Two mathematical models, which describe the dependence of critical pitting temperatures of 06ХН28МДТ alloy (analogue of AISI904L steel) in model recycled water with pH 4–8 and chloride concentration from 350 up to 600 mg/l on chemical composition and structure, are developed. They are based on multivariate regressions with pairwise combinations of features and a three-layer neural network of direct signal. Applying the developed mathematical models, it is found that the critical pitting temperatures of the 06ХН28МДТ alloy increase with an increase in the pH of model recycled water, the content of Cr, Mo, Cu, the volume of titanium nitrides within it, and a decrease in the medium austenite-grain diameter, the content of nickel within the standard, and chlorides in the media. At the same time, the analysis of the developed mathematical model, which is based on multivariate regressions with paired combinations of features (alloy and media parameters), reveals that the content of Cr and Ni in the alloy in combination with the medium austenite-grain diameter most significantly affect its pitting resistance in model recycled waters, and an effect of the Cr content in combination with pH and chloride concentration in the media is somewhat lower, but much higher than an effect of the Mo and Cu content, and the volume of titanium nitrides in combination with the media parameters. The developed mathematical model, which is based on a three-layer neural network of direct signal propagation, is recommended for predicting the pitting resistance of heat exchangers made of 06ХН28МДТ alloy or AISI904L steel when operating in recycled water. In addition, the developed mathematical model, based on multivariate regressions with paired combinations of features (alloy and media parameters), is recommended for selecting the optimal melts of this alloy or steel, which are most resistant to pitting in recycled water.