AI-based quality control system listens for clicks on assembly lines
When two interlocking parts of a product are joined together in a factory, they typically make a satisfying "click" sound. By listening for such noises, a newly-developed system can detect when faults have occurred on robotic assembly lines.
Created at Germany's Fraunhofer Institute for Digital Media Technology (IDMT), the technology incorporates audio sensor units that can be mounted either directly on a robotic device, or on a nearby wall or other surface. Each unit is "no larger than a pack of cigarettes," and contains a microphone, a battery, and a microprocessor that analyzes audio signals right within the module.
After initially being trained to recognize the sound of specific components being correctly joined, the system is subsequently able to tell when a robot has gone through the motions that should have resulted in a successful coupling, but the sound has not occurred. It then displays an error message, alerts a human supervisor, and records the incident.
Utilizing artificial intelligence-based algorithms, the setup is additionally able to disregard irrelevant background noise, plus it can distinguish between multiple target sounds. This means that a single sensor could be used to check for the successful joining of a variety of components.
"In effect, we give machines a sense of hearing for quality assurance," says Danilo Hollosi, Head of Acoustic Event Detection at IDMT. "This allows manufacturers to punctually recognize signs of damage at an early stage, to reduce unscheduled downtime, to harmonize shop floor workflows and to increase the effectiveness of the overall plant."