Resultados de esclerometría del colegio nacional 7221 La Rinconada

Fig. 6.7 Variation of Time-of-Flight signals with defect through-wall size in the nozzle inner radius.

height (that is the depth of the defect measured in its own plane) but not crack ori-entation, so the two techniques complement each other for surface-breaking defects.

Time-of-Flight Diffraction can size cracks on the inside of tubular members whereas ACPD cannot.

6.6 PWR nozzles

In a pressurised water reactor (PWR), the inner radius of a coolant nozzle is not normally particularly highly stressed. However, in the unlikely event of a loss-of-coolant accident, cooler water is injected and this will impose severe thermal stresses on the inner radius of the nozzle. This means that the critical defect size is small, and defects with size down to about 6mm, considerably smaller than those which might affect safety, may need to be detected and sized in a component up to over 300mm thick, to provide a handsome margin between critical sizes and targets for inspection.

Defects sought are those which grow in planes containing the nozzle bore axis and these are difficult to size with conventional ultrasonic means with access limited to the inside of the nozzle.

The way in which Time-of-Flight Diffraction signals vary with defect through-wall size in the complicated geometry of a PWR nozzle inner radius is shown in

Fig. 6.8 The nozzle radius region showing disposition of the probes and coordinates used for an azimuthal scan.

Figure 6.7. In this figure, the A-scans have been widely separated in forming the B-scan image so that individual traces can be seen more clearly. Defects less than 5 mm deep are difficult to detect by casual examination of such a plot and more sophisticated processing would be required. The larger defects, however, are clearly visible.

In the UKAEA Defect Detection Trials [Watkins, Ervine and Cowburn, 1983b], one specimen, Plate 4, was intended to represent the complex geometry of the nozzle inner radius. It was made from SA508 Class 2 specification steel purchased from a reactor pressure-vessel vendor. Spark eroded slits and welding cracks were delib-erately implanted and then the surface was clad in two layers, either automatically using strip feed or manually, with wire feed, as appropriate. The surface was then ground to an Ra value of 3µm. Details of the defects and their geometry are given in Watkins et al. [1983b].

Defects in Plate 4 were specified as extending no more than 30 mm below the surface of the base metal. Previous experience had shown that this region could be covered satisfactorily with a single pair of probes, which were mounted on gimbals 35 mm apart with their line of centres transverse to the local axial plane, as shown in Figure 6.8.

The probes were highly damped 12.5 mm diameter compression wave transduc-ers operating with centre frequencies between 2 and 4 MHz, generating a short pulse of between 2 and 3 cycles. To provide coupling for the ultrasound, the test block was immersed in water in a circular tank. This was spanned by a specially constructed scanner, shown in Figure 6.9, having its principal vertical axis along the bore of the

6.6. PWR nozzles 119

Fig. 6.9 Schematic diagram of the scanner for inspection of the nozzle inner radius.

test block.

Data collection and control of the scanning were carried out by a computer. A rectangular scan raster of ψ and V coordinates (see Figure 6.8) was obtained by selecting a particular value of V and then incrementingψ by equal amounts through 360^◦. At each point on this mesh of points a portion of the time trace, 12.5 µs long following the arrival of the lateral wave, was digitised and stored. With a digitisation rate of 20 MHz this gave 250 sample points per trace. Signal averaging was used, summing several time traces from each probe position, to improve the signal-to-noise ratio.

A search scan was first conducted with a raster spacing of 0.4^◦inψ and 4 mm in V , giving a step size on the surface varying from 2.9 mm to 4.5 mm in the cir-cumferential direction and between 4 mm and 5.6 mm in the axial direction. Equal increments ofψ and V give rise to step sizes on the surface which depend on the absolute position of the probes on the surface because of the effect of the local ge-ometry. The data from the search scan were analysed using an image processing display system linked to a computer. The B-scan presentation was used to reveal defect indications through either perturbation of the lateral wave signal or through the obvious presence of diffracted signals.

Having identified the defect locations with the coarse raster scans, a series of fine scans, in the neighbourhood of detected defects, was used with a raster of 0.2^◦inψ and 2 mm in V . Zero crossings of the time waveform following the principal pos-itive peaks were used as the timing references and absolute travel times were used to calculate defect depths from this information [Stringfellow and Perring, 1984].

Provided the defect edge nearest the surface was more than 5 mm below the

inter-Fig. 6.10 Coordinate system and probe deployment for inspection of the nozzle to shell weld and the nozzle inner radius (from Curtis and Stringfellow [1986]).

face with the cladding and almost parallel with the interface then defect depths could be found accurately. Detailed examples of measured crack profiles compared with those intended are given by Stringfellow and Perring [1984] who observed that, in all but three cases, the results obtained for the through-wall extent of the defects were within 2 mm of the actual values, while the remaining three cases were within 4 mm. This corresponds to an average oversizing error of 1.1 mm with a standard deviation of 1.8 mm. Apart from the two carbon cracks, which had very uncertain

6.6. PWR nozzles 121

definitions of length, the Time-of-Flight Diffraction length measurements gave an average undersizing error of 2.4 mm with a standard deviation of 7.4 mm. This dif-ference in accuracy between measurements of through-wall extent and defect length is expected because the through-wall extent is obtained from a time measurement whereas the defect lengths were inferred from the appearance of the signals as the probes were scanned, in effect using a dB drop method. It must be remembered that it is the through-wall extent of these defects which is of most importance in es-timating the structural integrity of the component. Time-of-Flight Diffraction was shown, in this work, to be intrinsically capable of providing the degree of accuracy required for realistic safety assessments of component integrity, even in geometries as complex as the PWR nozzle inner radius, with the added complication of a layer of anisotropic austenite.

In work on PISC II Plate 3, an actual nozzle-to-vessel weld of a pressurised water reactor, the inspection with Time-of-Flight Diffraction was aimed at detecting, locating and sizing defects in the weld region. The defects were expected to be lying in circumferential planes parallel to the nozzle bore axis but inspections were designed to detect defects with any skew about a direction parallel to the nozzle bore axis. This was achieved with a design in which there were two separate probe arrays each capable of being mounted on a scanner head and rotated about the nozzle bore axis. The radial array consisted of 20 probes mounted in a plane containing the nozzle bore axis. Of these 20 probes, 13 acted as transmitters and 7 as receivers and these are shown in Figure 6.10. With this design all parts of the weld region in its plane were covered by a minimum of four transmitter-receiver pairs.

The transverse array covered an inspection plane at right angles to the plane of the radial array and was, therefore, intended to be most sensitive to defects lying in an axial plane, that is defects transverse to the weld. Two identical sub-arrays were used each with 3 transmitters and 3 receivers.

The PISC II Plate 3 contained 43 defects of which 30 were deliberately implanted planar flaws ranging in size from 3 mm diameter circle to a square of side 60 mm.

Another 4 implanted defects were of a composite nature consisting of clusters of planar defects with overall dimensions of 50 – 60 mm. There were 9 unintentional defects with through-thickness heights of 2 – 4 mm. All the deliberately implanted defects were circumferential in orientation, that is parallel to the local orientation of the weld plane. The whole inner surface of the assembly was clad with about 5 mm of austenitic stainless steel [PISC, 1986c].

Scans with the radial array used 0.25^◦ steps, corresponding to displacements along the surface of about 3 mm at the weld centreline. At each position A-scans from 38 transmitter-receiver pairs were recorded, giving at least 4 transmitter-receiver combinations contributing to defect detection and location of sub-surface defects or those near the back wall, while giving up to 20 combinations of probes at mid-wall.

Pitch-catch reflection data were also recorded to assist in radial definition of defect positions. The radial coverage was from at least 695 mm out to 825 mm or more, giving inspection of at least 65 mm of the weld material and base metal either side of the weld centreline at about 760 mm radius.

Each trace was digitised at a sampling rate of 20 MHz. In order to get adequate signal-to-noise ratios, 128 traces were averaged for each probe pair and each probe position.

After analysis and reporting, five defects had been missed of which three were not more than 3 mm deep by 12 mm long. These three were not considered serious.

However, two defects missed were near-surface defects, each a 10 mm diameter cir-cle. These were just too deep to be seen by a closely spaced pair of probes such as that at 40 mm separation and they were just too shallow to be detected by a widely spaced pair such as those at 140 mm separation. It is clear that these defects would have been detected correctly with a pair of probes spaced at an intermediate value between 40 and 140 mm, say at 80 mm. The accuracy obtained for the through-thickness measurement was within±2 mm or better for about half the defects or within about±10% for the larger defects. Such errors were consistent with normal errors of measurement, whereas for the remaining defects, which were sized less well, the errors were due to misinterpretation of the various diffracted signals. For real reactor inspection the errors would be smaller because supplementary data on defect detection and sizing would be utilised.

The results for this inspection, together with those obtained on the PISC II flat plate (Plate 2), have been reported by Curtis and Stringfellow [1986]. They con-cluded that the Time-of-Flight Diffraction technique was capable of detecting and sizing defects in girth welds of pressurised water reactors with a high degree of reli-ability. To achieve similar accuracy and similar performance for near-surface defects in the inspection of nozzle-to-vessel welds it would be necessary for the clad inner surface of the vessel to be of higher quality than that of PISC II Plate 3.

Because the signals diffracted from the defect come essentially from the edges of the defect, the technique is less sensitive to the roughness of the defect faces than conventional pulse-echo techniques. Curtis and Stringfellow [1986] could find no difference between the diffraction responses from rough and smooth defects.

PISC II Plate 3 was also inspected by Risley Nuclear Laboratories using an au-tomated ultrasonic technique comprising high sensitivity pulse-echo detection and predominantly Time-of-Flight Diffraction sizing. These techniques were deployed from the clad inner surface of the nozzle and made use of digital data collection, analysis, and display. With this system Risley Nuclear Laboratories detected 30 out of the 31 intended weld flaws and correctly located all 3 of the nozzle corner defects.

With Time-of-Flight Diffraction sizing they achieved a mean size error of−1.3 mm and a standard deviation of 7.0 mm when their results were compared with the in-tended defect sizes of the 31 weld flaws [Rogerson, Poulter, Clough and Cooper, 1988]. This illustrates the way in which, for critical applications, the conventional pulse-echo techniques and the Time-of-Flight Diffraction method can provide di-verse ways of size measurement, thereby enhancing confidence.

For complex geometries such as the nozzle to vessel weld of a PWR inlet noz-zle, it has been found advantageous to use a mathematical model of the inspection geometry in order to display the signals in their correct relationship to the struc-ture [Poulter, 1986]. On the PISC II nozzle, Risley used Time-of-Flight Diffraction

6.6. PWR nozzles 123

Fig. 6.11 Close-up of the Time-of-Flight Diffraction technique crawler on the RTD plate.

sizing of the defects which they detected using pulse-echo techniques and found a mean sizing error of−0.14 mm, with a standard deviation of 3.0 mm, when compar-ing their results with the intended defect sizes [Poulter, 1986].

As well as scanners designed to fit standard in-service inspection masts for ge-ometries such as the nozzles of a pressurised water reactor, there is also a requirement for inspection devices which can be easily adapted to a variety of inspection tasks.

Such devices are usually, in effect, miniature vehicles which can traverse a compo-nent, carrying a probe assembly, under some form of guidance. For nozzles or pipes, the vehicle would usually be attached by straps or chains, allowing circumferential and possibly limited axial travel. Where that form of restraint is inconvenient, mag-netic attachment can be used, when the component is ferritic, and the vehicle can be guided by a marked track which it follows optically. A vehicle of this type, generally referred to as a crawler is illustrated in Figure 6.11 operating on the RTD plate. This plate, so named because it was supplied by Röntgen Technische Dienst, is a part of

Fig. 6.12 Images from MUSE data reconstruction, showing TOFD data in uncor-rected format and selected data mapped into the component (reproduced from Daniels et al. [1996]).

a boiling-water reactor (BWR) pressure vessel shell containing a nozzle.