[vc_empty_space][vc_empty_space]
Dosage optimization of polyaluminum chloride by the application of convolutional neural network to the floc images captured in jar tests
Yamamura H.a, Putri E.U.a, Kawakami T.b, Suzuki A.b, Ariesyady H.D.c, Ishii T.a
a Department of Integrated Science and Engineering for Sustainable Society, Faculty of Science and Engineering, Chuo University, Tokyo, 112-8551, Japan
b Department of Information Engineering, Hokkaido University of Science, Sapporo, 006-8585, Japan
c Faculty of Civil and Environmental Engineering, Institut Teknologi Bandung, Bandung, 40132, Indonesia
[vc_row][vc_column][vc_row_inner][vc_column_inner][vc_separator css=”.vc_custom_1624529070653{padding-top: 30px !important;padding-bottom: 30px !important;}”][/vc_column_inner][/vc_row_inner][vc_row_inner layout=”boxed”][vc_column_inner width=”3/4″ css=”.vc_custom_1624695412187{border-right-width: 1px !important;border-right-color: #dddddd !important;border-right-style: solid !important;border-radius: 1px !important;}”][vc_empty_space][megatron_heading title=”Abstract” size=”size-sm” text_align=”text-left”][vc_column_text]© 2019 Elsevier B.V.The optimization of the coagulation of floc remains challenging because it is affected by various factors, including the pH, turbidity, and alkalinity. A jar test is a reliable method to optimize the coagulation conditions; however, it is time-consuming and requires an experienced technician. A convolutional neural network (CNN) can be used to extract the specific image characteristics via learning processes to construct image classification models. In this study, we applied a CNN for predicting the jar test performance. Artificial water adjusted with kaolin and commercial humic acid was initially used. The floc image was recorded using a video camera during the jar test, and models for predicting the turbidity of the supernatants were constructed by inputting the recorded floc images and the level of turbidity as the training dataset. The learning curve denoted that all the models exhibited 100% accuracy during the learning process, clearly indicating that the CNN worked appropriately with respect to the extraction of the characteristics of the floc images and the determination of their settleability. Subsequently, the overfitting level was investigated by inputting the test data into the models. The maximum prediction accuracy of all the models exceeded 96%. The initial 100 s of the model’s learning of the floc images achieved a maximum accuracy of 99.6%, i.e., the images captured during rapid mixing were sufficient to ensure the reliability of the model. The applicability of the model to natural water samples was also investigated using samples obtained during different months. Even though the accuracy for natural water was lower than that for artificial water, 90% accuracy was achieved even when fluctuating natural water was used, demonstrating the applicability of a CNN to extract the characteristics of the floc governing its settleability and to immediately as well as automatically predict the state of coagulation.[/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Author keywords” size=”size-sm” text_align=”text-left”][vc_column_text]Classification models,Coagulation models,Convolutional neural network,Image characteristics,Jar test,Natural water samples,Polyaluminum chloride,Rapid mixing[/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Indexed keywords” size=”size-sm” text_align=”text-left”][vc_column_text]Coagulation modeling,Deep learning,Jar test,Rapid mixing,Turbidity[/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”Funding details” size=”size-sm” text_align=”text-left”][vc_column_text][{‘$’: ‘This work was supported by JSPS Kakenhi Grant Number 17H04941 and Tokyo Ohka Foundation for The Promotion of Science and Technology . We would like to thank the Yoshimi Water Purification Plant and the Maezawa Industries, Inc. for providing us with water samples. The authors declare no competing financial interests. Appendix A’}, {‘$’: ‘This work was supported by JSPS Kakenhi Grant Number 17H04941 and Tokyo Ohka Foundation for The Promotion of Science and Technology. We would like to thank the Yoshimi Water Purification Plant and the Maezawa Industries, Inc. for providing us with water samples. The authors declare no competing financial interests.’}][/vc_column_text][vc_empty_space][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][vc_empty_space][megatron_heading title=”DOI” size=”size-sm” text_align=”text-left”][vc_column_text]https://doi.org/10.1016/j.seppur.2019.116467[/vc_column_text][/vc_column_inner][vc_column_inner width=”1/4″][vc_column_text]Widget Plumx[/vc_column_text][/vc_column_inner][/vc_row_inner][/vc_column][/vc_row][vc_row][vc_column][vc_separator css=”.vc_custom_1624528584150{padding-top: 25px !important;padding-bottom: 25px !important;}”][/vc_column][/vc_row]