Mass Flow Measurement of Pneumatically Conveyed Solids Through Multi-Modal Sensing and Data-Driven Modelling

Bulk material is pneumatically conveyed in many industrial processes such as steel manufacturing, food processing, chemical engineering, cement production, and power generation. On-line continuous measurement of the mass flow rate of solids in pneumatic conveying pipelines is essential to balance the mass and energy and further to control the energy efficiency and raw material consumption. This thesis describes a novel method of mass flow rate measurement of pneumatically conveyed solids using multi-modal sensing incorporating data-driven models. A review of direct and inferential methods and the methods incorporating the data-drivel models is given with the technical requirements and limitations in the field of gas-solid two-phase flow.

The multi-modal sensing system is comprised of an array of ring-shaped electrostatic sensors, four arrays of arc-shaped electrostatic sensors, a differential pressure (DP) transducer, a temperature and humidity sensor and an accelerometer. The ring-shaped and the arc-shaped electrostatic sensors are capable of measuring the cross-sectionally averaged and the localized velocity and mass flow rate of solids, respectively. The DP transducer is capable of measuring the drop in line pressure caused by different air velocities, mass flow rates of solids and pipe orientations. The temperature and humidity sensors are integrated to measure the temperature and RH values inside the pipe under different ambient conditions. The accelerometer is installed to acquire the information related to different orientations of pipe.

The data-driven models, including artificial neural network (ANN), support vector machine (SVM), and convolutional neural network (CNN), are established through training with statistical features extracted from the post-processed data from the sensing system. Statistical features are shortlisted based on their importance by calculating the partial mutual information between the features and the corresponding reference mass flow rate of solids. The SVM model is also applied on the data of each sensor and different combinations of sensors to evaluate the importance of each sensor in terms of relative error in predicted mass flow rate measurements. Experimental work was conducted on a laboratory-scale test rig with different air velocities, pipe orientations and ambient conditions to measure the mass flow rate of solids. The data-driven modelling techniques are also applied for the measurement of coal and biomass ratios in the coal/biomass/air mixture. Experimental results suggest that the measurement system is capable to reduce the measurement errors to within ±12% under all the physical conditions.