Programmable networks brought by Software Defined Networks (SDN) are perceived by operators as cornerstone to reduce the time to deploy new services, to augment the flexibility and to adapt network resources to customer needs at runtime. However, despite the vulnerabilities identified due to the centralization of the intelligence on SDN, its research is more centered on forwarding traffic and reconfiguration issues, not considering to a great extent the fault management aspects of the control plane. The aim of this paper is to provide SDN with fault management capabilities by using autonomic principles like self-healing mechanisms. We propose a generic self-healing approach that relies on a Bayesian Networks for the diagnosis block and it is applied to a centralized SDN infrastructure to demonstrate its functioning in the presence of faults.