Explore the vast range of titles available.

The deep tank aeration process has attracted attention because of its unique characteristics: effective land use and high oxygen transfer efficiency. One operating problem being reported to be associated with this system is the flotation of sludge in final settling tanks. Unlike sludge bulking occurring in conventional activated sludge processes, which is usually a pure biological phenomena, supersaturation of dissolved gas in the mixed liquor in the deep aeration tank is believed to be a prime factor causing sludge flotation. Presently, efforts to understand these phenomena are mainly experimental, and no theoretical models have been proposed. Because of the complexity of the system that involved a combination of biological and physical processes, and the fact that the relationships among the process parameters are not clearly defined it is difficult to use traditional deterministic approach to study the phenomena of sludge flotation. Therefore, a stochastic time-series approach is selected to build a forecasting model for process operation and control. This dissertation uses the most popular time-series forecasting process--univariate Auto-Regressive Integrated Moving Average (ARIMA) and Multivariate Transfer Function (TF) methods to study the cause-effect relationship between sludge flotation and system operating parameters, and to develop forecasting models for plant operation control. The results demonstrated that the time series analysis technique is a powerful tool which provides an adequate description of the dynamic behavior of the deep tank aeration system. Both univariate ARIMA method and multivariate TF models can adequately simulate and forecast the dynamic behaviors of the system. While the ARIMA model outperformed the multi-TF models over the period with low output variability, however, the multi-TF models with the added explanatory power, is more accurate during the period with high fluctuations.