Compact crack arrest testing and analysis of EH47 shipbuilding steel Theoretical and Applied Fracture Mechanics

It is vitally important to measure the brittle crack arrest properties of shipbuilding steels to ensure that accidental damage will not result in total structural failure. Wide-plate test methods allow for direct measurement of the crack arrest toughness but this kind of testing is incredibly expensive. Therefore, there is a need for cheaper and simpler test methods which are able to measure a material ’ s brittle crack arrest toughness. In this work, Compact Crack Arrest (CCA) testing, which is standardised in ASTM E1221, has been successfully used to measure the crack arrest toughness of thick sections of EH47 shipbuilding steel. The results from this study have been compared to small-scale test methods. It was found that instrumented Charpy testing gives an overprediction of the CCA results, and nil-ductility transition temperature (NDTT) from Pellini tests gives a conservative estimate. The results presented in this study are discussed in terms of the effectiveness of the CCA test method for mea- surement of brittle crack arrest toughness and integrity assessment of large-scale structures.


Introduction
The issues associated with brittle fracture of shipbuilding steels were first brought to light by the premature failure of some of the Liberty ships during WWII [1]. In modern days, the requirements for shipbuilding steels are set by the International Association of Classification Societies (IACS), which includes assurance organisations worldwide. There is a drive to reduce carbon emissions of the shipping industry by using larger ships which can carry more cargo per journey [2]. These ships require stronger and thicker plates of steel for their hulls, and this carries with it an increased risk of brittle fracture, particularly in the case of bad weather or accidental damage [3,4]. A brittle crack can be prevented from causing catastrophic failure of the structure by ensuring that the materials used have a sufficient resistance to a propagating fracture i.e. high brittle crack arrest toughness [5].
The first standard which was developed for measurement of brittle crack arrest toughness was ASTM E1221:1988 [6], which utilises relatively large-scale Compact Crack Arrest (CCA) test specimens. In recent years, the most common method to measure brittle crack arrest toughness is using wide-plate testing [7][8][9][10][11][12][13][14][15][16][17] such as ESSO tests or double-tension tests which have been incorporated into International Standard ISO 20064:2019 [18]. It has been shown in previous studies that CCA testing gives a lower-bound approximation to the brittle crack arrest toughness, K Ia , whereas wide-plate testing usually enables the crack arrest toughness, K ca , to be directly measured [19]. Due to the experimental difficulties involved in CCA testing, there has been a limited amount of published results available in the public domain using this test method. The restricted amount of CCA test data which are publicly available in the literature are entirely limited to specimens with thicknesses of below 50 mm [20][21][22][23][24]. Some researchers noted that it was difficult to obtain valid results from this test method and there was high scatter in the test data, which is why wide-plate methods are generally preferred [14,22,25].
Although wide-plate test methods are more likely to give a better prediction of brittle crack arrest toughness, there are many advantages to using CCA testing. Primarily, CCA testing is much cheaper than wideplate testing. CCA testing requires less extensive instrumentation and set-up of the test; for example it is usually done isothermally rather than under a temperature gradient. Additionally, due to the smaller size and use of wedge loading, there is a lower force requirement for the test machine. For these reasons, it is of interest to carry out CCA tests on modern shipbuilding steels to determine their viability, with the aim of reducing the cost of brittle crack arrest testing.
In this paper, CCA testing was carried out on 80 mm thick EH47 shipbuilding steel to measure the brittle crack arrest toughness over a range of temperatures. The results from this study provide a unique set of data on relatively large thickness plates using the CCA test method. The obtained results have been compared to predictions of the crack arrest toughness from small-scale testing which are reported in previous research conducted by the authors [26]. These predictions rely on determination of reference temperatures and are based on the master curve approach. The specimen preparation, test procedure, experimental challenges, and the analysis of the test data have been comprehensively explained and are discussed in the following sections.

Compact crack arrest test methodology
The experimental studies in this work were carried out on EH47 shipbuilding steel, supplied by Lloyd's Register, which is widely used in industrial applications. EH47 steel was chosen due to the fact that it has been well characterised by other researchers which enables the CCA results obtained from this study to be compared to wide-plate test results available from other researchers' works [16,27]. The mechanical and fracture behaviour of the supplied EH47 steel employed in this study was characterised using small-scale mechanical testing by the authors in a previous publication [26]. The main material properties including the upper shelf Charpy energy, room temperature yield strength, σ Ys , and ultimate tensile strength, σ UTS , the nil-ductility transition temperature (NDTT), and average grain size are summarised in Table 1.
Compact crack arrest testing was carried out in accordance with ASTM E1221 [6] on 80 mm thick EH47 shipbuilding steel plates. In total, eight specimens were prepared and tested in this study, denoted CCA1-8. The CCA specimens consisted of full-thickness pieces of material which had a notched slot machined into them. The notch was opened using a wedge load through the hole in the specimen to initiate a brittle crack within a brittle weld bead deposited at the base of the notch.
The predictions of crack arrest toughness using small-scale tests, which were made in the previous paper [26], have been compared against the results from CCA test data obtained from the present study.

Specimen design
The mechanical properties of the material, shown in Table 1, were used to choose appropriate specimen dimensions which would meet the strict criteria given in the ASTM E1221 standard [6] and give the best chance of calculating K Ia . The key specimen dimensions are summarised in Table 2 and schematically shown in Fig. 1. As seen in Table 2, all eight specimens had the thickness of B = 80 mm, width of W = 260 mm, height of H = 156 mm, wedge loading hole diameter of D = 50.8 mm, length from loading hole to the crack mouth opening displacement (CMOD) measurement location of L = 65 mm, slot width of N = 8.28 mm, and initial crack length of a 0 = 55 mm. Also included in Table 2 are the normalised dimensions for each parameter with respect to the width of the specimen, X/W. According to the guidelines provided in ASTM E1221 standard [6], the thickness, B, of CCA test specimens must be sufficiently large to satisfy plane strain conditions, with the width of the test specimen, W, within the range of 2B ≤ W ≤ 8B; and the height, H, is 0.6 W.
In order to make the best use of the available material, it was decided to use a B/W ratio of 0.31, which gives a width of W = 260 mm for the plates with the thickness of B = 80 mm. The initial normalised crack length, a 0 /W, was kept very low to increase the stress intensity around the crack tip and facilitate brittle crack initiation. It was also decided to use an a 0 /W of just over 0.2 which is at the lower end of the allowed range recommended by the ASTM E1221 standard [6]. As seen in Fig. 1  the ASTM E1221 standard [6]. Six out of eight specimens were sidegrooved on both sides to promote a straight crack front. The remaining two specimens were side-grooved only on one face so that crack propagation monitoring wires could be soldered to the other face of the specimen to measure the crack speed as it propagated. As shown in Table 2, the side-groove depth of S = 10 mm was implemented on each side of the test specimens with the net thickness between side grooves of B N = 60 mm in those six specimens with double side-grooves, and B N = 70 mm in those with a single side-groove. As shown in Fig. 1, the test specimens were side-grooved at a 45 • angle using an Electrical Discharge Machining (EDM) technique.

Introduction of weld beads
A brittle weld bead was laid in the bottom of the slot to give an embrittled material in the starter notch to facilitate brittle crack initiation. The numerous challenges with producing suitable welds of high quality are explained in this section. Due to the large thickness of the specimens, magnetism was introduced in the specimen during the machining and welding processes. The magnetism caused the weld to be uneven due to arc blow or arc wander [28]. This was combatted through a combination of surface peening and careful welding technique. Before welding, the entire specimen was surface peened using a 2 lb ball hammer with a focus around the bottom of the slot where the weld would be laid. The manual metal arc welding technique was used to deposit a brittle weld bead at the bottom of the slot for each specimen using the hard-facing electrode Bohler FoxDur 350 with a heat input of about 1.5 kJ/mm. The weld was notched to 2 mm depth using the EDM technique to introduce a starter notch in the test specimens.
There were challenges with ensuring a sufficient thickness of the weld deposit as it was necessary to complete the weld in a single pass. It was also necessary to increase the heat input from that recommended in the standard (from 1 kJ/mm to 1.5 kJ/mm) to ensure the weld deposit was thick enough to be notched. Some of the specimens which had poor quality of welds were re-welded by initially using EDM to remove the old weld deposit before introducing the weld bead again. Although this gives a concern that the excess heat input may affect the test results, this was localised to the crack initiation region; hence this process is believed not to have affected the bulk of the specimen where the crack is propagating and, most importantly the arresting region, which is further away from the initial crack tip. Through an iterative welding process, the deposition of the weld beads on test specimens was finalised. An example of a test specimen with the weld bead, before performing the test, is shown in Fig. 2.

Test procedure
According to the experimental approach detailed in the ASTM E1221 standard, side-grooved CCA specimens are slowly cyclically loaded under crack-line wedge loading to incrementally increasing peak loads, in order to achieve a rapid run-arrest of a crack with a nearly straight crack front. The temperature range for the CCA testing was chosen based on the ductile to brittle transition behaviour of the material, with the NDTT (−50 • C) chosen as the initial test temperature and subsequent temperatures chosen iteratively. In order to perform CCA experiments, the sample was initially cooled down to the desired test temperature using liquid nitrogen, and subsequently loaded and unloaded repeatedly to a higher load point in each cycle. This loading and unloading sequence continued until a brittle crack initiated and arrested itself. This was evident through the loud noise it made and also a steep drop-off in the load reading as the crack jumped. Once the experiment was completed, the specimens were heat tinted to highlight the notch and arrested crack clearly on the fracture surface once the specimen was broken open, so that the arrested crack length could be measured. The arrested crack length was measured as the average of three points equally distributed across the thickness of the specimen, as suggested by the ASTM E1221 standard [6].
To facilitate the experiments, a bespoke test rig was designed and fabricated at TWI Ltd., UK, using high strength steel based on the design suggested in the ASTM E1221 standard [6], which is shown in Fig. 3. The test rig included cooling coils embedded into the support block through which liquid nitrogen was pumped to ensure a consistent temperature throughout the specimen. Due to the large thickness of the material being tested, the whole rig was scaled up, including the wedge and split pin assembly, to ensure sufficient strength in the test rig. This was directly scaled from the example dimensions which are given in the standard.
Each stage of the experimental set-up is shown in Fig. 3. Fig. 3-a shows the test rig, which is fully set up on TWI's 500kN capacity machine in Fig. 3-b. The specimen was inserted into the test rig and instrumented as shown in Fig. 3-c. The instrumentation included temperature measurements for all specimens and crack propagation monitoring wires which were used to measure the crack speed in two of the experiments. The case shown in Fig. 3-c is that with crack propagation monitoring wires applied to the surface of the specimen. As the crack propagates past the wire, it is broken and the signal is recorded and its speed can be measured. During the experiment, the low temperature was maintained by covering the specimen in thick layers of insulation, which is shown in Fig. 3-d.

Data analysis procedure
Eqs. (1)-(5) are used to calculate the stress intensity factor, K, shortly after arrest, which would be equivalent to K Ia when particular requirements are met, as described by Eqs. (6)- (10). In order to perform a CCA test, the wedge is loaded cyclically into the specimen with increasing peak load, and the CMOD is used to find K once an arrest event has occurred. According to the ASTM standard, the stress intensity factor for standard CCA specimen geometry can be calculated using Eqs. (1)-(5): where E is the elastic Young's modulus in MPa, B is the specimen thickness in mm, B N is the net thickness between the side grooves in mm, W is the specimen width in mm, δ is calculated from the CMOD using Equation (4) or Equation (5), and a is the crack length in mm (i.e. a 0 is the initial crack length and a a is the arrested crack length). In order to calculate K 0 , which is the stress intensity factor at initiation point (known as crack initiation toughness), the crack length is taken as a = a 0 , and δ = d 0 . Similarly, K a which is known as the crack arrest toughness is calculated using a = a a and δ = d a . δ, used in Equation 1, is calculated from the equations below using Fig. 4 as a reference to show how the parameters are calculated from the load-displacement curve over multiple cycles.
where n is the number of load cycles to the run-arrest event, δ 0 is the CMOD at the crack initiation point, δ p(n-1) is the CMOD at the start of the nth loading cycle when the load has been reset to 0, and δ a is the CMOD after the crack has arrested. It can be seen that as the crack propagates, the crack mouth widens before it arrests. According to ASTM E1221, when the following criteria are met K a can be taken as K Ia which is the critical stress intensity factor for crack arrest under the mode I fracture mechanics loading condition.
where N is the machined slot width, W-a is the uncracked ligament, σ Yd is the dynamic yield stress and σ Ys is the static yield stress. These criteria ensure that the specimen is of a sufficiently large size to satisfy plane strain conditions, and that the assumption of static behaviour during the crack jump event is appropriate, i.e. dynamic effects are not present.

Small-scale testing
Following previous research conducted by the authors [26], where small-scale testing was carried out, the predictions of the crack arrest toughness obtained from smaller size samples were validated against the large-scale results presented in this study. The small-scale tests in the previous study consisted of instrumented Charpy V notch testing to determine the post-fracture force during impact, and drop weight Pellini testing to measure the NDTT. These predictions rely on the determination of reference temperatures and are based on the principles of the master curve approach. The master curve is a statistical approach which gives a lower bound estimate of fracture toughness of ferritic steels for temperatures in the transition region and lower shelf from a limited amount of test data. This has been adapted by other researchers [25,[29][30][31][32] to predict brittle crack arrest toughness from small-scale testing using the following relationships: where T is the calculation temperature, NDTT is measured from Pellini testing [33], and T 4kN is the temperature at which the instrumented Charpy post-fracture force is 4kN. These curves can be compared to the fit to the CCA data, which uses T KIa as the reference temperature and is given in ASTM E1221 [6] by: where T KIa is the temperature corresponding to a median crack arrest toughness of 100 MPam ½ .

Validity of the test results
The results obtained from the CCA tests are summarised in Table 3 and compared against the predictions from small-scale testing in Fig. 5. While CCA2, CCA3, CCA4 and CCA8 specimens were tested satisfactorily according to ASTM E1221 and accurate crack arrest results were obtained from these four specimens, the remaining four samples resulted only in indicative crack arrest data as they were not fully qualified, due to a number of different reasons.
In the test on the CCA1 specimen, which was the first experiment conducted in this study, it was discovered that part of the loading rig had deformed during the loading of the specimen. This meant that the load recorded during the experiment was higher than the load applied onto the specimen, as some was lost to the loading rig. This did not affect the calculated crack arrest toughness as the load is not used as an input parameter into the calculation and the CMOD measurement was not affected. Although this issue was unlikely to have impacted the results, the test data obtained from this experiment was assumed to be indicative. The issue with the loading rig was fixed after this first test and this problem did not occur again in any of the other tests.
In the test on the CCA5 specimen, the crack did not propagate through the entire thickness of the test geometry, presumably due to the comparatively high temperature in this specimen, which meant that the crack was arrested easily in the material. Therefore, the results obtained from this experiment were considered to be indicative. Finally, CCA6 and CCA7 test specimens were side-grooved in a single side, to be able to accommodate wiring on the opposite plane side of the test piece for measurement of the crack propagation speed. Although this was taken into account during the calculation of K Ia , which includes consideration for the thickness between the side-grooves, the results obtained from these two test specimens were also considered indicative rather than strictly qualified to the standard.

Presentation of the test results
It is evident from Fig. 5 that there is a large scatter in the toughness results, and that not all of the test data points are valid. However, it can also be seen that both valid and indicative test data follow the same trend when the toughness data are correlated with the arrest temperature. Also seen in Fig. 5 is that the data points obtained from the CCA tests are entirely bounded by the two prediction lines which were made from small-scale testing. The prediction of the crack arrest toughness from NDTT gives a lower bound estimate, and the prediction from T 4kN gives an upper bound estimate with the CCA test data falling in between these two extreme trends. Also seen in Fig. 5 is that when the trend obtained from the test data points from this study is extrapolated to higher toughness values, the CCA trend remains between the prediction lines from NDTT and T 4kN . This suggests that the NDTT can be used to give a conservative estimate of CCA toughness results. However, more tests need to be conducted in future work to confirm the obtained trend from this study for higher and lower toughness values.

Fractography
Fractography analysis was carried out on all specimens post-testing. The fracture surfaces of one half of specimens CCA1-8 are shown in Fig. 6. It can be seen on the fracture surface of all specimens that heat tinting was an effective approach to mark the extent of crack growth during testing and before the specimen was broken open. The crack propagation occurred in a brittle manner in all specimens, which is evident from the relatively smooth and flat crack path observed in the fractography analysis on all test specimens. In specimens CCA6 and CCA7 (Fig. 6(f) and Fig. 6(g)), the crack showed tunnelling away from the surface, which meant that the crack propagation monitoring wires   were not broken as the crack propagated. This indicates that although these two specimens were instrumented for crack speed measurement, this could not be achieved in practice due to the lack of crack growth along the plane-sided face of the CCA specimen geometry.

Discussion
It was not possible to measure the crack velocity from these specimens. Due to the very low test temperatures (between −10 • C and −90 • C) it was necessary to keep the specimen under insulation which meant that it was not possible to use visual crack speed measurement methods (e.g. digital image correlation). For that reason, use of crack propagation monitoring wires is the recommended approach to measuring the crack speed. The focus of this research was on measurement of crack arrest toughness and ensuring that it was possible to obtain valid results before attempting to measure the crack speed, hence only two specimens were instrumented in this way. Now that the compact crack arrest testing approach has been proven successful, future work should focus on ensuring that the crack speed can be measured during these tests.
It is important to measure the crack propagation speed throughout the test as this is an important input into finite element modelling of brittle crack arrest which has been successfully carried out by other researchers [34][35][36][37][38][39][40][41][42][43][44]. The most common technique for finite element modelling of brittle crack arrest is using the nodal release method to replicate the propagating crack and extract the stress shortly ahead of the crack tip [39][40][41]45]. The critical stress just ahead of the crack tip is taken as the criterion for crack propagation, following a local stress approach [34,46]. Using this critical stress, predictive models can be developed. In future work, finite element modelling will be performed and validated using the experimental results.
The CCA results obtained from this study can be compared to those of other researchers who have carried out quantitative crack arrest testing on the same material. Although these experiments have never been done before on such thick sections of shipbuilding steels, other wide-plate tests have been carried out in previous studies using different methods [14]. Other researchers have consistently shown crack arrest toughness exceeding 6000 N/mm 3/2 at −10 • C (which is equivalent to 190 MPam ½ ) and this has been incorporated as a requirement of the International Association of Classification Societies [16,47,48]. Comparison of the results available on wide-plates with those obtained from CCA specimen geometries from the present study reveals that other test methods give a much higher threshold of brittle crack arrest toughness for the material examined in this study. This raises concerns that the CCA method may under-predict brittle crack arrest toughness results. Given the high strength and Charpy toughness of this material, it seems likely that it is the test method which is resulting in a low prediction of crack arrest toughness rather than the material itself. It would be necessary to carry out CCA testing and wide-plate testing on the same batch of steel in future work to have confidence in the agreement between the results of these two test methods. Alternatively, it may be possible to apply a correction factor to relate the CCA test results to wide-plate results for integrity assessment of large-scale structures. Last but not least, the presented results from this study will not be only of great interest to the shipbuilding industry but can be also employed in structural integrity assessment of other large-scale structures such as offshore wind turbine foundations [49][50][51][52][53][54][55][56][57][58][59].

Conclusions
CCA testing was successfully carried out on 80 mm thick EH47 shipbuilding steel to measure the brittle crack arrest toughness in this material. The results obtained from these tests were compared to the toughness predictions from small-scale testing and also wide-plate test results available from other researchers. The following conclusions were drawn from this study: • The present study has proved the possibility of achieving valid results from CCA testing on relatively thick (i.e. 80 mm) steel plates • The results show that Pellini NDTT data from small-scale tests can be used to obtain a conservative estimate of CCA test results • The results show that instrumented Charpy testing provides an overestimation of the CCA test results • Relatively smooth arrested fracture surfaces were observed in all CCA test specimens confirming that brittle crack arrest occurred in these experiments • The wide-plate test results available in the literature on the EH47 steel indicate that more conservative values of brittle crack arrest toughness may be achieved from CCA tests • More experiments should be conducted in future work on the same material batch to directly compare the CCA and wide-plate tests • Crack speed monitoring in CCA tests is challenging and may not be achieved by instrumenting a plane sided face of the specimen

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.