The safety of raw milk largely depends on using a clean milking system during the milk production. The milking system cleaning process widely used on dairy farms is a highly automated process called cleaning-in-place (CIP), which comprises of four cycles: i) warm water rinse; ii) alkaline wash; iii) acid wash; and iv) sanitizing rinse before the next milking event. Electrolyzed oxidizing (EO) water is an emerging technology, which consists of acidic and alkaline solutions by the electrodialysis of dilute sodium chloride solution. Previous studies in our lab showed that EO water can be an alternative for milking system CIP. Despite the progress made to enhance the CIP performance and evaluate alternative CIPs, the mechanisms behind the cleaning processes were still largely unclear. Therefore, this study was undertaken to evaluate the deposit removal rate during the EO water CIP process using a stainless steel surface evaluation simulator. Deposit removal data from the simulator formed the basis for developing mathematical models to describe the deposit removal process during the CIP process with EO water. Stainless steel straight pipe specimens were placed at the end of undisturbed entrance length along the simulator pipeline. The mass of milk deposits on the inner surfaces of the specimens were measured using a high precision balance after the initial soiling, and then after certain time durations within the warm water rinse, alkaline wash, and acid wash cycles. A unified first order deposit removal rate model dependent on nth power of remaining deposit mass was used for all three cycles. ATP bioluminescence method was also used as a validation approach at the end of each CIP cycle. Experimental results showed that the milk deposit on the inner surfaces of the specimens was removed rapidly by the warm water rinse within 10 s of rinse time. For the alkaline and acid wash cycles, the co-existence of a fast deposit removal at the beginning of the wash cycle and a slow deposit removal throughout the entire wash cycle was inferred. The proposed models matched the experimental data with small root mean square errors (0.23 mg/[(mg) (m(2))] and 0.07 mg/[(mg) (m(2))] for the upstream and downstream locations, respectively) and satisfactory percent error differences (3.67% and 0.93% for the upstream and downstream locations, respectively). Based on the experimental data and the proposed models, the time duration of the CIP process was shortened by 55% (10 s warm water rinse, 3 min alkaline wash and 6 min acid wash) and validated, which yielded an average deposit of 0.28 mg/[(mg) (m(2))] at the end of the CIP as compared with that of 0.29 mg/[(mg) (m(2))] at the end of the original CIP, to achieve a satisfactory CIP performance for the simulator. (C) 2015 Elsevier Ltd. All rights reserved.