In order to respond to today’s dynamic needs of customers, customized mass production systems have been more and more developed that, are facing with different challenges. Maintenance planning and scheduling is one of the most important manufacturing components in such systems, due to importance of availability and high investment for this kind of system. In order to consider real machine operation state, recently, predictive maintenance method is proposed. However, in traditional methods, historical failure data is the main source for this planning. In this paper, we propose a methodology for dynamic predictive maintenance for a real case in automotive industries with considering multi-component structural and positive economic dependencies between them. In our methodology, we propose to gather data science with mathematical optimization method. Prediction of Remaining Useful Life (RUL) of machine parts has been made by Artificial Neural Network method with considering sensors data. With this RUL values and other cost values and optimization model parameters, and by solving proposed mathematical model, an optimal schedule is achieved with minimization of maintenance costs. Through a dynamic proposed procedure, when a new data is received, RUL values and model parameters are readjusted and new optimal solution for maintenance planning and scheduling can be achieved. Further, some scenarios are defined for analyzing the dynamicity of the proposed procedure and relating results, conclusion and perspectives of these researched are discussed.