Sensor Validation Algorithms for Predictive Emission Monitoring Systems

This paper evaluates two sensor validation techniques—Feed Forward Neural Networks and Locally Weighted Regression—as complements to Predictive Emission Monitoring Systems (PEMS) for environmental monitoring. Using field data from a fluid catalytic cracking unit and ABB's IMP software, the study finds that Locally Weighted Regression offers superior performance, economic, and operational benefits for detecting sensor faults and maintaining emission prediction accuracy.