Читать книгу Paint Analysis - Roger Dietrich - Страница 20

Once the initial data of the first inspection and the process parameters have been collected, there should be an informed guess about possible root causes. An informed (or educated guess) must be distinguished from wild speculations which are the result of missing knowledge. The informed guess together with information from interviews directly lead to the design of experiment (DOE) and “prior planning prevents poor performance”.

In fact, the design of experiment (DOE) is a key feature of a successful problem solution. If the DOE is poor, the results lead to the wrong direction.

The DOE exactly defines the type and amount of analyses to be performed and the expected results. So, at the beginning of a DOE there is the educated guess or let us say the theory what could have happened. To evaluate this question, it is good to have background knowledge about paint technology, paint chemistry and an overview about the analytical methods and their limitations. This limits the group of people that are meant to do the DOE due to their education. Planning an error analysis starts with the question: How do I start? Unfortunately, there is no standard procedure in the sense of:

 error type A = method A -> solution

 error type B = method B -> solution

Compare it to the hiking trip onto a mountain. You would want to take the shortest path with less obstacles and hopefully no deviations. At the beginning of the trail a few paths lie in front of you. But how to decide which path is the most promising? When it comes to failure analysis, it is good to keep in mind that coatings fail because something went wrong. You may want to start with an Ishikawa diagram [1] or a FMEA [2] to sort the influences on the coating quality to visualize the possible weak spots in the process. The author himself is not convinced that this approach is the best way to deal with the issue, because these methods operate with probabilities and it has been shown very often that the most improbable root cause has been the key issue of a failure event. Therefore, it is not good advice to rule out one cause just because there is low probability for it. The experience shows that one should beware of the mental mistake of believing exclusively in monocausal connections according to the pattern “if A, then B”. In reality, it is often the coincidence of several factors that influence each other and only in a certain combination lead to a coating defect.

For example: The author was called to help in a coating crater issue that appeared erratically with a certain type of polymer parts whereas other parts coated on the same line have not been affected. Following the most probable assumption based on these observations the polymer quality, and the quality of the paint very soon came into the focus. It was argued that there are parts which can be coated without or with a minimum of failures on the same line which seemed to rule out the paint shop as a possible source.

After some strategic sampling and using the box-in-box approach (see Part III) it finally has been shown, that the real root cause was a wrong design of the paint shop facilities with respect to the compressed air which was used for spraying the paint: For financial reasons the design of the pressured air supply was limited to a certain air flow volume. But this was not the main problem. In addition to that the contractor, who built the pressured air supply, did not install an oil-free compressor (which is highly recommended for spraying air supply). However, this also did not cause the crater issue, because there were oil traps and filters in the compressed air pipe. The trouble started because the plant operator accepted a contract to paint some parts that demanded a pressured air flow that exceeded the capacity of the supply. The result was that once these parts were to be coated, the compressed air loaded with oil dust flooded the oil traps and filters and finally the oil arrived in the paint booth and was sprayed together with paint onto the parts. So, it was not the paint material, not the cleanliness of the polymer parts, it was not simply the insufficient design of the painting facilities, and not only the failure of the contractor with respect to the not-oil-free compressor. With the “right” parts, the failure of the plant design and the mistake of the contractor would have never been obvious. The painting issue only occurred when the “wrong” parts had to be coated with “too much” air flow and the filters have not been checked and changed for a long time so that they were overloaded.

This example shows that the visible failure sometimes has a root cause which is a sum of complex interactions of small hidden anomalies which cannot be described by probability approach. In fact, when it comes to design an experiment/analysis nothing can be ruled out at the beginning of the process. To the authors experience looking over 25 years in failure analysis the “human factor” has always to be drawn into account. If, like the above mentioned case shows, the person planning a compressed air supply for the painting plant tenders the design of the performance of this plant only according to monetary aspects without considering or knowing the technical requirements, then it is possible that compressor oil is continuously or erratically pressed through the lines, which then cannot be retained by any filter. As a result, paint failures can occur at irregular intervals and completely erratically, when the next human error (accepting a contract that exceed the parameters of the plant) drives the system to its limits.

But back to the planning of the investigations. When a failure disaster such as a crater problem arises, one is faced with many of possible causes. And as always in life there are simple and easy to understand but unfortunately wrong solutions. Very often, supposed connections are then made from observations such as “since we have been using the new batch, craters have been appearing! So, the cause lies in the quality of the paint batch”. As described above, it is basically correct to include such observations in the considerations. However often it turns out that, for example, someone has neglected => that the new batch was also painted on a completely different coating line, under different conditions,

 that the components that are painted with the new batch are different from those painted with the “old” batch,

 that a new painting robot was installed or similar parameters.

Therefore, it is recommended while collecting facts and documenting them, only sufficiently valid information is considered. The next chapters provide information on the options of doing this and how to use them. In short terms, it is about how to go from a well-founded assumption to a certain knowledge and avoid confusing the one with the other.