Towards a Modelling Methodology for Fault-Tolerant Multi-Agent System