Combinatorial testing is an important software engineering problem. Covering array generation (CAG), a discrete optimization problem, is the most popular field of research in combinatorial testing (CT). The problem has attracted many researchers in the community of software engineering. This paper presents two novel versions of discrete particle-swarm-simulated-annealing-based memetic algorithms named D-PSMA for CAG problem. We elaborate with both particle swarm optimization searching operators and simulated annealing local searching operators synthetically to equilibrate the exploration and exploitation capabilities. The concept of suboptimal solution acceptance rate is introduced for excellent suboptimal solution. After each iteration, multiple test cases are generated that improve the efficiency of test suite generation. Meanwhile, an improved boundary constraint strategy are proposed. In the experiment, we used different test suite generation strategies, boundary constraint strategies and discrete memetic algorithm in the general framework, and found the optimal combination for CAG problem. Extensive experiments demonstrate that D-PSMA has competitiveness when compared to other existing classical algorithms.