Interaction to chloride ions is only one of many challenging scenarios that reinforced concrete buildings may face during their lifespan. This exposure may cause concrete structures, particularly those near the seaside, to deteriorate and become less long-lasting. The chloride diffusion coefficient ( \(CD\) ) of nonsteady mode evident concrete may be accurately predicted using AI models that use empirical field data over an extended period of time. By isolating the most important factors, this technique has the potential to improve the evaluation of a concrete structure’s durability. This work shows how to estimate the \(CD\) of concrete under different exposure situations using an adaptive neuro-fuzzy inference system ( \(ANFIS\) ). The Bald Eagle Search ( \(BES\) ), Black Widow Optimization ( \(BWO\) ), and artificial rabbit optimization ( \(ARO\) ) techniques were used to enhance the forecast models, which were trained on a dataset involving 216 data points. Findings indicate that the \(ANFAR\) , \(ANFBW\) , and \(ANFBE\) models have promising futures as \(DC\) predictors for concrete under different exposure conditions, all while maintaining suitable R2 rates. The findings imply that ANFAR, ANFBW, and ANFBE may accurately anticipate concrete CD under diverse exposure scenarios. ANFBE attained R2 values of 0.9959 throughout training and 0.997 throughout testing, larger than other systems in spite of their higher capability during prediction accuracy. When exposure time (ET) and exposure condition (EC) are eliminated from the input set, RRSE, NMSE, and R2 values significantly rise and drop. Importantly, eliminating ET led to a remarkable impact on the target, where R2 reduced from 0.997 to 0.7761 and NMSE increased from 0.001 to 0.1834 in testing.
