Machine Learning-Based Stability Prediction and Analysis\nof Polypropylene Cu-MA Superhydrophobic Coating on the Aluminum Substrate

This study investigates the performance\nof a polypropylene myristic\nacid-modified copper (Cu) superhydrophobic coating under various conditions,\nincluding immersion in NaCl solution, abrasion tests, pH levels, and\ntemperature variations. We utilized specific machine learning models,\nsuch as random forest (RF) and XGBoost, to predict the wettability\nbehavior of materials under varying voltage and temperature conditions.\nGaussian noise data augmentation was employed to improve the model\naccuracy and prevent overfitting. The RF model demonstrated strong\ngeneralization capabilities, with consistently low MSE values and\nhigh <i>R</i><sup>2</sup> values across training and testing\ndata sets. In contrast, the XGBoost model showed a slight performance\ndecline when transitioning from training data to testing data, indicating\npotential generalization limitations. For variables such as immersion\ndays, abrasion cycles, pH levels, and temperatures, multiple polynomial\nregression models were developed to predict the contact angle, showing\na high degree of alignment with experimental data. The predictions\nof contact angle after 122 days of immersion in the NaCl solution\n(3.5 wt %), 350 abrasion cycles (each cycle of 18 cm), and exposure\nto 400 °C indicated sustained hydrophobic properties. The predicted\ncontact angles were 110° for immersion in NaCl, 400 °C,\nand for abrasion cycles it is 99°. These predictions closely\nmatch with the experimental results, which showed contact angles of\n103.7° for immersion, abrasion cycles and 97° for temperature\nexposure. The coating’s robustness is attributed to its micro/nanostructured\nsurface, which traps air and reduces water contact, and the chemical\nstability imparted by polypropylene myristic acid modification, ensuring\nlow surface energy and durability. This study demonstrates the effectiveness\nof higher-degree polynomial models in accurately predicting the performance\nof superhydrophobic coatings, underscoring the potential of advanced\nmodeling techniques in the development of durable, high-performance\ncoatings.