머신러닝 기반 저압 막여과 공정 자동진단 알고리즘 개발

Abstract

The water treatment process using membrane filtration removes suspended and insoluble substances through the sieving effect using the membrane pore size and produces water of target quality based on the principle of passing substances smaller than the pores. At present, the membrane filtration process is controlled and managed by data analysis and interpretation based on the expert knowledge of the technician. Therefore, it is currently difficult to automatically operate the membrane filtration process without a technician. the present study, we developed a model to classify the data of the normal filtration process that can occur in the membrane filtration water treatment process and the data of the abnormal filtration process, such as when there is a problem with the membrane. The role of the developed model is to replace the technician's data analysis and interpretation. The machine learning algorithm used in the classification model used the decision tree and K-NN algorithm optimized for classification. Each model applied reinforcement learning to continuously learn new data, and as the learning data was accumulated, the prediction performance converges to 100%. It shows the possibility of automatically diagnosing and controlling the membrane filtration process without a technician through the performance of classification model.