Forecast of stop line's duration in an Automotive Industry

Abstract

Advanced Manufacturing Processes (P.M.A.) require strict production and quality control of the final manufactured product [1]. In this scenario, productive losses occur, which leads to an increase in production costs. Therefore, a forecast of the profile and type of line stop is proposed, given the entry of a line stop event that occurs in real-time. Using the proposed data mining between the years 2018 to 2020 of the stop history, in the light of the CRISP-DM methodology, accuracy parameters were calculated in the measurement of the line stop time with the Random Forest approach [4 ], [16]. In a scenario of imbalance of the classes in the database, 51% accuracy was found, while with the balance of the same classes and repetition of tests for different values, 38% accuracy was achieved. Finally, the appearance of this profile of loss and time to be foreseen will assist in the decision making of the business manager when, for example, events of production stoppages for equipment repair will be necessary to happen, to prevent major line stoppages.