In this paper, an algorithm for sensor placement for distributed fault diagnosis is proposed. The main objective of this algorithm is to place the sensors in a system in such a manner that the partition of a system into various subsystems becomes easier facilitating the implementation of a distributed fault diagnosis system. This algorithm also reduces or minimized the number of sensors to be used or install thus reducing overall cost. Binary integer linear programming is used for optimization in this algorithm. A four water tank system has been used to demonstrate and validate the proposed algorithm.