Two-Phase Flow Measurement Model Of Coriolis Mass Flowmeter Based On Online Active Learning

Coriolis Mass Flowmeter(CMF) has high accuracy as measuring single-phase flow, so it is widely used. The accuracy of CMF is low under two-phase flow conditions, and correction algorithm is required. At present, many models for two-phase flow measurement are relatively complex and the amount of complex calculation is too large, so they cannot be deployed on embedded platforms. This paper proposes an online correction algorithm for metering two-phase flow based on a ZYNQ platform. By online learning and active learning, a high-accuracy neural network model can be obtained in a short time through incremental training. The iRDM algorithm is used to optimize the data set to ensure that the model can be generalized to all data while reducing the training time. The original data is pre-clustered by the K-means algorithm, and the neural network is divided into regions according to the clustering results. The task allocation method is used to improve the network prediction ability and perform real-time correction of the two-phase flow measurement of the CMF. Finally, through experimental verification, the real-time correction method proposed in this paper can make the accuracy of the CMF reach -3.1%~3.29% under the conditions of 210kg/h~890kg/h flow rate and 0~40% gas volume fraction.