top of page

How artificial intelligence uses acoustic emission data to monitor corrosion

Corrosion causes major damage to infrastructure in many industries, such as the aviation industry and the automotive industry.

In an interdisciplinary research project with the Competence Centre for Electrochemical Surface Technology (CEST), the Institute of Structural Lightweight Design at the Johannes Kepler University Linz, and the Department for Integrated Sensor Systems at the Danube University Krems, Senzoro investigates how an artificial intelligence may use acoustic emission data to detect different types of corrosion such as pitting corrosion.

What are the challenges in training an artificial intelligence to detect corrosion?

The first and perhaps greatest challenge is to obtain meaningful data on corrosion processes; in particular, data on the extent and type of corrosion happening on a monitored structure within a given timeframe. For obtaining such data, various experiments need to be designed and carried out. Here we benefit greatly from the unique structure of our research consortium, which includes experts in corrosion from a wide range of scientific backgrounds. For example, in experiments carried out together with our partners from CEST, a metallic coupon is exposed to a sodium chloride solution. Simultaneously, a voltage is applied to influence corrosion processes. During the experiment, we collect acoustic emission data, and use inductively coupled plasma mass spectrometry (ICP-MS) to detect dissolved metals as an indicator for corrosion. Moreover, we monitor the surface of the metallic coupon with a microscope.

After obtaining meaningful data about corrosion processes, the second challenge is to bring the data into a suitable form for training an artificial intelligence. As a part of this process, it is often useful to reduce the dimensionality of the data by reducing it to its essential features. For this purpose, we use and further develop Senzoro's "Onion Analytics Layer", which has already been successfully applied for processing data from roller bearings.

Experiment setup (Source: Senzoro)

How is our artificial intelligence trained?

To provide the artificial intelligence with accurate information about the corrosion processes taking place, we apply data fusion. That is, we combine all available data sources, such as element counts from the ICP-MS, measurements of the electric current, and microscopy images, to obtain more accurate information about the extent and type of corrosion than each individual data source could provide.

Fenced standard deviation of the ultrasonic signal and electrochemical measurements (Source: Senzoro)

How is our artificial intelligence going to detect corrosion?

While high quality and expensive to obtain data such as the data from the ICP-MS is needed during training to give our artificial intelligence quantifiable information about corrosion processes,

our artificial intelligence is ultimately trained to be able to detect corrosion from data which is more easily obtainable, such as acoustic emission data.

ICP-MS Data (Source: Senzoro)

- Funded by the Take Off program of the Bundesministerium für Verkehr, Innovation und Technologie (bmvit) -

Author: Dipl. Ing. Mag. Markus Loinig, Senzoro GmbH


bottom of page