Регресійний аналіз накопичення механічних домішок у прокатних емульсіях та алгоритм прогнозування рівня їх забрудненості

• Datum: 2026
• Hauptverfasser: Кухар, В.В., Спічак, О.Ю.
• Format: Artikel
• Sprache: Ukrainisch
• Veröffentlicht: Physico-technological Institute of Metals and Alloys 2026
• Schlagworte: прокатна емульсія; механічні домішки; холодна прокатка; регресійна модель; число омилення; система очищення емульсії
• Online Zugang: https://www.metalsandcasting.com/index.php/mcu/article/view/320
• Назва журналу: Metal and Casting of Ukraine

Introduction. The quality and contamination level of rolling emulsions are critical for process stability in cold rolling, surface cleanliness of steel strip and the service life of lubricant-coolant systems. However, the quantitative laws of mechanical impurity accumulation in industrial emulsions are still insufficiently described, which limits the development of engineering tools for predicting emulsion life. Methods. Laboratory and pilot-industrial tests were carried out for several commercial emulsions (Azmol Delta mark 1, Lubro DL ZPS, Quakerol Zap 4.0/5.0, Rollub 988-AR, Trenoil S740, Optimal Pro, etc.) used on a Tandem mill and a 1680 reversing mill at Zaporizhstal PJSC. Multi-factor regression analysis was applied to relate the content of mechanical impurities to emulsion operating time and concentration. In addition, the influence of saponification number on the regression coefficients was studied using standard Microsoft Excel tools and generative artificial intelligence as an auxiliary instrument for selecting approximation models. Results. It is shown that, for most emulsions, contamination can be described by a linear model with coefficients of determination R2 ≈ 0.58–0.91; the average growth rate of mechanical impurities is (1.2–3.0)·10-4 %/h. The concentration effect is either compensating or amplifying, depending on the emulsion brand, while the mill type and cleaning system strongly affect contamination kinetics. Generalised dependences of regression parameters on saponification number were obtained, and no universal optimum N for minimising contamination was found. Discussion. Based on the derived models, an engineering algorithm for monitoring and predicting the state of rolling emulsions was proposed. It enables the calculation of the allowable operating time before reaching a limit contamination level, optimisation of emulsion concentration and justification of bleed-and-feed intervals for different mill types. The proposed ap- proach provides a basis for adaptive monitoring of lubricant-coolant systems, improving the efficiency of cleaning units and reducing emulsion consumption without loss of cold-rolled product quality.