Based on bounded intensity process, an aperiodic imperfect preventive maintenance policy for numerical control machine tools with random maintenance quality under reliability limit was proposed. A hybrid model, which considers the effect of maintenance to virtual age and the fact that failure intensity increases as operational time increases, was presented. The treating average cost as an objective function, maintenance intervals and the number of optimum maintenance were given iteratively when increasing value of failure intensity and maintenance quality were both random variables with uniform distribution. Finally, a discussion for a real numerical example was presented to show the dependence of optimal results on different cost parameters.