Intelligent fault diagnosis for chemical process under imbalanced data: A non-local frequency-domain sparse hash autoencoder approach

Key Points

• Fault diagnosis improved accuracy in detecting anomalies within imbalanced data conditions—highlighting effectiveness.
• The approach utilizes a non-local frequency-domain sparse hash autoencoder to enhance diagnosis capabilities.
• Algorithm performance was assessed through comparisons against traditional methods in diverse chemical process scenarios.
• Findings may enable more reliable fault detection in industrial applications, though real-world validation is still needed.