Due to unpredictable mobility patterns, nodes in an ad hoc wireless network can get separated into partitions and reconnected a number of times. Whenever a larger partition divides into two smaller ones, there is a critical link that fails and thus divides the partition into two. In this paper, we propose an approach that utilizes network survivability concepts to detect the critical links in an ad hoc wireless network. Once the critical link has been detected, we propose two ways to delay or avoid their failure: change the trajectory of one or both nodes forming the critical link; and bring in another node to reinforce the link. It is assumed that the nodes know their location and periodically update their neighbors with their current locations. We have implemented these techniques in the network simulator ns-2 [4]. The simulation experiments show that the approach can significantly improve the network performance and reduce the degree of partitioning.