Image-based Novel Fault Detection with Deep Learning Classifiers using Hierarchical Labels

Nurettin Sergin,Jiayu Huang,Tzyy-Shuh Chang,Hao Yan

2024-03-27

Abstract:One important characteristic of modern fault classification systems is the ability to flag the system when faced with previously unseen fault types. This work considers the unknown fault detection capabilities of deep neural network-based fault classifiers. Specifically, we propose a methodology on how, when available, labels regarding the fault taxonomy can be used to increase unknown fault detection performance without sacrificing model performance. To achieve this, we propose to utilize soft label techniques to improve the state-of-the-art deep novel fault detection techniques during the training process and novel hierarchically consistent detection statistics for online novel fault detection. Finally, we demonstrated increased detection performance on novel fault detection in inspection images from the hot steel rolling process, with results well replicated across multiple scenarios and baseline detection methods.

Machine Learning,Artificial Intelligence

What problem does this paper attempt to address?

This paper focuses on the problem of using deep learning for new fault detection in industrial manufacturing systems. Existing fault classification models often assume that the types of faults are a known fixed set, but in reality, new types of faults may be encountered. The paper proposes a method to enhance the deep learning model's capability for detecting unknown faults by utilizing hierarchical label information, without sacrificing the model's performance on known faults classification. Specifically, the paper proposes two key points: 1. **Soft labeling technique during the training process**: Improving the novel fault detection technique of deep learning by injecting hierarchical structure information into the loss function, allowing the model to consider both classification accuracy and hierarchical structure consistency during training. 2. **Hierarchical consistency score for online detection**: Proposing a new statistical method for online detection of new faults, by detecting samples that are inconsistent with the known fault hierarchy, in order to improve overall detection performance. The paper demonstrates the superiority of this method on a defect detection image dataset of hot steel rolling process, and proves that the proposed strategies can improve the performance of deep learning-based fault detection methods. In addition, the authors discuss the importance of incorporating hierarchical information into training and detection statistics for anomaly detection, pointing out that although many fault sets have hierarchical structures, these information are often overlooked in practice.

Image-based Novel Fault Detection with Deep Learning Classifiers using Hierarchical Labels

Fault Detection in Nuclear Power Plants using Deep Leaning based Image Classification with Imaged Time-series Data

A Novel Visual Fault Detection and Classification System for Semiconductor Manufacturing Using Stacked Hybrid Convolutional Neural Networks

Automated Detection of Defects with Low Semantic Information in X-ray Images Based on Deep Learning

A Novel Deep Learning–based Fault Diagnosis Algorithm for Preventing Protection Malfunction

A Novel Nonlinear Mapping Based Compatibility Method for Zero-Shot Classification in Intelligent Fault Diagnosis

Enhanced Fault Diagnosis in Time-series Data by Deep Learning

Novelty Detection and Fault Diagnosis Method for Bearing Faults Based on the Hybrid Deep Autoencoder Network

Deep-Learning-Based Open Set Fault Diagnosis by Extreme Value Theory

Early Fault Detection Approach with Deep Architectures

Fault Classification Method of Information System Based on Cascaded Convolutional Neural Network Model.

Joint learning system based on semi–pseudo–label reliability assessment for weak–fault diagnosis with few labels

Learning From Unlabelled Real Seismic Data: Fault Detection Based On Transfer Learning

A novel deep learning approach for intelligent fault diagnosis applications based on time-frequency images

A Unified Label Noise-Tolerant Framework of Deep Learning-Based Fault Diagnosis via a Bounded Neural Network

Deep Imbalanced Separation Network: A Holistic Fault Detection Framework Considering Class-Imbalance and Partial Label-Unknown

Positive-Unlabeled Learning-Based Hybrid Deep Network for Intelligent Fault Detection

A Novel Deep Learning Network via Multiscale Inner Product With Locally Connected Feature Extraction for Intelligent Fault Detection

A novel multiscale hybrid neural network for intelligent fine-grained fault diagnosis

Deep Convolutional Neural Network Using Transfer Learning for Fault Diagnosis

Detecting and Diagnosing Incipient Building Faults Using Uncertainty Information from Deep Neural Networks