Unsupervised Fault Detection

Unsupervised fault detection aims to identify anomalies in complex systems without relying on labeled data, a crucial task for predictive maintenance and efficient operations. Current research focuses on developing robust algorithms, including those based on generative pre-training, contrastive learning, and graph convolutional networks, often incorporating techniques like moving window approaches and data augmentation to improve accuracy and handle diverse data distributions. These advancements are significantly impacting various fields, enabling more efficient monitoring of industrial processes (e.g., semiconductor manufacturing, chemical processes) and improving the reliability of critical infrastructure by facilitating early fault detection and prevention.

Papers

August 1, 2024

Predictive maintenance solution for industrial systems -- an unsupervised approach based on log periodic power law
Bogdan Łobodziński
Complex System UNsupervised Approach Predictive Maintenance Renormalization Group Industrial System Power Law Log Instruction Unsupervised Fault Detection

July 8, 2024

Unsupervised Fault Detection using SAM with a Moving Window Approach
Ahmed Maged, Herman Shen
Segment Anything Model Fault Detection Unsupervised Method AU Detection Unsupervised Fault Detection Window Approach

May 25, 2024

USD: Unsupervised Soft Contrastive Learning for Fault Detection in Multivariate Time Series
Hong Liu, Xiuxiu Qiu, Yiming Shi, Zelin Zang
Data Augmentation Multivariate Time Series Fault Detection Contrastive Fine Tuning Soft Contrastive Unsupervised Fault Detection

September 20, 2023

Generative Pre-Training of Time-Series Data for Unsupervised Fault Detection in Semiconductor Manufacturing
Sewoong Lee, JinKyou Choi, Min Su Kim
Time Series Time Series Data Generative Pre Trained Transformer Generative Pre Training Semiconductor Manufacturing Unsupervised Fault Detection

August 17, 2022

SensorSCAN: Self-Supervised Learning and Deep Clustering for Fault Diagnosis in Chemical Processes
Maksim Golyadkin, Vitaliy Pozdnyakov, Leonid Zhukov, Ilya Makarov
Self Supervised Learning Predictive Model Sensor Data Deep Clustering High Quality Sensor Industrial Monitoring Unsupervised Fault Detection

December 11, 2021

Spatial Graph Convolutional Neural Network via Structured Subdomain Adaptation and Domain Adversarial Learning for Bearing Fault Diagnosis
Mohammadreza Ghorvei, Mohammadreza Kavianpour, Mohammad TH Beheshti, Amin Ramezani
Domain Adaptation Unsupervised Domain Adaptation Temporal Graph Domain Adversarial Universal Domain Adaptation Graph Domain Adaptation Unsupervised Fault Detection

Unsupervised Fault Detection

Papers

Predictive maintenance solution for industrial systems -- an unsupervised approach based on log periodic power law

Unsupervised Fault Detection using SAM with a Moving Window Approach

USD: Unsupervised Soft Contrastive Learning for Fault Detection in Multivariate Time Series

Generative Pre-Training of Time-Series Data for Unsupervised Fault Detection in Semiconductor Manufacturing

SensorSCAN: Self-Supervised Learning and Deep Clustering for Fault Diagnosis in Chemical Processes

Spatial Graph Convolutional Neural Network via Structured Subdomain Adaptation and Domain Adversarial Learning for Bearing Fault Diagnosis