Model-based testing overcomes challenges in software testing by generating automated test cases from behavior models, e.g. finite state machine (FSM) and input/output transition system (IOTS). Despite the existence of methods for IOTSs, the problem of selection of test cases is an important and difficult topic. The current methods from IOTSs do not provide the same support offered by the existing theory for FSMs, as complete fault coverage. In this paper, we propose a test generation method for IOTSs based on the W method developed for FSMs. The basic idea is to generate a transition cover set and a characterization set and concatenate them to generate complete test suites for IOTSs in a bounded number of steps. The method generates test suites with complete fault coverage for a given fault domain and is targeted at a class of IOTSs, called mealy IOTS, which accepts inputs only in stable states. Results from a case study show the proposed algorithm can achieve better results than a classical method for IOTSs.