- 148 views
- 143 downloads
Effect of Purge Gas Flow Rate and Oxygen Impurity on Heel build-up and Prediction of Heel build-up Using Machine Learning
- Author / Creator
- Rahmani, Keivan
Heel buildup, i.e., the accumulation of non-desorbed/non-desorbable adsorbates and their by-products on an adsorbent, during cyclic adsorption/regeneration of volatile organic compounds (VOCs) onto activated carbon decreases its adsorption capacity and lifetime. In this this study, the effect of purge gas glow rate and oxygen impurity on heel build-up were investigated and machine learning was used to predict heel build-up.
In the first part, the simultaneous effect of purge gas flow rate and oxygen impurity during successive adsorption/regeneration cycles on adsorption capacity of and heel buildup on activated carbon during cyclic adsorption-desorption of TMB was investigated. Nine thermal desorption scenarios were investigated by varying nitrogen purge gas oxygen impurity level (< 5 ppmv, 10,000 ppmv, and 21%) and flow rate (0.1, 1 and 10 L/min or 1, 10 and 100% of adsorption flowrate) during thermal regeneration. The results show that increasing purge gas flow rate during thermal desorption improves adsorption capacity recovery and mitigates adverse effects of purge gas oxygen impurity. By increasing the purge gas flow rate from 0.1 to 10 SLPM, fifth cycle adsorption capacity increased from 30 to 40 wt% when 10,000 ppmv O2 was used as purge gas and from 10 to 37 wt% when dry air (21% O2) was used. Cumulative heel increased with increasing the purge gas oxygen impurity and decreasing its flow rate. In the least effective regeneration scenario (0.1 L/min N2 with 21% O2), 32.8 wt% cumulative heel was formed on BAC after 5 adsorption-desorption cycles whereas that of the best-case scenario (10 L/min N2 with <5 ppmv O2) was 0.3 wt%. Comparing the pore size distributions of virgin and used BAC indicated that heel is firstly built in narrow micropores (<8.5Å) and then starts to engage mesopores. Thermogravimetric analysis (TGA) of regenerated samples revealed that oxygen impurity leads to formation of high boiling point and/or strongly bound heel species. TGA confirmed that in presence of oxygen, higher purge gas flow rates could reduce the amount of heel but promote chemisorbed heel formation. These results can be used for optimizing the regeneration conditions to boost activated carbon’s long-term performance in cyclic adsorption.
In the second part, two machine learning (ML) algorithms (XGBoost and deep neural network (DNN)) were applied to predict volatile organic compounds (VOCs) cyclic heel buildup on activated carbons (ACs). A dataset consisting of 411 experimental tests of cyclic adsorption/desorption of different VOCs on ACs with distinct properties was used. Our study revealed that cyclic heel buildup can be predicted with acceptable accuracy using both ML algorithms by considering the adsorbent characteristics, adsorbate properties and regeneration conditions. The DNN algorithm showed better performance in prediction of cyclic heel buildup (R2 = 0.94) than XGBoost (R2 = 0.81). To verify the ML algorithms results and gain some insight into heel buildup relation to adsorbent’s nature characteristics, partial dependency plots were generated using adsorbate properties and regeneration conditions, partial dependency plots were generated. The proposed ML-based heel prediction methods can be used to: (i) optimize adsorption/desorption operating conditions to minimize heel buildup on activated carbon in cyclic adsorption processes and (ii) quickly screen various adsorbents for efficient adsorption of a particular family of VOCs.
- Graduation date
- Fall 2021
- Type of Item
- Master of Science
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.