No abstract

Pressure fluctuations and X-ray computed tomography (CT) measurements were utilized to characterize the flow behavior of gas-solid fluidized beds using polyethylene particles in three Plexiglas columns with diameters of 10cm, 20cm, and 30cm. Air was used as the gas phase. Gas-solids flow dynamic under ambient conditions was characterized from statistical analysis of pressure fluctuation data and CT images. The time-averaged voidage distribution, bubble phase area fraction, bubble diameter and bubble number distribution varying with the bed heights were extracted from all the three columns. Bed scales had significant effect on the hydrodynamics. Scale up effects on the gas-solids two-phase flow behavior were discussed.

Hydrodynamics of a polyethylene fluidized bed were studied at different operating pressures (191-2908 kPa) and constant temperature of 30 • C. Minimum fluidization velocity (U mf ) decreased with increasing of operating pressures. Measured U mf agree well with calculated U mf as derived from the standard deviation of pressure fluctuations using Puncochar's Method (1985). As expected, with the increase of superficial velocity, amplitude and standard deviation of pressure fluctuation series also increase, which is indication of more vigorous bubbling behavior inside the bed. Power spectral density of pressure fluctuation series indicates similar frequency distribution of the bubbling behavior at different superficial velocities. The dominant frequency from the power spectral was found to be around 1 Hz. Bubble diameter and bubble velocity were estimated from X-ray fluoroscopy images. Bubble diameter and bubble velocity increase with increasing bed height and superficial gas velocity due to bubble coalescence and more gas flowing upwards. At the same fluidization number, the average bubble size slightly decreases with increasing pressure at all bed heights because there is less gas flowing and the bubbles coalesce less rapidly. The bubble velocity is observed to have a small decrease from 191 kPa up to 2200 kPa, and then a substantial increase due to the fact that at higher pressure, the solid circulation increases at the bed.

Flow development and flow dynamics were systematically investigated using local solids concentration measurements in a pair consisting of a downer (0.1 m I.D., 9.3 m high) and a riser of the same diameter (0.1 m I.D., 15.1 m high). Both statistical and chaos analysis were employed. Values for the Kolmogorov entropy (K), correlation dimension (D), and Hurst exponent (H) were estimated from time series of solids concentration measurements. Axial distributions of chaos parameters were more complex in the downer than those in the riser, especially in the entrance section. Flow in the downer was more uniform with a flatter core in all the radial profiles of chaos parameters. The radial profiles of K varied significantly with increasing axial levels due to different clustering behavior in the wall region of the downer. In both the riser and the downer, anti-persistent flow in the core region and persistent flow behavior near the wall were identified from the profiles of H. Different flow behavior in the region close to the wall in the downer and riser was characterized from the combination of the three chaos parameters. Relationships between chaos parameters and local time-averaged solids holdup in the core and wall regions of the developed sections in both the downer and riser were also analyzed.

Fluid Catalytic Cracking (FCC), a key unit for secondary processing of heavy oil, is one of the main pollutant emissions of NOx in refineries which can be harmful for the human health. Owing to its complex behaviour in reaction, product separation, and regeneration, it is difficult to accurately predict NOx emission during FCC process. In this paper, a novel deep learning architecture formed by integrating Convolutional Neural Network (CNN) and Long Short-Term Memory Network (LSTM) for nitrogen oxide emission prediction is proposed and validated. CNN is used to extract features among multidimensional data. LSTM is employed to identify the relationships between different time steps. The data from the Distributed Control System (DCS) in one refinery was used to evaluate the performance of the proposed architecture. The results indicate the effectiveness of CNN-LSTM in handling multidimensional time series datasets with the RMSE of 23.7098, and the R2 of 0.8237. Compared with previous methods (CNN and LSTM), CNN-LSTM overcomes the limitation of high-quality feature dependence and handles large amounts of high-dimensional data with better efficiency and accuracy. The proposed CNN-LSTM scheme would be a beneficial contribution to the accurate and stable prediction of irregular trends for NOx emission from refining industry, providing more reliable information for NOx risk assessment and management.