A consistent set of data is needed in any process for the purposes of control, cost accounting, hazard reduction, and so on. What we call here by the term rectification refers to the adjustment of process measurements to eliminate noise and/or random gross errors. Because artifical neural networks (ANN) are networks of basis functions, they can serve as good nonparametric models of processes. We describe using ANN to rectify data in dynamic processes. To control the rectification and evaluate the results we use simulated rather than actual data. By showing that the rectified data give unbiased estimates of the true process variables (which are not known for plant data), and that the estimated variances of the variables are reduced by rectification, we hope to build up trust that the procedure of using ANN is valid. Two examples indicate what can be accomplished.