To attain a complete elimination of the oxide from steel strips surface a mandatory production step - the
pickling process - is needed in order to make further treatments (e.g. hot deep galvanizing) and to maintain
quality standards. One of the main problems during this process is the difficult evaluation of the pickling state.
If on the one hand it is evident that high strip quality requires a complete elimination of the oxide on the strip
surface, thus under-pickling cannot be tolerated, on the other hand a too long permanence of the strip in the
pickling bath can lead to an erosion of the steel strip surface itself (over-pickling). Both these problems lead to
more expenses or losses: in facts under-pickled strips need to be re-processed, while the quality of over-pickled
ones must be downgraded. The main issue in the detection of over and under pickling defects by means of
historical data analysis lies in the unbalance of correctly pickled strips, which are the most, and the class of
defective ones, which makes the application of standard classification methods ineffective.