Читать книгу Paint Analysis - Roger Dietrich - Страница 23

The evaluation (data oriented) and the assessment (process oriented) of the analytical results are two steps that must be strictly separated. Both steps can and should only be carried out by experts.

The data evaluation includes:

 the purely technical processing of the data

 checking the calibration

 checking for plausibility and measurement errors

 the comparison with databases,

 the correlation of the measured data with tables

Whether, for example, an infrared spectrum is “reasonable” or whether a beautiful but completely meaningless spectrum has been produced by preparative errors or previously unknown material properties, can only be decided by someone who understands the technology and has the theoretical materials background.

In addition, the comparison with databases (uncritically and automatically used) bears considerable risks and imponderables. One reason for this is, that database comparisons are purely mathematical routines. To put it pointedly, the computer compares zeros and ones. The result is a selection of database spectra with a hit quality index, which reflects the mathematical probability that the measured spectrum matches a stored spectrum. There is no plausibility check. This often leads to very bizarre results. If you rely on this probability indices list without checking them and build a theory about the course of the damage and how to remedy it, it is not unlikely that you will not find the cause in a harmless case or, in the worst case, that you will cause great damage by taking the wrong corrective actions. Furthermore, the algorithms only compare the measured spectrum with stored data. However, databases usually become obsolete faster than the seasons change. If, for example, a binder is entered in the database under a trade name X, then the manufacturer may be sold to another company, the product name has been changed or the same binder has been sold under a different trade name with a slight modification. Thus, the corresponding database entry is worthless. Thirdly, databases fail when analysing mixtures! A spectrum of a primer with a high filler content is then “identified” as barium sulphate, for example, because the signals of the barium sulphate stand out prominently in the spectrum. As a result all other paint components are no longer recognized by the database.

The evaluation of the measured data is followed by the assessment. The assessment should be carried out in several stages and is process oriented. Finally, it has to answer the core question of whether all outstanding questions have been answered by the analyses carried out, whether correctives actions can be derived from them or if further investigations are necessary:

Every analytical method used to test a sample for its composition has certain so-called “blind” spots. This means that each method (due to the physical limitations of the methods) will deliver only an excerpt from reality.

No analytical method can determine exactly the complete, comprehensive and “true” composition of a sample. To achieve a maximum of information a concert of different methods is needed. Analysing a sample by SEM-EDS, for example, leads to the detection of the elements in the uppermost μm of the tested area. But it does not tell anything about the molecular structure. Detecting the elements carbon, oxygen and calcium by SEM-EDS e.g. leaves the molecular options of calcium carbonate as well as a calcium soap of an organic acid. To distinguish between these possible structures infrared spectroscopy is needed.

This means, when selecting the analysis methods which is used to get closer to the real composition of a sample, it is good to know these “blind spots” (i.e. the information that cannot be detected) and weigh up to what extent the “omission” of this information represents a relevant statement or not. The more preparation steps, analysis steps and evaluation steps are applied from the initial sample to the test result, the more information is lost. It is good to be aware of this fact when evaluating all analysis results (from whatever laboratory).

And another fact follows from this insight: Analysis results that are not carried out on the same samples, with the same preparation methods and the same analysis procedures are not comparable in every sense.

The following gives an example to illustrate this:

Paint craters occurred on a paint shop after the production output was increased. Initial tests and random analyses indicated that the paint air (i.e. compressed air) could possibly be contaminated with paint wetting interfering substances (PWIS). The best way to check this is to take a sample and analyse the activated carbon filters used in the compressed air line. For this reason, two independent laboratories examined the same activated carbon filter sample of the compressed air supply. The aim of these tests was to obtain information about whether paint wetting interfering substances (PWIS) are present in the compressed air.

Table II.2: TD GC-MS and TOF-SIMS analyses of the same activated carbon filter from a paint shop
	Laboratory A	Laboratory B
Sample	carbon filter granules used and loaded
Analytical method	thermodesorption GC-MS	extraction + TOF-SIMS
Detected substances	cyclohexane, pentane, 1,2,2-trichloro-1,22-trifluoroethane, 2-methylbutene, ethyl acetate, methylcyclopentane,methylcyclohexane, xylene	palmitic acid ester, stearic acid ester,tetraethylene glycol dicaprylate, polydimethylsiloxane

The results of both analyses are listed in Table II.2, and seem to be unequal and contradictory. How can this happen? Did one or even both laboratories work inaccurately? The answer is “No!” Rather, the method selection limits the results to one aspect of the real composition.

One of the laboratories used the TOF-SIMS method for the analysis of the extracts from the activated carbons, because it has an extremely high sensitivity to PWIS (paint wetting impairment substances). It was deliberately accepted that volatile compounds are not detected by the TOF-SIMS method.

The second laboratory, however, used a combination of thermo-desorption and GC-MS to examine the activated carbon. Some PWIS cannot be detected by thermo-desorption, because they are simply not thermally desorbable. After thermo-desorption/extraction (which are only separation methods but no analysis methods), this laboratory carried out an analysis and identification using GC-MS. It must be taken into account that GC stands for gas chromatography which is another separation method. By this separation method, the mixture of substances absorbed by the activated carbon is separated on a so-called column into individual components or groups of components (see Part IV, Chapter 5.12.2). Here again, there is discrimination in the sense that not all substances that are injected at the beginning of the column necessarily leave the column at the end. Depending on the choice of column and the temperature program during the chromatographic separation of a mixture, very different results can be obtained.

The GC-MS analysis selected for the second laboratory ends with a mass spectrometric analysis of the substances separated from the GC column. Here the result of the identification depends very much on how it was carried out. In many cases, the mass spectra obtained are automatically compared by the computer according to purely mathematical criteria for similarity with stored spectra from a database and this degree of mathematical agreement is output in a so-called “hit quality index”. To our experience, however, a purely mathematical match of 80 % may occur, but the substance “found” above this level may be complete nonsense in the context of the investigation. This means that the quality of the identification depends on whether the substance being searched for is actually present in the database. However, many PWIS are not included in commercially available databases.

Based on the originally formulated question, which substances are present in the compressed air, the available results must be assessed as follows:

 It can be assumed that the detected substances were present in the compressed air (provided the activated carbon did not already contain these substances at the time of installation).

 If a substance was not detected, this does not mean that the compressed air is free of it!

The following must also be considered: A conclusion from the composition of the substances adsorbed on the activated carbon to the complete composition of the compressed air can only be drawn if

 all substances present in the compressed air are adsorbed on the activated carbon without discrimination,

 all adsorbed substances are completely and non-discriminatingly transferred to the analysis,

 the chosen analytical method detects all substances without discrimination.

However, precisely these requirements are not met in the present case. This example shows which pitfalls lurk in a supposedly simple analytical task. Of course, this can be applied to all methods.

To ensure that the results can be interpreted and implemented in the process, the analytically obtained facts must be combined with knowledge. This intelligence can be the knowledge of internal specialists (process knowledge, expert knowledge) as well as the expertise of specialists outside the company. It is important to carefully distinguish between assured knowledge and assumptions, and it is no good advice to leave the knowledge and intuition of the employees on site at the production lines unused. A qualified brain should generate a quintessence from the information from all these sources, this involves answering the following questions:

 What is the desired objective?

 Which of the substances found in the defect originates from the process and which is to be classified as an impurity?

 Which substances are used in which process step?

 Can the detected substances explain the origin of the defect plausibly and without contradiction?

 Are there any gaps in knowledge that need to be filled by additional analysis of further samples?

If all measurement data and process data can be combined consistently without contradictions, then the solution is not far away. The last step is to check the knowledge gained in practice and to define remedial measures.

In the case of surface and material defects, the instrumental analysis of samples is a necessary, helpful but not sufficient method to get to the bottom of the cause of the defects. Even though it is associated with costs, it is always cheaper than shirt-sleeved experiments based on speculation. An effective problem solution can only be generated when internal expertise, external knowledge and measurement data are brought together to form a synthesis.