On the probability distribution and reconciliation of process plant data

In all methods of plant data reconciliation, measurement errors have been assumed to be random variables with normal distribution. However, the propagation of measurement errors through different devices is a nonlinear process that distorts the probability distribution of the original signal. The reconciliation problem mathematical formulation is here redefined to take into account these distortions. Comparison with state of the art techniques shows that there are sizable differences of practical importance.