(Microsoft PowerPoint - expos\351-restitutionSENSO.ppt)
Transcription
(Microsoft PowerPoint - expos\351-restitutionSENSO.ppt)
Quality of NDT measurements and accuracy of concrete physical properties Denys BREYSSE, Mathilde LARGET, Zoubir Mehdi SBARTAI, Jean-François LATASTE (Univ. Bordeaux) Jean-Paul BALAYSSAC (Univ. Toulouse) NDT – CE 2009, Nantes, 30th june-3rd july 2009 1. Challenge of non destructive techniques and difficulties Provide an estimate of material indicators enabling to reassess the structure, to check safety or reliability Indicators of interest are : strength, modulus, porosity, water content, chloride content, carbonation depth... Estimates means : average values with level of confidence, c.o.v., characteristic values (reliability analysis need c.o.v., but they often lack real reliable data) NDT – CE 2009, Nantes, 30th june-3rd july 2009 To analyse measurements means: To estimate the values of required indicators from the measurements (observables) accounting for - all potential influent factors (bias) - effects of variability and uncertainty NDT – CE 2009, Nantes, 30th june-3rd july 2009 2. Assessing variability at various scales What has been done in the laboratory AND on site At a local scale Point repeatability Local repeatability (within specimen) Variance V1 Variance V2 NDT – CE 2009, Nantes, 30th june-3rd july 2009 V1 Point repeatability evaluation V2 V2i Local repeatability measurement (within sample, inside « homogeneous area » V2j V2k V3 NDT – CE 2009, Nantes, 30th june-3rd july 2009 Global variability (between specimens, between areas) Spatial variability of properties (indicators X / observables Y) Ym(x) = Yref + Yglob(x) + Yloc + εY Var (Ym(x)) = Var (Yref) + Var(Y)glob(x) + Var(Yloc) + Var (εY) V1 =0 V2 V3 NDT – CE 2009, Nantes, 30th june-3rd july 2009 3. Calibration of models between indicators and observables Data and results come from a huge experimental program (ANR-SENSO) implying : - five research teams, - laboratory and on-site programs, - many different techniques (about 50 observables) Strategy was defined such as to : - build an extensive database (material indicators / NDT observables) - quantify variability at various scales - calibrate Y = fM (X) models for all indicators - study quality of measurements, enabling selection of « best observables » - analyse complementarity of observables - develop data fusion for helping assessment NDT – CE 2009, Nantes, 30th june-3rd july 2009 quantify variability at various scales Obs 36/54(Bordeaux) 6 cv V1 cv V2 5 cv V3 4 3 2 1 0 LABO Saint-Nazaire Bordeaux on site Selection of « best » observables - accuracy : low V1 - sensitivity : a priori, the highest sensitivity, the better… but can be different Suppose Y1 = a + b Sr + c Esat Y2 = a’ + b’ Sr + c’ Esat when combination is considered, what imports is maximizing [ b c’ – b’ c ] NDT – CE 2009, Nantes, 30th june-3rd july 2009 4. Assessment of variability and of its consequences Models have been calibrated – let us assume they are EXACT (no model error) Y1 = 1128 + 4.87 Sr + 26.4 10-3 Esat Y2 = 2644 + 8.77 Sr + 39.1 10-3 Esat Y3 = 1.94 – 0.015 Sr + 0.053 10-3 Esat Y4 = 0.541 – 0.0016 Sr + 0.0014 10-3 Esat Y5 = 0.956 + 0.00379 Sr - 0.0032 10-3 Esat Yi = ai + bi Sr + ci Esat If one assumes Sr is known (controlled) Esat = fMOD-1 (Yi) quality of estimation depends on n (test number) and on variance of measurement (V2 or V3) minimal number of tests for a given confidence and accuracy n (kα/2 / ci ∆E(Esat))² . V NDT – CE 2009, Nantes, 30th june-3rd july 2009 Determination of Esat acoustic waves & & resistivity & radar & #$ ' # "# ∆%$ ∆ $$ ! But… do not forget model error ! (think to « calibration ») Bilinear regressions have been chosen ai, bi and ci have some statistical uncertainty : f.i., for Y2 = 2644 + 8.77 Sr + 39.1 10-3 Esat (+/-290) (+/-1.45) (+/- 10 10-3) and r² = 0.65 NDT – CE 2009, Nantes, 30th june-3rd july 2009 & Assessing condition combining several observables 4,5 6000 4 5000 4000 3000 3,5 4-4,5 3 3,5-4 2,5 3-3,5 2 2,5-3 1,5 2-2,5 2000 1000 0-1000 Esat 20 0 40 60 100 Sr Esat Y2 = velocity of compression waves 0-0,5 40 20 0 Sr 10000 1000-2000 10000 80 16667 2000-3000 23333,33333 Y3 = log(electrical resistivity) porosité (%) 20 15 Case of porosity/saturation 10 5 Re 2 Ra 1 US 3c US 6 Ra 7c Re 7 saturation (%) 0 0 10 20 30 NDT – CE 2009, Nantes, 30th june-3rd july 2009 40 50 60 70 80 1-1,5 0,5-1 60 23333 3000-4000 0 43333 4000-5000 80 0 30000 5000-6000 36666,66667 1,5-2 100 0,5 36667 50000 1 90 100 Assessing condition – problems with variability porosité (%) 20 15 10 5 Re 2 Ra 1 US 3c US 6 Ra 7c Re 7 saturation (%) 0 0 10 20 30 40 50 60 70 80 90 100 Specimen A Specimen B porosité (%) Specimen C porosité (%) 20 20 15 15 10 10 5 Re 2 Ra 1 US 3c US 6 Ra 7c Re 7 5 saturation (%) 0 0 10 20 30 40 50 60 70 80 90 100 Re 2 Ra 1 US 3c US 6 Ra 7c Re 7 saturation (%) 0 0 NDT – CE 2009, Nantes, 30th june-3rd july 2009 10 20 30 40 50 60 70 80 90 100 Et maintenant, un dépouillement « en direct » Les données (salle du 3ème étage) – moyenne sur les 3 dalles - Vitesse US transmission : 4746 m/s +/- 20 m/s - (log) Résistivité quadripole 5 cm (Ω.m) : 1,77 +/- 0,005 - (log) Résistivité quadripole 10 cm (Ω.m) : 1,81 +/- 0,005 - (log) Résistivité Wenner (Ω.m) : 1,90 +/- 0,01 - capacité (GE) : - 483,4 +/- 8 - amplitude radar : 0,38 +/- 0,01 NDT – CE 2009, Nantes, 30th june-3rd july 2009 Avertissement ! Les résultats qui suivent font l’objet d’une « manipulation » NDT – CE 2009, Nantes, 30th june-3rd july 2009 Sr - porosité 30 p Esat 25 20 15 10 5 Re 2 Ra 1 Re 1 Re 7 Ca 1 US 6 Sr 0 20 40 60 80 100 Parfaite cohérence des mesures entre elles Parfaite complémentarité Bonne capacité prédictive… mais NDT – CE 2009, Nantes, 30th june-3rd july 2009 Résultat obtenu avec l’exploitation immédiate des relations de corrélations et des mesures 30 30 p Esat 25 25 20 20 15 15 10 5 Re 2 Ra 1 Re 1 Re 7 Ca 1 US 6 p Esat 10 5 Sr 0 Re 2 Ra 1 Re 1 Re 7 Ca 1 US 6 Sr 0 20 70 120 Résultat immédiat absurde 20 40 60 80 nécessité d’une calibration NDT – CE 2009, Nantes, 30th june-3rd july 2009 100 Avec la même calibration 40000 Esat (Mpa) 35000 30000 25000 Re 2 Ra 1 Re 1 Re 7 Ca 1 US 6 20000 15000 Estimation Sr-Rc 10000 5000 Sr 90 0 20 40 60 80 Estimation Sr-module d’Young Rcsat (MPa) 100 100 80 70 Re 2 Ra 1 60 Re 1 Re 7 Ca 1 US 6 50 40 30 20 10 Sr 0 20 40 60 80 Besoin d’essais semi-destructifs NDT – CE 2009, Nantes, 30th june-3rd july 2009 100 Conclusions ANR-SENSO très vaste Base de Données Méthodologie pour analyser les conséquences de l’hétérogénéité à différentes échelles Mesure de différentes variances (ponctuelle, locale, globale) des indicateurs vers la variabilité Le CND permet, après calibration, d’estimer les propriétés du matériau et sa variabilité, et de quantifier la précision de l’estimation. L’estimation de la résistance mécanique requiert de recourir à quelques mesures semi-destructives (rebond, arrachement…) Ces techniques ouvrent de larges portes pour les calculs de fiabilité des structures Le point principal demeure la question de la calibration des modèles. Elle a été conduite au laboratoire (et son utilité vient d’être illustrée) et sur site. NDT – CE 2009, Nantes, 30th june-3rd july 2009 30 p 25 20 15 10 5 US 3a US 3c US 1' IE 1 d Ra 1 Re 1 Re 2 Ca 1 US 6 Re 7 0 20 30 40 50 60 70 80 NDT – CE 2009, Nantes, 30th june-3rd july 2009 90 Sr 100 30 p US 3a US 1' IE 1 d Re 2 Re 7 25 20 US 8 US 11 Ra 1 Ca 1 US 3c US 1 Re 1 US 6 15 10 5 Sr 0 20 40 60 80 100 120 NDT – CE 2009, Nantes, 30th june-3rd july 2009 140