The main purpose of this work is the study of the application of data reconciliation to measurements obtained from various units of an ammonia Brazilian plant (FAFEN/BA). The whole process includes 55 steady state mass balance equations, 51 measured and 25 unmeasured variables. We developed a data reconciliation program using FORTRAN77, where restrictions are linearized so as to use matrix projection method (Crowe et al., 1983), on an iterative approach. Gross errors are identified and eliminated with the aid of an adaptation of the MIMT (Measurement Iterative Modified Method) algorithm. In this paper we simulate the predicted variance reduction for the measured variables and the estimated standard deviation for the unmeasured ones, and study the placement of new instruments in order to improve the estimation accuracy.