Multi-View Deep Learning-Based Comprehensive Evaluation Model for Coating Protective Performance

Organic protective coatings are extensively employed to mitigate metal corrosion, yet accurate quantitative evaluation of their performance degradation during service still poses a considerable challenge. This study aims to develop a multi-view deep learning framework for five-grade quantitative assessment of organic coating protective performance using routine electrochemical and mechanical parameters. This study aims to develop a multi-view deep learning framework for five-grade quantitative assessment of organic coating protective performance using routine electrochemical and mechanical parameters. Based on abundant laboratory-accelerated corrosion test data, independent single-view sub-models were constructed for three complementary descriptors, including the mid-frequency phase angle (θ_{10 Hz}), open-circuit potential (OCP), and adhesion strength (A_s). Prediction outputs from individual sub-models were fused through correlation-weighted voting, where weighting factors were determined by quantitative parameter-degradation correlations across various coating systems. The proposed framework achieves reliable five-level grading (excellent, good, fair, poor, failure) of coating protective performance. This methodology provides an effective data-driven framework for the predictive assessment of organic coating protective performance in practical engineering applications.