Risk management in a construction project is a key but extremely complex issue. To answer it, we developed Stochastic Multi-agent simulAtion for Construction projeCt (SMACC), a simulation tool for construction project that considers risks. SMACC proposes, using an agent-based model with a stochastic approach, to assess risk consequences for each project stakeholder. This dynamic model enables to build indicators, which can be used as criteria in a decision support process. A former version of SMACC had some flaws such as a too quantitative approach, a poor modelling of resources or a high rigidity of the process description. Furthermore, formally and conceptually, the SMACC model had a weak structuration of knowledge. An improved version based on an ontology has been developed, called SMACC2. The paper presents this new version as well as a real case of application: the building of the crossroad of Bab El Karmadine at Tlemcen.