Evaluation of instrument calibration monitoring using artificial neural networks

Traditional approaches to sensor validation involve periodic instrument calibrations. These calibrations are expensive both in labor and process downtime. Many periodic sensor calibration techniques require the process to be shut down, the instrument taken out of service, and the instrument loaded and calibrated. This method can lead to damaged equipment, incorrect calibrations due to adjustments made under nonservice conditions, and loss of product due to unnecessarily shutting down a process. A less invasive technique of determining sensor status using data from nuclear and chemical process systems is described in this paper