'Fingerprinting' saves metallic equipment.

Over the past decade, scientists have developed techniques to fingerprint everything from DNA found in blood at crime scenes to magnetic particles in credit cards. Now, chemical engineers at Washington University have discovered a method to fingerprint one of the biggest villains in industrial processing corrosion.

Using a multi-faceted signal processing technique called wavelets, or wavelet transforms, Babu Joseph and Rodolphe L. Motard, professors of chemical engineering, and Xue-dong Dai, a graduate student in chemical engineering, have created a method that detects two different types of corrosion in metallic equipment - one caused from pitting; the other from crevices.

Their new technique is designed for situations where liquids and metals are in contact in flowing streams, such as pipes in a variety of industries, from oil and gasoline manufacturing to power plants. It would not be applicable for corrosion in automobile mufflers or paint finishes, for instance, or other machinery exposed to various environmental substances, but that does not rule out the possibility that wavelets some day may be a part of such preventative systems.

The wavelets are sophisticated algorithms that interpret electrochemical (and many other) signals through time and frequency of occurrence. In experiments at Motard's Computer-Aided Process Engineering Laboratory, the researchers were able to decipher electrochemical "noise," signals that previous signal-processing methods could not detect. With wavelet analysis, they identified what kind of corrosion was taking place in the metals and at what rate it was occurring.

The technique could help prevent corrosion-induced equipment breakdowns that cause accidents and hamper productivity. The monitoring sensor would tell process-control engineers where corrosion is taking place in a pipe, vessel, or machinery part and when to replace the part before damage reaches a certain level. It is...