Investigating the Efficiency of Machine Learning Methods for Simulating the Effects of the Paper-Pulp Industry on Aqua Ecosystem Pollution

The paper and pull industry has grown exponentially since 1960 and it is one of the main reasons for water pollution. Due to the rapid extension of the paper and pull industry and its considerable role in aqua ecosystem pollution, analyzing and managing the related pollutant factors are essential. This is not an easy task since sewer space limitations for using monitoring equipment. In addition, laboratory analysis of pollutant factors takes a long time and may affected by measurement error or some undefined induced error. To overcome these difficulties, this paper aims to use machine learning tools for analyzing the pollutant space. The chemical oxygen demand (COD), mixed liquor suspended solids  (MLSS), and pH are considered the main parameters for analyzing pollutant systems. First, the experimental values of MLSS and COD for different hydraulic retention times (HRT=12, 18, and 24) are obtained. After that, the efficiency of linear regression, generalized additive model, neural network, and support vector regression for simulating and predicting the trend of MLSS and COD are investigated. In addition, these methods are considered for predicting pH in the membrane-aerated biofilm reactor (MBR) and the membrane-aerated biofilm reactor (MABR). The numerical results show that  NN is a highly accurate method for predicting COD and MLSS and GAM can predict pH accurately. In addition, the results indicate that HRT=18 is the most accurate and stable time retention for analyzing COD and MLSS.

Investigating the Efficiency of Machine Learning Methods for Simulating the Effects of the Paper-Pulp Industry on Aqua Ecosystem Pollution Keywords:

Wastewater treatment , COD , MLSS , membrane bioreactor