Combination of job scheduling and maintenance activity has been widely investigated in the literature. We consider a non-identical parallel machines batch processing (BP) problem with release dates, due dates and variable maintenance activity to minimize total tardiness. An original mixed integer linear programming (MILP) model is formulated to provide an optimal solution. As the problem under investigation is known to be strongly NP-hard, two meta-heuristic approaches based on Simulated Annealing (SA) and Variable Neighborhood Search (VNS) are developed. A constructive heuristic method (H) is proposed to generate initial feasible solutions for the SA and VNS. In order to evaluate the results of the proposed solution approaches, a set of instances were randomly generated. Moreover, we compare the performance of our proposed approaches against four meta-heuristic algorithms adopted from the literature. The obtained results indicate that the proposed solution methods have a competitive behaviour and they outperform the other meta-heuristics in most instances. Although in all cases, H + SA is the most performing algorithm compared to the others.