TY - GEN
T1 - Process Dynamics Modeling on Polishing Unit of Artificial Neural Network-Based Produced Water Treatment System
AU - Aisyah, Putri Y.
AU - Soehartanto, Totok
AU - Finazis, Riski F.
AU - Afif, Khoiruddin
AU - Lokeswara, Rajendra
AU - Umamah, Fyrda
N1 - Publisher Copyright:
© 2021 IEEE.
PY - 2021
Y1 - 2021
N2 - Modeling on the polishing unit of the PT. SIPL produced water treatment system was carried out using an Artificial Neural Network (ANN). ANN is used because there is no mathematical model capable of modeling biological processes in polishing units. The ANN used is a feedforward backpropagation structure, with the Levenberg-Marquardt algorithm. The ANN model has 4 input nodes at the input layer, temperature, pH, COD, TSS and 1 output node at the output layer in the form of COD effluent. The method to determine the best hidden node does not yet exist, so 20 variations of the hidden node are carried out to find the best validation RMSE value. The simulation results show that the best RMSE value is in the ANN with a 4-13-1 structure (Input layer-hidden layer-output layer). The resulting RMSE value is 14.7919, which is better than the reference ANN model which is 59.48. The ANN model used is also able to produce COD values below the maximum level of wastewater quality standards. Thus, the dynamics of the polishing unit process can be shown by the ANN model 4-13-1 with the influencing parameters, temperature, pH, COD, and TSS.
AB - Modeling on the polishing unit of the PT. SIPL produced water treatment system was carried out using an Artificial Neural Network (ANN). ANN is used because there is no mathematical model capable of modeling biological processes in polishing units. The ANN used is a feedforward backpropagation structure, with the Levenberg-Marquardt algorithm. The ANN model has 4 input nodes at the input layer, temperature, pH, COD, TSS and 1 output node at the output layer in the form of COD effluent. The method to determine the best hidden node does not yet exist, so 20 variations of the hidden node are carried out to find the best validation RMSE value. The simulation results show that the best RMSE value is in the ANN with a 4-13-1 structure (Input layer-hidden layer-output layer). The resulting RMSE value is 14.7919, which is better than the reference ANN model which is 59.48. The ANN model used is also able to produce COD values below the maximum level of wastewater quality standards. Thus, the dynamics of the polishing unit process can be shown by the ANN model 4-13-1 with the influencing parameters, temperature, pH, COD, and TSS.
KW - artificial neural network (ANN)
KW - polishing unit
KW - produced water
UR - http://www.scopus.com/inward/record.url?scp=85135040650&partnerID=8YFLogxK
U2 - 10.1109/ICAMIMIA54022.2021.9807734
DO - 10.1109/ICAMIMIA54022.2021.9807734
M3 - Conference contribution
AN - SCOPUS:85135040650
T3 - 2021 International Conference on Advanced Mechatronics, Intelligent Manufacture and Industrial Automation, ICAMIMIA 2021 - Proceeding
SP - 23
EP - 27
BT - 2021 International Conference on Advanced Mechatronics, Intelligent Manufacture and Industrial Automation, ICAMIMIA 2021 - Proceeding
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2021 International Conference on Advanced Mechatronics, Intelligent Manufacture and Industrial Automation, ICAMIMIA 2021
Y2 - 8 December 2021 through 9 December 2021
ER -