Regularization-based Continual Learning for Anomaly Detection in Discrete Manufacturing

Maschler, Benjamin; Pham, Thi Thu Huong; Weyrich, Michael

doi:10.1016/j.procir.2021.11.076

Cited by 24 publications

(16 citation statements)

References 17 publications

(33 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• Due to only very low numbers of identical industrial machinery, high standards of data protection and low levels of cooperation between different enterprises, sufficiently large datasets and diverse for successful training are hard to acquire [19]. • Due to increasing demand for frequent reconfigurations [20], changing processes and dynamic environments, quickly outdating datasets once acquired only provide short-term representations of the problem space necessitating continuous data collection and algorithm retraining [21]. Transfer learning offers mitigation to those challenges by enabling algorithms to train not only on datasets characterizing the task at hand, but on related ones, e.g.…”

Section: Transfer Learningmentioning

confidence: 99%

See 1 more Smart Citation

Transfer learning as an enabler of the intelligent digital twin

et al. 2021

Self Cite

View full text Add to dashboard Cite

Section: Transfer Learningmentioning

confidence: 99%

“…Studies show that the quality of this fit greatly influences the resulting algorithm's performance [25]. Furthermore, the factors that support or hinder the adaption of algorithms proven to solve similar use cases are still largely unknown [21], making diligent testing of proposed solutions a key priority.…”

Section: Scenario 1: Design Of Deep Neural Networkmentioning

confidence: 99%

Transfer learning as an enabler of the intelligent digital twin

et al. 2021

Self Cite

View full text Add to dashboard Cite

“…The term 'deep industrial transfer learning' refers to methods utilizing previously acquired knowledge within deep learning techniques to solve tasks from the industrial domain [7,10]. It can be used to facilitate learning across several smaller, less homogenous datasets [11][12][13], thereby mitigating two central problems of conventional deep learning in industry [10]:…”

Section: B Deep Industrial Transfer Learningmentioning

confidence: 99%

“…In [13], EWC, online EWC [20] and synaptic intelligence (SI) [21] are compared regarding their ability to predict the state-of-health of lithium-ion batteries. Despite the overall good results, again, the generalization capabilities showed clear limitations.…”

Section: B Deep Industrial Transfer Learningmentioning

confidence: 99%

“…In [12], EWC, online EWC, SI and learning without forgetting (LwF) [22] are compared regarding their ability to detect anomalies in a discrete manufacturing process. Here, a saturation-like effect is limiting the continued addition of new knowledge for longer task sequences, raising concerns regarding the approaches' long-time functionality.…”

Section: B Deep Industrial Transfer Learningmentioning

confidence: 99%

See 1 more Smart Citation

Towards Deep Industrial Transfer Learning for Anomaly Detection on Time Series Data

Maschler

Knodel

Weyrich

2021

2021 26th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA )

Self Cite

View full text Add to dashboard Cite

Deep learning promises performant anomaly detection on time-variant datasets, but greatly suffers from low availability of suitable training datasets and frequently changing tasks. Deep transfer learning offers mitigation by letting algorithms built upon previous knowledge from different tasks or locations. In this article, a modular deep learning algorithm for anomaly detection on time series datasets is presented that allows for an easy integration of such transfer learning capabilities. It is thoroughly tested on a dataset from a discrete manufacturing process in order to prove its fundamental adequacy towards deep industrial transfer learning -the transfer of knowledge in industrial applications' special environment.

show abstract

Continual prune-and-select: class-incremental learning with specialized subnetworks

et al. 2023

View full text Add to dashboard Cite

We introduce a new continual (or lifelong) learning algorithm called LDA-CP&S that performs segmentation tasks without undergoing catastrophic forgetting. The method is applied to two different surface defect segmentation problems that are learned incrementally, i.e. providing data about one type of defect at a time, while still being capable of predicting every defect that was seen previously. Our method creates a defect-related subnetwork for each defect type via iterative pruning and trains a classifier based on linear discriminant analysis (LDA). At the inference stage, we first predict the defect type with LDA and then predict the surface defects using the selected subnetwork. We compare our method with other continual learning methods showing a significant improvement -mean Intersection over Union better by a factor of two when compared to existing methods on both datasets. Importantly, our approach shows comparable results with joint training when all the training data (all defects) are seen simultaneously * .

show abstract

Regularization-based Continual Learning for Anomaly Detection in Discrete Manufacturing

Cited by 24 publications

References 17 publications

Transfer learning as an enabler of the intelligent digital twin

Transfer learning as an enabler of the intelligent digital twin

Towards Deep Industrial Transfer Learning for Anomaly Detection on Time Series Data

Continual prune-and-select: class-incremental learning with specialized subnetworks

Contact Info

Product

Resources

About