement by DMT may  replace a costly load test. Often load tests executed to adjust assumed FEM parameters. 29

30. Research embankment D=40 m(Venezia) Comparison M DMT vs M back-calculated from LOCAL vertical strains measured under center Sliding Micrometers installed every meter 30 SETTLEMENTS Measured 36 cm DMT predicted 29 cm (  = 24%)

31. Silos on Danube's Bank (Belgrado) 31 SETTLEMENTS Measured 63 cm DMTpredicted 77 cm (+22%) (D. Berisavijevic 2013)

32. Exponent n used for obtaining the normalized parameters Kd and Qcn (used for predicting CRR) Q cn = [(q c -  v )/ p a ] (p a /  ' v ) n n = 1 a welcome simplification – avoids the iterative procedure to determine Q cn and n, an additional soil unknown (n=0.5-1) Due to arching : q c (z) increases less than linearly Blade no arching : side ratio  6 32 Uniform NC sand Kd = (p 0 -u 0 )/  ' v

33. 33 Determining “n” (0.5 to 1) not straightforward Flow chart - Iterations by computer Robertson & Wride – Canad.G.J. 1998

34. (  Giacheti, Mayne, Chang, Failmezger, Fonseca…) In general The correlations involving pwp (e.g. Cu=…) are generally not applicable. The correlations for estimating moduli/ settlements generally applicable, since DMT is a load test determining a pressure-settlement relation. DMT CORRELATIONS IN NON-TEXTBOOK SOILS (Residual, Unsaturated…) 34

35. Liquefaction (spontaneous, earthq. induced etc.) : complex, involves many soil properties 35  Many unknown. Qc alone is not enough. Need multi- parameter triangulation CRR = f (Qcn, Kd, Ed, (Id), Go/M DMT ). East Germany 1000s hectares higly unstable (fence : keep out). Often high sudden settlements - spontaneous collapse - “butterfly triggered” Dangerous to compact them, rigs may sink. Vibrations >> than earthquake. C is stable for any Earthquake. But - if there is cementation (or SH) - can A be stable ? Can we rely on cementation (SH)? Also : Qc  Dr Kd  (Dr+SH) : need both to separate If Fragile : Cement destroyed by strong Earthquake – and more or less also by direct push tests? Not reflected by a  Kd. But reflected by a high Vs (high Go/M). If Ductile : Cement not destroyed, reflected by a  Kd

36. 36 CALCAREOUS SAND (Plouasne – France – 1991) Id  sand - Kd very low

37. 37 WATERDEPTH 0 -100 m PUSH CAPACITY 7 ton SEAFLOOR DILATOMETER Ballast 7 ton buid locally Ballast connected by bolts to the pushing system (60 Kg) 6-7 pushrods already charged vertically on top – can add more Limitation : Cannot push CPT (no 1 m stroke)

38. 38 END Thank you

39. MODULI CAN ONLY BE POSITIVE - INHERENTLY What to do if you have several Ed<0? (Tell operator to be more careful when taking DA,DB) Shoud not happen ! However : Verify if you have just one Ed<0 or many Ed<0 If “Just one” you may accept (but cancel that depth) If “many” Ed<0 : it means DA, DB too high. Reduce them gradually (eg in proportion) until all Ed positive (maybe one negative) (“Regular” DA,DB 0.15 0.45 bar)

40. REPRODUCIBILITY of DMT Cestari (SGI), Lacasse (NGI), Lunne (NGI), Marchetti (Aq) 40 NC clay Onsoy, Norway

41. Nuno’s thesis Giacheti H.L. et al. (2006) “Flat Dilatometer Testing in Brazilian Tropical Soils”, Proc. 2nd Int. Conf. on Flat Dilatometer, Washington D.C.: 110 D.A. Brown, J. Vinson, 1998. Comparison of strength and stiffness parameters for a Piedmont residual soil. Proceedings of the First International Conference on Site Characterization - ISC’98, Atlanta, Georgia, USA, Vol. 2, pp 1229-1234. G.K. Martin, P.W. Mayne, 1998. Seismic flat dilatometer tests in Piedmont residual Soils. Proceedings of the First International Conference on Site Characterization - ISC’98, Atlanta, Georgia, USA, Vol. 2, pp 837-843. Chang, M.F. 1988. In Situ Testing of Residual Soils in Singapore. Proc. 2nd Int. Conf. on Geomechanics in Tropical Soils, Singapore, Vol. 2: 97-108. Dec. Wang Chainchye E. and Borden Roy H."Deformation Characteristics of Piedmont Residual Soils" Jnl ASCE GE, Vol. 122, No. 10, October 1996, pp. 822-830 Failmezger, R.A., Rom, D. & Ziegler, S.B. 1999. SPT? A better approach to site characterization of residual soils using other in-situ tests. Behavioral characteristics of residual soils, ASCE Geotech. Spec. Publ. No. 92, 158-175. Mayne, P.W. 2005. Unexpected but foreseeable mat settlements of Piedmont residuum. Int. Jnl of Geoengineering Case Histories, http://casehistories.geoengineer.org, Vol. 1, Issue 1, 5-17. A. Viana da Fonseca, M.M. Fernandes, A.S. Cardoso, 1988. Characterization of a saprolitic soil from Porto granite by in situ testing. Proceedings of the First International Conference on Site Characterization - ISC’98, Atlanta, Georgia, USA, Vol. 2, pp 1381-1387. SDMT in NON-TEXTBOOK SOILS : SOME EXPERIENCE General comment : estimating by DMT modulus / settlement should be appli- cable whatever the material (glass, lead, residual, artificial fills…), because DMT is a load test determining directly a pressure-settlement relation.

42. 42 Significant – systematic research in CC by Koreans (sand) “Effects of stress history on CPT and DMT (Lee 2011, Eng. Geology) Qc: 1.10 - 1.15 larger Kd: 1.30 - 2.50 larger  Kd ++ reactive to SH than Qcn

43. PRESTRAINING CYCLES simulate AGING (grain slippage) CC TEST N. 216 IN TICINO SAND They applied prestraining cycles in calibration chamber. Found : K D (DMT) 3 to 7 times more sensitive to AGING than penetration resistance K D + 20 % q D + 3 % Similarly higher sensitivity of Kd to SH & aging observed by Jamiolkowski (ISC'98 Atlanta) 43

44. Reconstituted sand specimens at same Dr “Testing on sand specimens reconstituted even at exactly the same density is highly questionable” (Ladd 1977). Already known at the time (SOA 1977 Tokyo) Hoeg (2000) showed that… “natural sands had dilative behaviour while reconstituted specimens with the same void ratios were contractive” Hence the need of testing sand in situ 44

45. Vp - Seismic Dilatometer (SPDMT) s wave p wave DMT Vp (m/s) Depth (m) lower sensor upper sensor Pianola (AQ) 2013 Two vertical axis sensors provide two independent Vp profiles (pseudo interval) 45

46. 46 Recent results indicate : the CPT-clean-sand curve not unique but comprised in a wide band band - f(K d ) E.g. CRR = 0.12 (SH low), but 0.24 (SH high). Factor 2 ! The CPT “consensus” curve is generally conservative. BUT can be v. uneconomical in prestressed/ aged sands where Qcn predicts liquef just because Qcn does’not feel SH. 0.12-0.24 are both right ! explains historical controversies by researchers Kd may help to choose the line in the band But then may be even + direct to obtain CRR from f(Kd)

47. 47 CONCLUDING REMARK 1 The most important of the evidenced aspects is probably the capability of DMT to sense Stress History, not easily felt by other tools. SH is necessary for reasonable predictions of CRR and of settlements. Migrating from the SPT-blowcount N to Q cn for predicting CRR and settlements has had the merit of eliminating that part of scatter due to the poor repeatability of the SPT. The use of K D should reduce even that part of scatter due to the scarce ability of conical tips to distinguish between freshly deposited sands and prestressed or aged sands.

48. 48 CONCLUDING REMARK 2 DMT is approx 1½ costlier than CPT. Hence, if one considers only the cost, CPT is preferable. However CPT may be sufficient only if your geotechnical report does not include a section “settlements prediction”. If settlements are important to the designer, CPT cannot help, and DMT should be executed, because of the well known weakness of CPT – unable to provide realistic settlement predictions (eg Powell & countless others). Frequently adopted is a combination of both, since they complement and since the penetrometer can push both.

49. STORY of the correlation OCR-K D in clay Theoretical 1993 Finno Experimental 1980 & 1995 Theoretical 2004 Yu It is often said that DMT interpretation is empirical... 49

50. DERIVATION OF Cu CORRELATION Once having OCR  Ladd’s SHANSEP 77 SOA Tokyo (Ladd : best Cu not from TRX UU but from oed  OCR  Shansep) Using m  0.8 (Ladd 1977) and (Cu/  ’ v ) NC  0.22 (Mesri 1975) OCR, Ko, Cu well founded derivation. They use OCR confirmed by theory + SHANSEP  50

51. Situation of CRR repeats for modulus : different curves f(SH), but SH missing. How to select curve? OCR??? RATIO  = E/Qc= 2.5 to 25 f(OCR?) Jamiolkowski concludes : "without Stress History it is impossible to select reliable E (or M) from Qc“ (Isopt-1, '88, Vol. 1, p.263) May use Kd to select curve, or use directly M dmt. 51