From the available literature, several possible solutions for feedwater flow measurement have been identified. These include ultrasonic flow meters (e.g., the LEFM by Caldon Inc., commonly used in nuclear power plants worldwide, or the Krohne UFM 500, for non-nuclear applications and successfully used in the oil industry), and non-ultrasonic solutions such as cross correlation measuring and “virtual” flow meters based on artificial neural networks.
Despite the ultrasonic solutions being the dominating ones in the nuclear industry, the current project aims at investigating the possible added value and/or the technological limits of using a combination of artificial neural networks and other computational intelligence techniques with cross-correlation and noise analysis for a better estimate of such a critical process performance parameter, as shown in Fig. 1. Enhancement of feedwater flow measurement by means of computational intelligence techniques is expected in at least one of two possible scenarios: either by providing a higher accuracy than existing methods, or by providing an acceptable level of accuracy but at a much reduced cost. Either direction is expected to be beneficial to the market of the feedwater flow measurement.