A Korean research team has developed a system that monitors manufacturing processes by analyzing sound from multiple devices in a factory. Companies will be able to reduce costs in building a smart factory and efficiently manage their manufacturing facilities.



On August 2^nd, Seoul National University College of Engineering announced that a team led by Professor Sung-Hoon Ahn of Dept. of Mechanical Engineering developed a monitoring system that converts sound data into 2D form with a log-mel spectrogram and uses Convolutional Neural Network (CNN), using a microphone installed at the factory. The research was published in the Journal of Manufacturing Systems.





 

This system is able to simultaneously monitor multiple devices at a loud factory. Unlike previous methods where training data is acquired through controlling movements of a specific device, this system compiles operational sounds from various devices for its training data. Dr. Jisoo Kim of SNU Dept. of Mechanical Engineering, the lead author, explained: “Previous technology installed sensors inside the devices to monitor theme at a time, but this new system can monitor multiple devices from the outside.” Prof. Ahn, corresponding author of the article, also said: “SMEs will also be able to monitor their manufacturing equipment at a lower cost. It is a new economical method for establishing smart factories.”