During the hot tandem rolling production process, the urgency of rolling tasks requires the manufacturing workflow to respond efficiently to market requirements. Product quality decisively influences a company's competitiveness and customer satisfaction. This study examined production scheduling optimisation and roller-changing decisions in hot rolling systems, considering different production tasks with different quality grades and delivery deadlines under multiple attribute constraints. First, a stochastic wear degradation and reliability prediction model for the roller was developed based on the different material attributes of the rolled pieces and the diversity of the production processes, all of which have varying effects on the roller wear. Second, a combined decision model for production scheduling and roller changes was developed under the dual constraints of quality grade and delivery deadlines, from which the optimal scheduling sequence for the rolled pieces and the reliability threshold for roller changes were determined with the aim of maximizing production profitability. A hybrid optimisation algorithm combining the genetic algorithm with tabu search was employed to enhance search efficiency. Numerical experiments with data sourced from 1,009 rolled pieces produced in an actual steel factory were conducted to validate the effectiveness of the proposed model and method. The results showed that the proposed strategies and models could increase profits by 2.62% under the same conditions and reduce the number of roller changes, ensuring that the reliability of the system before roller changing was no less than 0.97, while still meeting the required quality grades.