Optimal Transport Image Representation and Deep Covariance Alignment (CORAL) for Control Valve Stiction Detection

2026-07-27 · Source: cs.CV updates on arXiv.org · Field: Manufacturing & Industrial — Automation & Robotics, Smart Manufacturing & Industry 4.0, Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A novel stiction detection methodology, combining optimal transport (OT) imaging with Deep Correlation Alignment (Deep CORAL), addresses domain shift in industrial control loops. Control valve stiction, a cause of poor control-loop performance, is typically detected by data-driven methods that struggle to generalize from simulated to real-world data. This approach converts closed-loop signals, specifically controller output and process variable, into two-dimensional OT images to capture dynamic behavior. A convolutional neural network (CNN) encoder learns domain-invariant representations by optimizing a combined objective: cross-entropy loss on labeled simulation data and Deep CORAL loss between simulation and unlabeled industrial data. Evaluated on an independent test set of 20 industrial benchmark loops, the methodology successfully diagnosed 18 loops, achieving 100% recall across 13 stiction cases, 90.00% accuracy, and a 92.86% F1-score. This significantly mitigates domain shift, offering robust stiction detection.

Key takeaway

For Machine Learning Engineers developing fault detection systems for industrial control loops, this methodology offers a robust solution to domain shift challenges. You should consider integrating Optimal Transport imaging with Deep CORAL into your models to improve generalization from simulated to real-world data. This approach achieved 90.00% accuracy and 100% recall on industrial stiction benchmarks, suggesting it can significantly enhance the reliability of your predictive maintenance efforts.

Key insights

Optimal Transport imaging combined with Deep CORAL effectively mitigates domain shift for robust control valve stiction detection in industrial settings.

Principles

• Domain shift hinders real-world model generalization.
• Image representations can capture dynamic system behavior.
• Covariance alignment improves cross-domain feature learning.

Method

Convert closed-loop signals into 2D Optimal Transport images. Train a CNN encoder with combined cross-entropy and Deep CORAL loss to learn domain-invariant features for stiction classification.

In practice

• Apply OT imaging to time-series data.
• Use Deep CORAL for domain adaptation.
• Integrate CNNs for feature extraction.

Topics

• Control Valve Stiction
• Optimal Transport Imaging
• Deep CORAL
• Domain Adaptation
• Convolutional Neural Networks
• Industrial Automation
• Fault Detection

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Computer Vision Engineer

Related on AIssential

• Optimal Transport Image Representation and Deep Covariance Alignment (CORAL) for Control Valve Stiction Detection — Combining optimal transport imaging with Deep CORAL effectively mitigates domain shift for robust control valve stiction detection in industrial settings.
• YOTOnet: Zero-Shot Cross-Domain Fault Diagnosis via Domain-Conditioned Mixture of Experts — YOTOnet enables zero-shot cross-domain fault diagnosis in mechanical equipment using a domain-conditioned mixture of experts.
• Leveraging Deep Learning for Object and Position Recognition of Load Carriers for Autonomous Logistics Vehicles — Deep learning with RGBD data and geometric knowledge enables accurate autonomous load carrier pose estimation for logistics.
• Why Domain Matters: A Preliminary Study of Domain Effects in Underwater Object Detection — A new framework defines underwater domains using measurable image, scene, and acquisition properties for robust object detection evaluation.
• Geometric Dictionary Learning of Dynamical Systems with Optimal Transport — DOODL learns shared spectral manifolds of dynamical systems for robust, data-efficient operator estimation and system identification.
• Pulling The REINS: Training-Free Safety Alignment of Video Diffusion Models via Representation Steering — REINS steers video diffusion models towards safe generation by manipulating linearly encoded safety information in hidden states at inference time.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.