The reliability of the data which characterize the behavior of a plant is critical to the effective monitoring and improvement of plant performance. It is thus essential that gross errors in these data, which can arise from measurement problems or inadequate mathematical models, are detected and eliminated before the performance of the plant is evaluated. Procedures for the detection of gross errors based on back propagation neural nets have recently been shown to be superior to those based on conventional statistical tests, especially where data processing is dependent on highly nonlinear models. The global detection of gross errors in process systems appears to be a relatively simple problem that can be accommodated with equal efficiency by back propagation, probabilistic, and learning vector quantization neural nets. The location of errors based on the constraint residuals of process systems, on the other hand, poses a more formidable problem that is not handled well by standard back propagation nets. For these types of problems other systems, such as learning vector quantization neural nets, that are significantly more efficient than back propagation neural nets are recommended.