·Table of Contents
·Reliability and Validation 1
Surveillance and certification of industrial x-ray films - an international project for quality assurance
U. Ewert, J. Stade, H. Heidt , BAM Berlin,
B. Vaessen, J. Snels, M. Ailliet, W. Markie , Agfa-Gevaert,
Nobody would drive his car in the rain without using wind screen wipers, - why a radiograph which is too grainy should be tolerated if the faint image of a dangerous crack in a pressure vessel is to be detected ? The safety and reliability of industrial plants, components and products depends on the quality of the non-destructive testing methods. In radiography the sensitivity for the detection of flaws is strongly correlated with the image quality of radiographs and this is influenced in a decisive manner by the properties of an X-ray film system. Therefore these properties have to be achieved and supervised with great care.
Already in the seventies the films on the market were investigated and classified in four different classes . The sudden increase of the price of silver 1980 forced all manufacturers to react upon and to add chemicals to the emulsion which gave the films still the desired good properties if the suitable developer was used. But now the films were sensitive against the use of developers of other manufacturers and could show in this case noticeable changes in gradient, granularity and speed. Hence, the film, developer and the processing procedure have to be considered as a system which determines the properties. As a consequence of this situation and to ensure the quality of radiographic testing, a system of standards as ISO 11699, EN 584, ASTM E 1815-95 and JIS K 7627-97 which fix the minimal demands on industrial X-ray film systems, has been elaborated in close co-operation of film manufacturers, users and independent institutions. These standards became valid within the last 4 years. The film system classes defined in these standards allow to choose the film system suitable for a certain inspection procedure. However, surveillance and certification of film systems is necessary. In the following some inherent problems and difficulties of this task, results of measurements and conclusions are described and solutions are proposed.
2. The new situation
The radiographic image quality which is attainable with suitable X-ray film systems is still the measure for all new X-ray imaging systems and detectors as fluoroscopic systems, line scanners, imaging plates and flat panel detectors. These systems have been developed in the meantime to a high stage and their image qualities respectively the combination of their detection sensitivities with other properties allow the solution of various special inspection problems, whereas the film is preferred in a large field of applications because of its image quality, reliability, robustness, mobility and archivability.
But to keep film systems at this position, the design parameters of new or changing products (film and chemistry) have to be kept at the classified level and the film systems have to be surveyed at the moment of introduction and during product lifetime. The survey must be performed at design and during manufacturing in an open co-operation according to ISO 9000 philosophy. Therefore BAM has come up with a proposal for a new film certification programme with classification of the film system and surveillance of the production process and production consistency thereby covering the total of the certified system.
With regard to mixed systems, i. e. film systems in which films and chemistry of different manufacturers are used, the qualification and surveillance is already an issue because nobody can take responsibility for the system. System components may not be matched, and over time changes may occur which will influence the classification and the quality performance of these film systems. Mixed film systems will need to be addressed in a different way.
3. Film classes and visibility of details
For the further discussion some facts are reminded and estimations are made to give an impression about the influence of chances in a film system on the detection sensitivity.
The human eye can discern differences of the optical density down to DD = 0.01 at density D = 2. This is of the same order of magnitude as the standard deviation sD of the optical density of NDT X-ray films, as for example the limiting values of the granularity sD of the classes C1 to C6 of the standard EN 584-1 vary between 0.018 to 0.039. Expressed in percent the granularity of the different film system classes differs only between 14% and 28% from class to class. These values are comparable to the quite usual variation of the granularity of 10% to 26% if a film is processed in a developer of another manufacturer (so called "mixed system").
An additional insight can be obtained, if one calculates the smallest thickness difference Dd of steel which can be discerned on a radiograph whose granularity is just as high as the limiting value sD of the corresponding class of the standard EN 584-1. A signal-to-noise ratio of S/N ³ 2 is assumed as limiting condition for a clear perception of a density difference. The estimation of the smallest perceptible thickness of steel at those granularity levels is given in Table 1.
Flaws under these dimensions will be below the critical signal-to-noise ratio of 2:1 for a given film system class. One can see that for instance for a borderline film of class C5 a flaw must be already 21% deeper for perception than for a C4 film. In case of an increase of granularity by 20% -as for some mixed systems- the estimation results in perceptible flaws which are 20% deeper as the values in Table 1.
|Dd in mm
|Table 1: Estimated minimum values of thickness differences Dd of steel of small lateral extensions visible in granularities equal to the limiting values of the film system classes.|
170 keV, 10mm steel, D-Do = 2
These estimations give only the order of magnitude of the effects caused by the influence of the film system class and chemistry on the visibility of flaws, but they help to realize that this influence is of importance.
4. Steps toward a system for quality assurance for industrial X-ray films
4.1 A round robin testThe results of this first round robin are :
With the standard EN 584-1 manufacturer and user have a tool which gives definitions of the film system properties, methods and prescriptions of the procedures to measure them and the demands which have to be fulfilled for the different film system classes. The problems pointed out in the above sections indicate however, that it can be only the common base and frame of a system for quality assurance for industrial X-ray film systems. A continuous surveillance and certification of the film systems on the market within a system of quality assurance is necessary. This system should be open to all manufacturers and users of X-ray films. As a result of discussions between manufacturers, users and the independent third party institution BAM a system was proposed which is based on round robin tests to harmonize the equipment and procedures for the measurement of film system parameters and on surveillance of the production. The check of measuring equipment and procedures is an important requirement for further steps to provide the users with constant film quality. The tests allow to analyse the reasons of different results in different laboratories and the development of better measuring procedures and algorithms.
Such a round robin test comprehends investigations on the potential influence of the exposure process, the process of development, the process of measuring and on the influence of the algorithms for evaluation. It has been performed already at BAM and Agfa-Gevaert .
- Neither the exposure process nor the processing in the two laboratories caused significant differences in the values for the gradients.
- Differences were found for the gradient D-Do = 4 due to different calibration of densitometers at densities above D = 4 .
- The values for the granularity measured by BAM were always greater than those measured by Agfa-Gevaert. The difference originated in different numeric procedures for the sD calculation.
These first results of the round robin test confirmed the necessity of a round robin test concerning the equipment and procedures as a base for quality assurance. The difference in the results for granularity were of such a magnitude that the same film type could be classified into different film system classes by BAM and Agfa-Gevaert. Each partner of the round robin trusted in his results and was convinced that his equipment and algorithms had the best performance. But of course, it is not a question of wrong or right. It is simply the problem that EN 584-1 allows a variety of procedures for the sD measurement. This analysis had to be done in an additional cycle of the round robin test with exact and fine definition of each step of the algorithm.
4.2 Analysis of procedures
Both laboratories received slightly different results for the gradients although both laboratories used the same procedure as prescribed in the standard EN 584-1. By comparison and exchange of measured data and the crosswise evaluation it became clear that two differences in the procedures contributed to the different values of the gradients:
- while in the algorithm of BAM the data set for the film curve was supplemented by density zero (without fog) for the dose zero this was not done in the algorithm of Agfa .
- the densitometers of the laboratories measured different values for densities over D = 4.5. The instrument of Agfa was better calibrated here.
Differences in the granularity can arise due to deviations in the aperture of the microdensitometer and analysing software. A deviation of a few micrometers from the correct size of 100 mm would result in a few percent difference in the values for granularity. As the real mechanical size of the aperture of microdensitometers is reduced to the effective size in the image plain by the magnification of the microdensitometer, some efforts are necessary for the measurement of this size. With a special procedure developed at Agfa-Gevaert, the effective sizes of the apertures of both instruments could be determined. From the results of these measurements it became clear that the differences in the values for granularity had to have other reasons. Therefore, the algorithms for the calculation of the granularity were analysed. For this purpose the same film samples were measured in both laboratories and the data sets were exchanged for crosswise evaluation with the own algorithms and those of the partner. One result was that the conversion of the measured values of specular density into diffuse density had to be performed by direct calculation of each measured value from the Callier curve. It turned out to be important in which way this conversion was done and when it happened in the algorithm. The difference in the results could rise up to 4%.
The way how the granularity was calculated could also influence the result. The standard EN 584-1 prescribes only a spatial filtering of the measured values before sD is calculated. Due to the great number of checked film samples to be tested normally by the manufacturer, it can happen that some of the measured film samples contain scratches or dust which would result in far to high granularity values in these cases. Therefore, sD was calculated as the median at Agfa-Gevaert as originally proposed by Buhr et all. and included in the working draft ISO 10505 WD #3 . The result of this median calculation is robust against minor scratches, dust and spots on the film sample. Depending on the number of groups into which the data set is subdivided for the calculation of the median this algorithm can yield results which are not equal to the value of sD as result of the normal standard deviation.
4.3 New algorithm for the calculation of granularity
Based on the experiences of the round robin the following algorithm for sD -measurement is proposed:
This algorithm should be one base of the surveillance and certification of industrial X-ray films.
- step exposure of 6 or more films
- measurement of specular and diffuse density
- third order polynomial fit for calculation of the conversion function from specular density into diffuse density
- scan of exposed steps of density ~2, at least 20 mm per scan line, step width 0,1 mm
- conversion of the scan data into diffuse density values
- digital filtering of this values by a high pass filter with a cut off frequency of 0,1 lp/mm
- grouping of the complete data set of more than 1200 data points into at least 60 groups with a group length of 2 mm
- for each group the standard deviation is calculated and the median of all single group values is determined including the suitable statistical factor demanded by the ISO working draft
5. Results of sample test measurements
Parallel to the round robin test the BAM made additional measurements on film systems procured on the market to get an overview of the situation and a broader statistical base. The film systems of the manufacturers (original systems) were processed according to the prescriptions of the manufacturers. For mixed systems, where film and developer originate from different manufacturers, one of these procedures was chosen without claiming it as optimal for this special mixed system. For reasons of comparison to former measurements the procedures and algorithms of BAM were not changed.
While the results of these measurements indicate nothing striking for the gradients they indicate for some film systems some remarkable changes of granularity. The results prove the necessity of surveillance and certification of X-ray film systems and the harmonization of the measurement equipment and procedures by round robin tests. In the following the results concerning granularity and gradient-to-noise ratio will be given.
The first figure shows the granularities of the systems as they were measured by the manufacturers themselves (BAM measured corresponding values) in a round robin of 1988 which was one important base of the standard EN 584-1. Exact the results of this round robin were used to define the limiting values of the film system classes.
Fig 1: Granularity values of film systems measured by the manufacturers in a round robin 1988. These values were used to define the limiting values of the system classes.
Fig 2: Granularity values of film systems measured by BAM 1998/1999/2000. The system classes are cited according to the manufacturers
The results of measurements of the granularity made by BAM in the years 1998/1999/2000 are shown in figure 2. It can be seen that the granularity of some of the film systems of the manufacturers has increased nearer to the class limit and that some of the values are even beyond it but within the +10% error range of the standard EN 584-1. If mixed film systems are included as in figure 3 the granularity varies considerably, mostly to the bad side.
Fig 3: Granularities of film systems including mixed systems measured by BAM 1998/1999/2000. The original systems are classified according to the manufacturers.
Fig 4: Granularity values of film systems measured by the manufacturers in a round robin 1988. These values were used to define the limiting values of the system classes.
Concerning the gradient/granularity-ratio a similar trend can be stated. In figure 4 the results of the round robin 1988 measured by the manufacturers are shown. Again these values were used to set the limiting values of the film system classes. These results can be compared with values BAM measured for the systems in the years 1998/1999/2000 as shown in figure 5. Again some values are nearer to the class limit and some systems are beyond it.
Fig 5: Gradient/granularity-ratio of film systems measured by BAM 1998/1999/2000. The system classes are cited according to the manufacturers.
Fig 6: Gradient/granularity-ratio of film systems including mixed systems measured by BAM 1998/1999/2000. The original systems are classified according to the manufacturers.
The effect of mixed systems is shown in figure 6. By comparison with figure 5 it can be seen that the gradient/granularity-ratio of a mixed system becomes mostly worse than for the original system but in some cases it becomes even better.
Changes in the granularity are related with changes of the speed of a film system. With respect to the results above it would be of interest how they are mirrored in a diagram where the granularity is put into relation to the dose for a certain density. In figure 7 this is done in a special way: The ordinate gives the values of 2/sD for the film systems as a kind of signal-to-noise ratio over the square root of the dose for D-Do = 2 as a measure related to the quantum noise. On a first glance one would expect a straight line as relation between these two quantities, but as figure 7 shows this is valid only on the whole but not in detail. The relation is ambiguous as for some dose values, more than one value of 2/sD is possible. This is especially true if mixed systems are included and has consequences for the application of the film test strips according to EN 584-2. These film test strips are used for a quality check of a classified film system where film and developer originate from the same manufacturer, their use as quality check for unknown combinations of mixed systems has to be discussed.
Fig 7: The values of 2/granularity (~ signal/noise-ratio) in relation to the square root of the dose for D-Do = 2 (~ quantum noise) for different film systems inclusive mixed systems.|
The results of this international project for quality assurance of film systems as they are described in the sections above give evidence for the necessity of surveillance and certification of industrial X-ray films. As shown above is that only possible with an obligatory round robin test for the harmonization of the equipment and the algorithms for measurement. Obviously, the prescriptions of the standard EN 584-1 are not precise enough to enable identical results in different laboratories, therefore a revision of this standard seems necessary. A proposal for a possible procedure to measure the granularity is given in section 4.3. The participation of other film manufacturers and users in a round robin test for the measurement of such film system properties would be a step forward towards an international accepted surveillance and certification system. Interested parties are invited to such a round robin test and the further analysis of the present status of film systems.
The continuous surveillance of all important film system properties by the manufacturers supported by periodical measurements of film samples and chemistries from production by independent third party institutions like BAM will be a contribution to the stability of the radiographic testing quality. Together with an obligatory round robin test for the harmonization of the measurement equipment the surveillance will be a presupposition for a long term certification of each film system.
- D. Schnitger, E. Mundry : "Über die Klassifizierung von Röntgenfilmen"
Amts- und Mitteilungsblatt der Bundesanstalt fÜr MaterialprÜfung 1, 1970, No.4
- J. Stade, U. Ewert, D. Schnitger, J. Snels, W. Markie, M. Ailliet: "Klassifizierung am Markt befindlicher Röntgenfilmsysteme", Proceedings of the annual conference 1999 of the German Society of Non-destructive Testing, Celle 10.-12. May 1999, Vol. 68, 2, p.663-673
- E. Buhr et all. PTB Mitteilungen 101 3/91 p. 183-191 , 1991,
ISO 10505 WD#3 Photography - Root-mean-square (rms)-granularity of photographic film - method of measurement