Application of the selected classification models to the analysis of the settling capacity of the activated sludge – case study

Abstract

Abstract. The study presents the development of classification models for sedimentation of activated sludge using the artificial neural networks (ANN), logistic-regression (RL), and linear discrimination model (LDM). The input consisted of indicators of wastewater quantity and quality (biochemical oxygen demand, chemical oxygen demand, total suspended solids, total nitrogen and total phosphorus at the inflow to the wastewater treatment plant) and operational characteristic of the bioreactor (pH, temperature of activated sludge, mixed liquor suspended solids, concentration of oxygen in the nitrification chamber, amount of PIX dosing).The prediction quality of the developed models was measured with: sensitivity, specificity, and computed errors. The calculations of the sedimentation were performed for sludge volume index (SVI). The results indicate that successful predictions were obtained using ANN, RL and LDM methods, which is supported by the fit of computations to measurement results. The study shows that for the wastewater treatment plant of concern, sedimentation properties can be obtained using only the loads of organic compounds, mixed liquor suspended solids, temperature, pH of activated sludge, concentration of oxygen in the nitrification chamber and amount of PIX dozing. Other analysed variables appear to be statistically insignificant for the sludge volume index.

Effect of Physicochemical Wastewater Parameters and Abiotic Factor on Activated Sludge Sedimentation Capacity

“…The influence of temperature on the improvement of sedimentary abilities was also demonstrated by Rössle and Pretorius [5], who developed a non-linear regression relationship for the sedimentation forecast based on the results of temperature and sediment concentration measurements. The dependences obtained in the paper were also confirmed by the author's research [11], who using the logistic regression method demonstrated the influence of the increase of oxygen concentration in the nitrification chamber and of sediment temperature on improving the sedimentation abilities of the activated sludge.…”

“…Mathematical models have been used on numerous occasions to predict sludge sedimentation capabilities [8][9][10][11][12][13][14][15], whereby due to the complex nature of the processes occurring in activated sludge they were black-box models, in which the structure of the model is generated at the stage of learning without having to know the physical laws that describe the analyzed phenomenon. The disadvantage of these models is that one cannot clearly determine the impact of particular explanatory variables on the dependent variable.…”

Effect of Parametric Uncertainty of Selected Classification Models and Simulations of Wastewater Quality Indicators on Predicting the Sludge Volume Index

“…In this study, in order to compare the prediction capabilities of classification models, we considered the use of the logistic regression method and the Gompertz model. These methods are commonly used in medicine, microbiology and the social sciences, and a review of the literature [1,3,4] shows that the Gompertz model has shown much greater capacity to classify both economic events and disease cases than logistic regression. Therefore, analyses should be performed to compare the capabilities of these models to predict sedimentation and to identify the model whose use in everyday practice will assist the plant operator in making appropriate decisions, thereby improving the efficiency of the facility's operation.…”

Using Data Mining to Predict Sludge and Filamentous Microorganism Sedimentation

“…The data in Table 1 indicate that the values of the sewage quality indicators (BOD, COD, TN, TP, TSS, N-NH 4 ) varied in a broad range, affecting significantly the activated sludge in terms of concentration and settleability and, as a result, also its substrate load. The process data in this research paper were the basis of theoretical considerations of which the aim was to develop models for predicting such parameters of the activated sludge as concentration, substrate load, sludge index, and quality indicators [5,6,20].…”

“…In this category, multiple regression is the simplest method and the more complex methods include artificial neural network, support vectors, genetic programming, regression trees and others. Since the processes occurring in activated sludge are very complex, statistical models are used in the simulation of its settling properties, as reported in [5][6][7]. It is worth noting that continuous control of the parameters determining the sludge settling process is essential to the course of the sewage treatment process.…”