Neuron soundware monitors an automatic feeding system to prevent costly outages in automotive components production
The goal of the provided service is to keep the manufacturing process for recognized European automotive manufacturer running smoothly and detect failures that often arise with the automatic feeding system moving baskets around.
Challenge: provide early warnings for operators to shorten their reaction time and minimize automatic feeder outages
The production machine produces metal parts with a tact time of approx. one piece per minute. The machine is tightly integrated with the material and parts handling system (automatic feeding system, parts holders, loading units) into an automated production line.
In one part of the line, the automatic feeder is picking up baskets and moving them from place to place. Sometimes the baskets are stuck together, which the automatic feeder cannot detect. The feeder lifts these “stuck” baskets and the lower basket is released during lifting and falls down. Unfortunately, it no longer falls into the exact position and remains somehow rotated. During the subsequent removal of the pallet, an accident will occur and the basket must be repaired or discarded.
Production operators are not always present, as they have to supervise the rest of the production line. Incidents happen unexpectedly, on average twice a day. Due to the tact time of production, the response time must be very short (120 s). If the operators do not react within the limit, the baskets are damaged, which means downtime for the entire production line, which can subsequently jeopardize JIT deliveries to customers.
IoT technology allows operators to solve machine incidents in time
Neuron soundware monitors the robots instead of human supervision. It consists of sound sensors collecting data, the industrial IoT devices to process them, and Artificial intelligence for continual evaluation of the sound data. When an incident occurs, the solution provides visual and audible warnings on-site to prevent further damages to the baskets and the long-term failure of the entire working cell.
Neuron soundware solution learns continuously to increase the failure detection accuracy
“In order to provide the customer with information for instant decisions, we opted for neuron networks, which learn from acoustic data to recognize the noise made by a falling pallet. The longer the neuron network continues learning, and the more times it captures this noise, the more data will be available for improving the learning process and error detection accuracy”, explains Petr Ivančák, Project Manager from Neuron soundware.