TY - JOUR
T1 - On‐line management of siloxane breakthrough in carbon contactors using changepoint detection
AU - Hepburn, Caroline A.
AU - Daneshkhah, Alireza
AU - Simms, Nigel
AU - McAdam, Ewan
PY - 2015
Y1 - 2015
N2 - Siloxane‐induced damage is responsible for a significant proportion of the cost of producing electricity from biogas. The performance of siloxane removal vessels is currently determined by discrete sampling followed by analysis at an external laboratory, making refined process control difficult to achieve. This article introduces the use of on‐line siloxane detection. Siloxane breakthrough curves within this article identified that upon approaching media exhaustion, a concentration of 30 mg/m3 was reached within a short time frame. Consequently, changepoint analysis was evaluated as a method to predict the onset of breakthrough. Breakthrough data were collated from vessels that exhibited both broad‐fronted and sharp‐fronted breakthrough profiles. When applied to broad‐fronted profiles, changepoint analysis provided a 25–45 h early warning of breakthrough. In this case, probabilistic changepoint detection methods facilitated by on‐line analysis provide sufficient process lag time to enable duty/standby switch over before the critical siloxane concentration is reached.
AB - Siloxane‐induced damage is responsible for a significant proportion of the cost of producing electricity from biogas. The performance of siloxane removal vessels is currently determined by discrete sampling followed by analysis at an external laboratory, making refined process control difficult to achieve. This article introduces the use of on‐line siloxane detection. Siloxane breakthrough curves within this article identified that upon approaching media exhaustion, a concentration of 30 mg/m3 was reached within a short time frame. Consequently, changepoint analysis was evaluated as a method to predict the onset of breakthrough. Breakthrough data were collated from vessels that exhibited both broad‐fronted and sharp‐fronted breakthrough profiles. When applied to broad‐fronted profiles, changepoint analysis provided a 25–45 h early warning of breakthrough. In this case, probabilistic changepoint detection methods facilitated by on‐line analysis provide sufficient process lag time to enable duty/standby switch over before the critical siloxane concentration is reached.
KW - Bayesian change point
UR - https://www.scopus.com/pages/publications/84938744627
U2 - 10.1111/wej.12133
DO - 10.1111/wej.12133
M3 - Article
SN - 1747-6593
VL - 29
SP - 383
EP - 390
JO - Water and Environment Journal
JF - Water and Environment Journal
IS - 3
ER -