Ablation Performance of the SAU-PGA-CNN-BiLSTM Model for Industrial Fault Diagnosis

93.6% multiclass diagnosis accuracy and a 94.0% F1 score were achieved by the proposed SAU-PGA-CNN-BiLSTM model on the Tennessee Eastman Process benchmark. The dataset, authored by Babar Hayat and last updated in June 2026, contains ablation performance results for a sensor-aware deep learning framework designed for industrial fault detection and diagnosis.

Use Cases

• Benchmarking fault diagnosis models based on the reported 93.6% accuracy and 94.0% F1 score.
• Analyzing model calibration improvements based on the reported reduction in negative log-likelihood, Brier score, and expected calibration error.
• Studying the effect of sensor-aware attention mechanisms on model performance for industrial process data.

Strengths

• Performance metrics are explicitly quantified, including 93.6% accuracy and a 94.0% F1 score.
• Model calibration improvements are documented with specific reductions in negative log-likelihood (0.197 to 0.182) and Brier score (0.101 to 0.095).
• The model was validated on the established Tennessee Eastman Process benchmark.

Limitations

• The dataset is very small at 5.5 KB, suggesting limited scope.
• Row count and column-level documentation are absent; field semantics must be inferred after download.
• The description focuses on model performance, so the specific data content and structure are unclear.

Provenance

Source

figshare

Collection Method

Likely contains ablation study results from the author's research on the SAU-PGA-CNN-BiLSTM model.

Freshness

Last updated 2026-06-02 17:36:45; freshness should be verified.