Jam detection in waste sorting conveyor belt based on k-Nearest Neighbors - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Jam detection in waste sorting conveyor belt based on k-Nearest Neighbors

Résumé

Waste sorting is a complex task because the proportion and composition of incoming waste (humidity, weight …) is unknown, so the operators suffer from the events. No efficient method is used to detect when a conveyor belt will be jammed by product overload. This research aims to miss as few jams as possible to avoid production shutdowns and to ease the operators, who can patrol up to 12 km per shift and have to manually unstuck the waste after a jam. Missing fewer potential jams means more sorted waste and less incinerated waste. With the possibility to classify the data by waste sorting experts, a supervised approach, a k-Nearest Neighbors (k-NN) is chosen to confirm if it is possible to use Artificial Intelligence to detect a potential jam. The details of the method implemented are given and the results obtained on field data are analyzed.
Fichier non déposé

Dates et versions

hal-04109630 , version 1 (30-05-2023)

Identifiants

Citer

Calliane You, Olivier Adrot, Jean-Marie Flaus. Jam detection in waste sorting conveyor belt based on k-Nearest Neighbors. 32nd European Safety and Reliability Conference ESREL2022, Aug 2022, Dublin, Ireland. ⟨10.3850/981-973-0000-00-0_esrel2022-paper⟩. ⟨hal-04109630⟩
21 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More