R.B. Newell

Fuzzy adaptive control of a first-order process

Fuzzy adaptive control of a first-order process with a varying gain and time constant is demonstrated. The fuzzy adaptive controller is based on a fuzzy model-based controller that uses an explicit fuzzy model of the process. Adaptation is achieved by the addition of on-line identification of the fuzzy process model. Starting with a crude initial model the adaptive controller matches the model to the process to self-tune the controller. The model is further modified when the operating point is changed to adapt the controller to the new process conditions. Two different fuzzy identification algorithms are both shown to provide a successful fuzzy adaptive controller. The robustness of the algorithms in the presence of process noise is investigated and improved by the addition of weighting factors.

Modelling anaerobic degradation of complex wastewater. I: model development

Complex wastewater is defined in this context as containing significant levels of fats, proteins or particulates. High-rate anaerobic treatment of complex wastewater is becoming more popular due to economic and environmental advantages. However, treatment is complicated because of the high levels of solids and fats and lack of knowledge of degradation mechanisms. This paper is the first part of a study of a structural model describing degradation of complex wastewater. A structural anaerobic model is developed which extends previous research, of which the most recent described the anaerobic degradation of soluble protein (Ramsay, I.R., 1997. Modelling and Control of High-Rate Anaerobic Wastewater Treatment Systems. Ph.D. Thesis, Department of Chemical Engineering, The University of Queensland, Brisbane, Australia). This study incorporates hydrolysis of particulates and long chain fatty acid (LCFA) β-oxidation. A set of physico-chemical equations describes gas transfer and ionic reactions and structured biochemical equations describe biological and enzymatic reactions. A generic parameter estimation methodology for application to full-scale systems is also proposed.

Fuzzy identification and control of a liquid level rig

The most difficult problem in fuzzy control is the development of a control rule set that is both complete and correct. We describe a fuzzy adaptive controller that can learn a control algorithm on-line and adapt it to changing process conditions. The controller makes use of fuzzy identification techniques for learning and adaption. Its control algorithm is based not on an explicit control rule set, but on predictions obtained from a fuzzy model of the process. The learning properties of the controller are demonstrated by its application to the control of a laboratory liquid level rig. Experimental results are given, and a comparison is made with traditional PI/feedforward control techniques.

Mathematical modelling of prefermenters—I. Model development and verification

A dynamic mathematical model of prefermentation is presented. Prefermentation is becoming an increasingly popular unit operation in biological nutrient removal wastewater treatment plants. The aim of the prefermenter model is to provide a tool for optimisation of prefermenter design and operation. Model assumptions, equations and parameters are detailed. The major specification for the prefermenter model is that it is relatively simple, but can still describe the effects of the main design and operating parameters on volatile fatty acid (VFA) production in prefermenters. The model is verified by comparing its prediction with experimental data presented in the literature. Predicted and measured steady-state effluent VFA, soluble COD and ammonia-N concentrations are compared for varying hydraulic and solids residence times. The correspondence between experimental data and model prediction is good.

A systematic approach for reducing complex biological wastewater treatment models

Abstract Biological wastewater treatment systems comprise a variety of processes which occur at vastly different rates. Biological growth, mass transfer, hydraulics and chemical reactions all occur simultaneously and are all inter-dependent. In this paper we address the question “to what extent can we de-couple these processes, and what are the associated issues? We aim to introduce people who work with biological wastewater treatment models to analytical tools which may be used for model reduction. We present a quantitative technique to compartmentalise states into fast, medium and slow. From this we have provided an algorithm for eliminating state variables from a model based on whether they affect the process in the selected “time scale of interest”. Through the technique presented we provide a means of quantifying the interaction between state variables, the “speed” of a state and whether it is a candidate for reduction. A simple case study of a biological wastewater treatment process is presented. We were able to reduce four biological and 19 settler differential equations into algebraic equations. This resulted in significant savings in integration time. Application of the technique also highlighted the strong coupling between the slower biomass states and the rest of the model.

Dynamic simulation of bioreactor systems using orthogonal collocation on finite elements

Abstract The dynamics of continuous biological processes is addressed in this paper. Numerical simulation of a conventional activated sludge process shows that despite the large differences in the dynamics of the species investigated. the orthogonal collocation on finite element technique with three internal collocation and four elements (OCFE-34) gives excellent numerical results for bioreactor models up to a Peclet number of 50. It is shown that there is little improvement in numerical accuracy when a much larger internal collocation point is introduced. Over and above Peclet number of 50, considered to be large for this process. simulation with the global orthogonal collocation (GOC) technique is infeasible. Due to the banded nature of its structural matrix, the method of lines (MOL) technique requires the lowest computing time, typically four times less than that required by the OCFE-34. Validation of the hydraulics of an existing pilot-scale subsurface flow (SSF) constructed wetland process using the aforementioned numerical techniques suggested that the OCFE is superior to the MOL and GOC in terms of numerical stability.

Robust model-order reduction of complex biological processes

This paper addresses robust model-order reduction of a high dimensional nonlinear partial differential equation (PDE) model of a complex biological process. Based on a nonlinear, distributed parameter model of the same process which was validated against experimental data of an existing, pilot-scale BNR activated sludge plant, we developed a state-space model with 154 state variables in this work. A general algorithm for robustly reducing the nonlinear PDE model is presented and based on an investigation of five state-of-the-art model-order reduction techniques, we are able to reduce the original model to a model with only 30 states without incurring pronounced modelling errors. The Singular perturbation approximation balanced truncating technique is found to give the lowest modelling errors in low frequency ranges and hence is deemed most suitable for controller design and other real-time applications

Automatic control of effluent quality from a high-rate anaerobic treatment system.

An anaerobic fluidized bed reactor was overloaded with up to 42 g molasses 1- day-. The effluent quality, expressed in total acids concentration, was successfully regulated by an automatic control algorithm, with alkaline consumption as the controlled variable and feed rate as the manipulated variable. The effluent quality could be maintained at any required level, from 100 to 400 mg acids 1-, depending on the setting of the control parameters. Moreover, it is possible to keep the acid addition within a certain range, setting higher and lower thresholds in the controller. This option requires the use of equalization capacity and is particularly suitable for handling surges of higher than average concentration waste. At lower waste concentrations, the feedrate is automatically increased, allowing the reactor to run at maximum loading with the effluent still satisfying quality requirements.

Mathematical modelling of prefermenters—II. Model applications

This paper investigates several potential applications of a recently developed dynamic mathematical prefermenter model. Three case studies are presented. In the first case study, the model is used to analyse and optimise the performance of a prefermenter. The full-scale complete-mix prefermenter in Penrith. Australia. is used as an example. The model call predict the optimal residence time with respect to maximal mass of volatile fatty acids and soluble COD produced per day. The second study focuses on the effect of different feed types on prefermenter performance under different operating conditions. Four different feed types are defined for this case study, the main difference being their total COD concentrations. According to the simulation results, in-line prefermenters should generally be operated at a longer solids residence time than side-stream prefermenters for maximal rates of volatile fatty acids production. Optimal hydraulic retention limes predicted by the model are considerably lower than what is currently used in practice for side-stream prefermenters. In the third case study the model is used to predict the dynamic behaviour of a prefermenter after a sudden disturbance. Experimental results from a pilot-scale activated primary tank undergoing dynamic transients and the corresponding simulation results are shown. The disturbances were initiated by adding glucose to the feed and by a step change in hydraulic residence lime. The three case studies illustrate that the prefermenter model is a versatile tool for optimising the design and operation of prefermenters

Dynamic modelling and simulation of activated sludge process using orthogonal collocation approach

This paper deals with the improvement in modelling of the dynamics of activated sludge wastewater treatment process using a distributed parameter approach. A computational algorithm, based on the global orthogonal collocation technique, for the activated sludge process is developed in this work. Steady-state and dynamic simulations are performed based on this algorithm. The system configuration considers backmixing or intermixing, which can represent the actual process more accurately than the idealised flow schemes commonly employed for the design and modelling of the activated sludge reactors. Based on a dispersed plug flow model for activated sludge bioreactor formulated in this work, a Peclet number in the order of 0.5 to 5 and 5 to 7 internal collocation points are recommended for modelling channel-type activated sludge bioreactor. The dynamics of substrate is predicted well whereas the prediction for biomass is fair, when compared to the experimental data. It is also demonstrated in this paper that the proposed algorithm can give superior prediction of process dynamics than the commonly-used tanks-in-series technique. Dynamic responses caused by disturbances in the influent wastewater which propagates at various spatial positions along the bioreactor can be correctly predicted through simulations.