FRACTURE CONTROL OF ENGINEERING STRUCTURES ~ ECF 6 MECHANISM BASED STATISTICAL REQUIREMENTS FOR FRACTURE TOUGHNESS TESTING K. Wallin” and K. Térrénen” A mechanism based statistical brittle fracture model is applied to determine the reliability of "brittle" fracture tough- hess testing. Guidelines for optimizing the number of tests needed for sufficient reliability in the results are evaluated on a theoretical basis. It is shown that brittle fracture testing, which usually is easier to perform than ductile fracture testing, requires more tests to ensure the same degree of reliability, urRopUeTION ‘The evaluation of the critical flaw size in the assess— ment of the structural integrity of e.g. offshore and Pressure boundary components is currently based on fracture mechanics. The largest hazard in causing unstable crack growth is that of brittle cleavage fracture. Fracture mechanics testing is, however, often expensive and time consuming to perform. Therefore it is of importance to be able to optimize the number and Size of test specimens needed to ensure reliability of the design parameters extracted from the results, such as lower bound curves etc. Usually it is thought that 2 to 5 full thickness teste per heat is sufficient, regardless of fracture mechanism. Much of the scatter in test results has commonly been attributed to some errors performed in the testing procedure. “Metals Laboratory, Technical Research Centre of FinlandFRACTURE CONTROL OF ENGINEERING STRUCTURES ~ ECF 6 the scatter due to the material inhomogeneity and fracture mechanism has not been examined thoroughly. Rccatly it has been shown through mechanism based Recetetical evaluation that, for brittle fracture, the qeobtare mechanism based scatter is actually quite iarge (Wallin (1))- In this paper results from the statistical evalu- ation are studied. The aim is to be able to tell how ASny and what type of tests should be performed, in many oe tne large scatter, to ensure a certain degree view viability of the Lower bound values or equivalent Ceatture toughness parameters. The required number of fracture Gifverent for ductile and brittle fracture mechanism. qt is shown that brittle fracture testing, which usuatly ia eagier to perform than ductile fracture RELAY, requires more tests to ensure the same degree of reliability. PARAMETER VALIDITY cme main feature of most fracture mechanics parameters The mare able to describe materials fracture toughness ig to pe Single parameter. Such interrelated parameters wee or example the plane-strain fracture toughness ire, ‘the critical J-intergral Jy¢ and the critical SiScx' tip opening displacement CTOD,- All the different parameters are assumed to repre- sent the stress and strain fields ahead of a sharp sont, theecnuse of their definitions they should all be seSSiry independent and show identical scatter for 2 Jeomeent stress state, An important factor affecting constant ability of the different parameters is the way Brownich they are determined. Depending on the method wy a the parameter chosen one will observe quite angeorent effects regarding specimen size and crack (bhgih dependence as well as the test result scatter. Sond often used toughness parameters and their Some cterietics as suggested by the authors are listed in Table 1. tn Table 1, Kyo and Kg are the valid and the invaq lia fracture toughdGss, deGermined by ASTM E399, rosa. pigeively: Both parameters are determined by construct pect ’Satpercent secant to the P-V-curve- Ky and Ky are ABS 2 ained by using the Kzq LEFU-formulas together with seco rag maximum and load St ductile fracture initi— 564FRACTURE CONTROL OF ENGINEERING STRUCTURES ~ ECF 6 TABLE 1 - Commonly used fracture toughness parameters. Gyre] Paramerer|scarren| size errect [crack | RELIA- uenora | srurry EFFECT Lert] Kyo medium |saal1 emal1 | medium % snalt [unpredictable | medium | seit K, snait | unpredictable | medium | smaii KK smatt [unpredictable | large | smelt epru| Jy0 [medium [emait emall | medion a mediun |smait smait | large crop, |meaium | ema1t small | large 5e targe medium anal | large crop, jiarge |meaiun emai | large om snail | arge/ targe | enatt unpredictable analt |tarse/ unpredictable | large | salt mediun/|medium nediun/ | medium large large nediun/|meaiun mediun/ | medium large large ee ation, respectively. The parameter Jyc is the valid ae tele reeeree teal value’ devormined Sy AST! E813. The Sa eee vane and CTODy are the values of Param ere grhl and the Erack tip Spening displacenent se tear ae elaeions ductile and brittle respec SST Fem aba, lane, the respective parameter ELELY: Sp enonaing to minimum loed. Finally Jy and eeeeee ahhe Bor the value of the J-integral and crack CTO eta for aacement at the onset of brittle tip Gpening n*hae been preceeded by sone anount of ductile eearing A summary of thickness effects connected to dif~ ferent parameters is presented in Fig. 1. Most of the Gbserved size effects are due to invalid tests. The 568FRACTURE CONTROL OF ENGINEERING STRUCTURES ~ ECF 6 only test parameters to be regarded as valid are Kyo SY ctebe, ay and CTOD;. ‘The other parameters do 2oF srrespond td the initiation of crack extension. Tn~ ceeeg they describe rather the thickness effects on seetinen plasticity together with ductile crack growth. Specie they cannot be regarded to represent parameters Me eutbing true initiation fracture toughness. Inter= seeing enough, even the valid parameters seen to record thickness effects. the effect of normalized crack length on different paraneters is schenatically presented in Fig: 2. it is paramthat the “valid” parameters show the smallest erack length dependence. ordinary Kygrteste according to ASTH £399 applies Lupn-formulas afd allows the use of the 95 percent Lert {°hrocedure. Since the 95 percent secant usually secant, Proteinly plasticity effects, it's application descr ives Teld an optimum description of the materials does not Yeture toughness. A more Tealistic toughness actual, fractained by calculating the J-integral or CTOD w2 Uke Sceual crack initiation point, ductile or BE ERT 2Sinds will lead to physically more realistic toughness values. ne preferable paranctors to use are thus Jo, cron Pehee ana’ sitn sone reservations Jy 28 CPOBer Zk. Spoitcacion of these parameters yielder coe ease eee oe brittle fracture, a relatively noel 18, Seeag mamvees [il sil § cetiac obra sae tee eres arctfece (watiin (2))+ Move. effects Specimen Mee oes be dealt with theoretically can, on the oemeyon of the Wotenodel presented By see Pa ta) (CLEAVAGE FRACTURE hrittle cleavage fracture differs mechanism vise Bohpletely from ductile fracture. Cleavage fracture is cone eked’ by a critical stress induced statistical iathanism, governed by the fracture of brittle Beccipicates like carbides (Curry et al. (4), Curry (5) ere fbeenfield et al. (6))- As such, cleavage fracture ane Spe affected besides by changes in the stress win Pe tions also by the probability of finding a weak Gatticre (3). The specimen size will affect the Patghness partly by changing the constraint at the toughness End partly by changing the effective volume Sra he number of weak particles at the crack tip. 566FRACTURE CONTROL OF ENGINEERING STRUCTURES ~ ECF 6 ‘The new theoretical WST-model based on carbide induced brittle fracture (3) and Wallin et al. (7) allows one to write the probability of cleavage fracture in the case of a sharp crack as (2) crake aa P= 1 - exp - (cL “min )*, a) Kop = Kmin Se wes: Gi) By be Se Sieber pecanlaien. Fo the stresd” intensify factor, Rejeisra lower Limiting Picture toughness dnd kee ie S!Ehickness and constant temperature one can write (2) Kop, = Knin + (Kony ~ Knin) * ga", @ where 8, and B, are the respective thicknesses. Writing ean. (1) as Ky = Kpin + (Kop > Kmin) * {1n 1/4 @) I-Pe one can construct a go called failure probability dia~ gram yielding a very clear description of the materials fracture toughness scatter. An example of one such diagram is presented in Fig. 3 for Welding Institute fracture toughness round robin data. Combining eqn. (1) and eqn. (2) the thickness correction for a single toughness value becomes = Kin + (Kp, - 4 Xp, 7 Kain + (Kay ~ Knin? “ By applying eqn. (4) it is possible to compare fracture toughness results obtained from specimens of different thickness. It hag been proposed that constraint effects are not very pronounced when dealing with brittle cleavage fracture (2). Actually it has been demonstrated (2) that eqn. (4) is guite sufficient to describe the size effect as long as the K calculated from J-integral fulfills the reguirement B > 25 + J/of1oy- Resides that it is possible to examine the experi- mental scatter with the method described above it can ser