Optimising the automated vision inspection system of battery sheets
Examensarbete för masterexamen
Please use this identifier to cite or link to this item:
Bibliographical item details
|Type: ||Examensarbete för masterexamen|
|Title: ||Optimising the automated vision inspection system of battery sheets|
|Authors: ||Sivakumar, Guru Prasath|
|Abstract: ||The utilisation of lithium-ion batteries in the field of Electric Vehicles (EVs) has increased
its global demand. Northvolt, a Swedish lithium-ion battery manufacturer
is setting up Europe’s largest battery manufacturing (in terms of battery capacity)
facility at Skellefteå, Sweden.
The electrode manufacturing process stage in the battery production line at Northvolt
Ett will perform the quality operation through a vision inspection system, which
is being developed in-house at Northvolt Labs. This thesis is focused on identifying a
suitable machine learning model through which the anomalies can be detected; also,
to investigate how the big data collected through the vision inspection system can
be further utilised for improving the battery manufacturing process at Northvolt Ett.
In total, five research questions on anomaly detection and big data usage were formulated.
Qualitative study was performed to understand the various stakeholder
requirements, challenges in detecting anomalies, manufacturing process constraints
etc. at Northvolt. Experts from different working backgrounds at Northvolt were
interviewed and over 50 research articles were analysed to answer the research questions.
Deep learning was found to be a suitable method for anomaly detection and various
deep learning techniques such as Convolutional Neural Network (CNN), Generative
Adversarial Network (GAN) can be used to detect anomalies and predict the type
of anomaly on a high-level classification. A framework is proposed for transforming
the data collected through vision inspection into meaningful value for improving the
battery manufacturing process at Northvolt Ett. Overall, it was understood that
the behaviour of the machine learning/deep learning model is greatly influenced by
the richness of the input data.
Keywords: Vision Inspection, Anomaly Detection, Deep Learning, Convolutional
Neural Network, Big Data Analysis.|
|Keywords: ||Vision Inspection, Anomaly Detection, Deep Learning;Convolutional Neural Network, Big Data Analysis|
|Issue Date: ||2021|
|Publisher: ||Chalmers tekniska högskola / Institutionen för industri- och materialvetenskap|
|Collection:||Examensarbeten för masterexamen // Master Theses (IMS)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.