Use case in production

DeepCara: Acoustic-based anomaly detection using deep neural networks

Majid Mortazavi

Data Scientist, Siemens AG

Majid is a T-shaped Data Scientist with a result-oriented mindset. He is a mechanical engineer by formation holding a PhD in computational materials science with years of experience researching solutions on identifying sustainable energy storage solutions and atomic-level understanding of molecules used in pharmaceutical domains. Majid also has had a short Data Science engagement in e-commerce before joining Siemens. He currently works at the intersection of manufacturing and artificial intelligence for boosting automation, improving production processes by implementing algorithmic solutions namely for anomaly detection, visual inspection, and process optimization.

Majid Mortazavi
Majid Mortazavi
Session description

At the manufacturing line, the robot assembly often fails to assemble clamps into the designated slots due to several reasons causing excessive manual re-work. The goal is to develop an anomaly detector based on acoustic data to identify incorrect assembly of clamps in real-time. Triggered by the robot's digital inputs, the clicking sounds were recorded using a microphone and sent to a trained neural network for anomaly detection.