ИИ и методы Тагути помогли оптимизировать твердость никель-железо-фосфорных покрытий

News Coatings Technologies Ni–Fe–P coating hardness optimised via AI and Taguchi methods Researchers combine Taguchi experimental design with regression, artificial neural networks and support vector machines to predict and optimise the hardness of electroless Ni–Fe–P coatings on copper. Metaheuristic optimisation and post-deposition annealing further increase the microhardness. Electroless Ni–Fe–P coatings are of interest for surface engineering applications where corrosion resistance, hardness and specific functional properties are required. In a recent study, researchers deposited such coatings on copper substrates following a Taguchi L27 experimental array and used the resulting dataset to develop predictive models for microhardness. Three modelling approaches were compared: linear regression, artificial neural networks (ANN) and support vector machines (SVM). Based on the coefficient of determination (R²), the ANN model was identified as the most reliable predictor of coating hardness. Analysis of variance (ANOVA) confirmed that the NiSO₄ concentration and its interaction with FeSO₄ are the most significant parameters governing microhardness. To identify optimal deposition conditions, the authors combined the Taguchi method with two metaheuristic algorithms, namely a Genetic Algorithm and a Bonobo Optimizer. Microstructural analysis and post-annealing behaviour The optimal coating was characterised by scanning electron microscopy (SEM), energy-dispersive X-ray analysis (EDX) and X-ray diffraction (XRD) to explain the mechanisms behind the improved hardness. The as-deposited Ni–Fe–P coating produced under optimal conditions exhibited a moderate phosphorus content, leading to the formation of Ni and FeNi₃ phases. In parallel, the coating displayed a characteristic cauliflower-like microstructure that further contributed to increased hardness. Additional gains were achieved after annealing the optimal samples at 450 °C, which promoted the formation of Ni₃P precipitates and further increased hardness values. The study illustrates how the combination of statistical design of experiments, machine learning models and metaheuristic optimisation can be used to systematically improve the properties of functional coatings, offering a transferable workflow for the development of other electroless coating systems. Source: Mandal, R. et al., Statistical and Machine Learning Approaches for Predicting Hardness of Electroless Ni–Fe–P Coatings with Taguchi and Metaheuristic Optimizations. Journal of Bio- and Tribo-Corrosion 12, 150 (2026).