Abstract
The development of coal-water slurry (CWS), a new type of coal-based chemical product in China, has garnered increasing attention as a potential substitute for petroleum resources. The Coriolis mass flow meter is widely used in industrial measurement due to its low uncertainty and its ability to simultaneously measure fluid density and mass flow rate, with a single-phase measurement error as low as 0.1%. However, significant errors still exist in multiphase flow measurement scenarios. To address this issue, we designed and constructed a CWS liquid-solid two-phase flow measurement platform to investigate the flow measurement errors of CWS in Coriolis mass flow meters under various conditions. A deep learning correction framework was developed to mitigate the significant measurement errors in liquid-solid two-phase flow. Based on the theoretical support provided by repeatability experiments, two correction models were established: (1) An error correction model based on a BP neural network was developed, which provided corrections for the measurement errors of CWS liquid-solid two-phase flow. The first correction results showed that the corrected error of the predictive model was 3.98%, a significant improvement compared to the 5.11% error measured by the X company's meter. (2) Building on this, a second correction model was established through algorithm optimization, successfully reducing the corrected error of the predictive model to 1.01%. Through this study, we aim at providing a new technical approach for Coriolis mass flow meters in the field of liquid-solid two-phase flow measurement, enhancing measurement accuracy, reducing costs, and offering more reliable data support for industrial process control and scientific research.