No abstract

Corrosion process in metallic cans is one of the most significant issues, which is of utmost importance in food packaging; meanwhile, it can be studied through both economical and hygienic aspects. Designing a method to determine this process in an acidic food like tomato paste might be a good standard to monitor this phenomenon in food products. Accordingly, in this article, tin‐free steel (TFS) and tinplate (TIN) cans containing tomato paste were incubated at three different temperatures (25, 35, and 45 °C) and three filling percentages (85, 90, and 95%). Corrosion rate was then studied via image processing method within six months, thereupon corrosion modeling was carried out using adaptive‐network‐based fuzzy inference system (ANFIS). Time, temperature and filling percentage were considered as the input parameters, on the other hand the corrosion rate was assumed as the output. Multiple training cycles and various membership functions were used in the form of trial and error to have the models optimized. Having the models been examined, trapezoidal membership functions with 3–3–3 functions for TFS cans and 2–2–2 functions in Tin cans were determined as the best models. High coefficients of determination (R2 = 0.97) between experimental and predicted values by optimum models presented a robust method for monitoring corrosion in metallic cans. Practical applications Food products might be preserved in metallic cans for more than 2 years without any significant changes in organoleptic properties. One disadvantage is that cans are more susceptible to corrosion compared with other packaging materials. The absence of appropriate methods and the high costs of analytical instruments are major problems encountered in evaluating the corrosion process. On the other hand, developing straightforward and modern techniques may be an efficient solution in this area. This study was devoted to the prediction of the corrosion process using two prevalent can plates employed in the food industry. ANFIS modeling was implemented in tomato paste, which exhibits acidic and corrosive behavior. ANFIS modeling may be an onset for further research in this area to engineer and manufacture intelligent sensors by which on‐line monitoring in complex and nonlinear interactions is undertaken.

No abstract