Hydraulic systems are widely employed across diverse industrial production processes. Nevertheless, complexity of their system structure presents challenges in developing a maintenance strategy. This paper develops a maintenance strategy for hydraulic systems by proposing an integrated approach that takes into account epistemic uncertainty and multi-source information. Initially, a hydraulic system is modeled using a fault tree, which is subsequently converted into a generalized stochastic Petri net model. A Monte Carlo simulation algorithm is proposed to deal with the epistemic uncertainty that arises from the interval-value failure rates of basic events within its complex structure. As a result, importance measures are calculated for each component. Next, experts are invited to evaluate maintenance cost of components, and their evaluation results are aggregated. Moreover, importance measures and maintenance cost are used to construct an original decision table, and an improved combinative distance-based assessment method is developed to obtain the maintenance strategy for the system. Finally, a case study is conducted on a hydraulic system of a tipping truck with side-pressing mechanism to demonstrate the generality of the proposed methodology.