VISUALIZING PROCESS INFORMATION AND THE HEALTH STATUS OF WASTEWATER TREATMENT PLANTS A Case Study of the ESPRIT-Project WaterCIME Peter Dannenmann Deutsches Forschungszentrum für Künstliche Intelligenz GmbH P.O. Box 2080 67608 Kaiserslautern, Germany* peter.dannenmann@dfki.uni-kl.de Hans Hagen University of Kaiserslautern P.O. Box 3049 67653 Kaiserslautern, Germany hagen@informatik.uni-kl.de Abstract In this case study we present an approach to support the operators of wastewater treatment plants by supervising the health status of the plant with an online diagnosis software and by visualizing the process data and the diagnosis results. In addition to measuring the available on-line data, we simulate the behavior of the biological processes in order to retrieve data on-line that normally only can be determined by analyzing water samples in the laboratory. Based on the automated analysis of these data we provide to the operator a view of the states of the biological processes within his plant. This information allows an early reaction in the case of deviations from an optimal process behavior. Thus the plant can always be operated in an optimal state so that we can be sure that the treated water always meets the environmental standards that are required. Keywords: information visualization, process control, simulation, diagnosis * Peter Dannenmann conducted this work while staying with Astrium GmbH, P.O. Box 28 61 56, 28361 Bremen, Germany 1 2 1. Introduction Process supervision is an important topic in wastewater treatment. Downtimes of a plant are expensive and in this domain they even may cause environmental problems. Anyway, most of the current supervision systems restrict to displaying only large numbers of sensor-values which require an experienced operator to determine the current status of the cleaning process. In case of an error normally there are lots of messages about limit-violations in some process-parameters that only give little hint to the actual failure. What is even more important, many of the process-parameters are not even on-line available. Here it is an important task to supply the appropriate information about the process-state to the operator and visualize it in a way that covers the vital information and is easy to understand. In this paper we propose an approach to support the plant-operator by visualizing the health status and information about the biological processes within the plant in a clearer way than it is currently done by just displaying raw measurement values. For this purpose, the plant’s “Supervision, Control and Data Acquisition System” (SCADA System) is linked to a decisionsupport system based on dynamic simulation of the processes within the plant. This permits to acquire deeper information about the process-state from the simulation model than we can get only from on-line measurements. With the support of such a knowledge-based system that uses on-line measurements as well as simulated values of the cleaning-process, deviations from an optimal process state can be detected early. The decision-supportsystem then will provide suggestions for re-establishing an optimal processstate. The information about the current state of the cleaning process as well as suggestions for re-establishing an optimal cleaning process are visualized to the operator. These suggestions can be verified by simulating the future development of the cleaning-process. The work described in this case study has been performed in the scope of a research project sponsored by the European Union within the “European Strategic Programme for Research and development in Information Technologies” (ESPRIT). In this project, named WaterCIME (ESPRIT Project 8399), a European consortium of eight companies developed new methods to improve water management information systems by developing an open software framework combining advanced software tools. One of our tasks within this project was the integration of advanced technologies for supporting the operation of wastewater treatment plants into this framework. Visualizing Process Information 3 In the next chapter we will describe the simulation- and diagnosiscomponents that are the basis for the decision-support system. Here we introduce a representation for the biological processes within wastewater treatment plants in the form of simulation-models. The graphical representation of these models visualizes the structure of the underlying biological processes much more adequately than it is done today in a set of differential equations. Afterwards we describe in chapter three the methods how the information from this decision-support system is visualized to the plant operator. In chapter four we discuss the experiences we made during a test installation of the system. We conclude our paper with the lessons learned from the use of this decision-support system in operating a wastewater treatment plant. A short outlook on future trends will close our discussions. 2. Simulation- and Diagnosis Components of the Decision-Support System 2.1 A Simulation Model for Representing Complex Biological Processes in Wastewater Treatment Modern wastewater treatment plants are designed to eliminate several pollutants from the incoming water. Besides removing organic substances an important task is the removal of nitrogen and phosphorous. In modern plants this is done in a nitrification / denitrification process for organic components and nitrogen-compounds and by chemical and biological phosphorouselimination. The current standard for describing these processes is set by the International Association on Water Quality (IAWQ). Currently most simulation is based on the “Activated Sludge Model No. 1” and the „Activated Sludge Model No. 2“ (ASM1 and ASM2) (Henze et. al. 87, Henze et. al. 95) that describe the biological processes as a set of eight differential equations (equal to eight processes covered by the model) with thirteen parameters (for thirteen substances in the wastewater that are affected). Additionally the sludge-thickening process is also described in a similar way. The sludge in the wastewater that has passed the biological stage settles down in “settling tanks”. Most of that sludge is re-fed into the biological stage and only a small percentage is removed from that circulation and taken to the surplus-sludge treatment (Benefield, Molz 84). In order to get a realistic model of the sludge-circulation in the whole cleaning process a 4 model for the settling of the sludge is needed. In contrast to the biological processes described above, in the settling tanks only the physical processes of the sinking sludge are of interest. In the literature there are several models for these processes. They start with models of ideal settlers that completely separate solid and soluble fractions (SIMBA 95, Otterpohl 95) but can also be quite complicated ones that subdivide the settling tanks in several layers with different sludge concentrations and that describe transport-processes of the several fractions between these layers (Härtel 90, Takacs et. al. 91). Since the presentation of the differential equations describing these processes to the plant-operators is not useful, the information contained has to be presented in a more intuitive manner. Therefore we developed an ASM1 representation (Lunde 96) in the simulation tool CSS (Columbus Simulation System) (CGS-Project 95), that has been developed at Astrium. In addition to that model of the biological part of the plant we also modelled the settling tanks. These are the components that are vital to the cleaning process, so we restricted to them. The graphical editor of the simulator allows modeling the processes in a very clear manner. It supports a structural decomposition of the system by its ability to model it hierarchically. The differential equations are hidden in function blocks which now build up a structural representation of the sewage plant instead of a functional one. The simulation-model is decomposed according to the different stations the water passes during the purification-process (Figure 1). In every block the decomposition of the wastewater into the fractions of solid and soluble substances in the water is computed, so that at every point in the simulationmodel a characterization of the wastewater is available that can be visualized to the operator or can be used within the diagnosis component. The computation of the characterization of the wastewater is done by a set of blocks that represent the biological processes taking place within a particular stage of the sewage plant (Figure 2). In addition to the possibility to get an overview of the health-status of the whole cleaning-process (see chapter three) here the user can get detailed information about the single process parameters that are either measured on-line or simulated (as shown in figure 2). With such a structured approach the processes taking place within the sewage plant can be displayed more clearly than in their original form and they are easier to understand to the operators. Visualizing Process Information 5 Figure 1. Simulation model of the sewage plant Figure 2. Simulation model of the circulation tanks 2.2 The Diagnosis Component in the Architecture of the Decision Support System The decision support system (Abels 95) integrates the simulation component described above with the diagnosis software CONNEX-II (CONNection-matrix EXpert system), developed at Astrium (Sielaff 92). This architecture is used for supervising the cleaning-processes in the sewage 6 plant. It is the results of the diagnosis process that are visualized to the plantoperator to support him keeping the cleaning processes in an optimal state. Figure 3 describes the architecture of the decision-support system: The acquisition of the on-line measurements is performed by a SCADA-system that is coupled to the sewage plant’s data bus. This system also serves as a database for the measured values and additionally performs some preprocessing where it is necessary (e.g. when the depending modules need average values of some sensors instead of the actual ones). The simulation-system is fed with the on-line measurements and controlinputs of the plant. Here one simulation-model is run in parallel to the cleaning process. This model provides via the SCADA-system those data to the diagnosis-component that else could only be gathered by laboratorymeasurements. Database Process data DataPreprocessing DiagnosisModule (CONNEX-II) Control DiagnosisUserInterface Modelvisualisation & Control Modellvisualisation & Control Diagnosismodel Simulationmodel Diagnosisknowledgebase Figure 3. Architecture of the decision support system A second model within the simulation-system is used for simulations into the future. If the diagnosis-component detects a non-optimal process state, this model can be used to verify the suggested recovery-strategies. Based on a given state of the on-line simulation, the user can simulate the effects of different control-strategies on the future behavior of the cleaning-process. Here he can check the results of the proposed actions either in the simulation-model directly or the diagnosis-component can perform these checks automatically and visualize the results to the operator. All on-line and simulated sensor values are fed from the SCADA-system into the diagnosis-component. With a real-time inference engine this module detects deviations from the optimal process-state, informs the user and suggests actions to improve the quality of the cleaning-process. Visualizing Process Information 3. 7 Visualizing the Process State to the Operator of a Wastewater Treatment Plant The main user-interface of the decision-support system is the diagnosis user-interface (Figure 4). It has been integrated with the SCADA-system and provides comfortable access to all related modules. Here the diagnosisprocess can be controlled and the health-status of the cleaning-process can be visualized (Abels, Dannenmann 97a). Figure 4. Main user interface of the decision support system for wastewater treatment plants Besides access to the process-parameters (Figure 5), the analysis-results of the diagnosis-component as well as suggestions for recovery-actions (Figure 6) can be displayed here. However, since these data are mere text, advanced visualization techniques are needed, to give the user a clearer impression of the diagnosis system’s findings. This is done by integrating a virtual environment into the user-interface. Within this environment, the socalled “System Explorer,” a virtual model of the wastewater treatment plant is available in VRML format. All relevant components within this model are linked to their counterparts within the simulation-model, which in turn serves as a basis for the knowledgebase used for determining the healthstatus of the whole cleaning process. Thus we get a consistent representation of the wastewater treatment plant as well in the simulation-model as in the VR-model and the representation of the diagnosis-results. 8 Figure 5. Displaying raw sensor values in a listbox Figure 6. Displaying diagnosis-results in a list-box with corresponding explanation In operating the plant, this allows the user now to get a more detailed view of the processes in the plant: First a conventional user-interface is available. The operator can display a list of sensors that are available within the plant where he can check the sensor-values directly. In addition the location of the sensors within the plant is shown in an overview-window. This helps the operator in selecting the correct sensors if there are several sensors of the same type distributed all over the plant. In combination with the values retrieved from the simulation-model a good overview of the current parameters of the cleaning-process is available to the plant operator. For further support in plant operations, the diagnosis-component permanently monitors all measured and simulated process-parameters and performs a system-diagnosis on the cleaning-process. First, the results are displayed within the main user-interface in a textual manner. If a deviation from an optimal process-state or even a failure within a component is detected, a message is displayed to the operator that states the possible reason of the problem. In contrast to the conventional process-supervision systems that only state limit violations at several sensors, this is a big advantage to the operators. Additionally, an explanation text is available, that can give a deeper explanation of the reasons of the failure and countermeasures suggested by the diagnosis-component (Figure 6). At the same Visualizing Process Information 9 time the component where the problem occurred is highlighted within the virtual representation of the plant (Figure 7). Figure 7. Highlighting a faulty component in a turbine of the process-air supply-system This helps the operator to get a clearer view of what had happened and which measures he can take to get the cleaning-process back to an optimal state. If a technical problem has occurred, the operator gets important information about the location of the faulty component within the plant and can arrange the repair of that component quite quickly. 10 Due to its modular architecture, the decision-support system can also be ported to other wastewater treatment plants without much effort. In order to install such a system on another plant, three tasks have to be performed: The simulation model has to be adapted to the new plant, the VR model also has to be adapted and the diagnosis knowledgebase has to be adjusted to the process parameters within the new plant. With the use of component libraries in a manner similar to (SIMBA 95), the modelling effort for the simulation model restricts to selecting components that reflect the architecture of the actual plant and connecting and parametrising them. These parameters are known for every plant, so that they just have to be entered into the model. The virtual model of the plant can also be adapted to a new architecture in the same manner. Also here a set of component- or subsystem-models is available, that permits a quick and easy generation of a virtual model of a wastewater treatment plant. The third adaptation, the adjustment of the diagnosis knowledgebase is a little different. Since we do symptom-based diagnosis, the symptoms have to be adjusted to the new plant and the new simulation model. However, since the structure of such a set of symptoms is the same for all sewage plants (the processes within the plant are always the same), only the new parameters that describe the symptoms have to be adjusted to the specific plant. Additionally, the possible failure modes have to be linked to the new virtual model in order to allow a proper visualization in the new environment. This also can be done easily in an editor integrated into the supervision system to give the model-developer an easy access to all modelled components. 4. Experiences During the Test Installation of the Decision-Support System The modules contributed by Astrium Space Infrastructure and the University of Kaiserslautern were tested within the WaterCIME project’s test-environment at our project partner’s wastewater treatment plant (Abels, Dannenmann 97b). For evaluating the modules, we used operational data provided by the plant’s existing supervisory system. On this basis we ran the modules in parallel to the existing control system of the plant. During the tests the decision-support system showed its capabilities to provide clearer information to the plant’s operators than the conventional supervisionsystem. The operators considered the system’s user interface as very userfriendly. The system gave them easy access to information they else have to Visualizing Process Information 11 gather from different sources in conventional control systems. However, since the controllers of the plant were quite conservative, they first had to get used to the decision-support system. Then, after a short time, they esteemed especially the way how a summary of the plant’s operational state and suggestions for control actions to keep it in an optimal state were presented to them. In addition, they found it particularly valuable how the system presented to them in a graphical environment the locations of faulty components together with proposed corrective actions. However, since on-line simulation of biological processes is still a topic directly at the frontier of research in wastewater treatment, the simulation model of the cleaning-process could not be kept as close as possible to the biological processes as intended over a longer period of time without recalibration. This was a drawback to the application and caused us to shift our focus for further applications on the coupling of the visualization-system to the diagnosis-component alone. Here the integrated diagnosis- and visualization-system proved its usefulness in providing the user detailed visual information about the location of faulty components in a complex technical system and thus supporting a quick repair and reducing downtimes of the affected system. 5. Conclusions In this paper we presented an approach to support the supervision of wastewater treatment plants by integrating simulation-, diagnosis and visualization-components into a plant’s supervision-software. A prototype of the presented system has been implemented in a sewage plant, where it was tested with real process-data. On one side the tests revealed weak points in the available process-models describing the biological processes within wastewater treatment. On the other hand, the tests showed the usefulness of enhancing modern diagnosis systems with advanced visualization techniques to display the location of faulty components to the users, thus providing valuable information for a quick repair of the affected system. Our further work will now concentrate on a generalization of the techniques for coupling simulation, diagnosis and visualization in the product development process for technical systems. We will transfer the techniques developed within this project into an integrated product development environment, where we use simulation- and visualization-techniques for an early prototyping of technical systems under development. For the later phases it is our objective to use simulation techniques to retrieve 12 diagnosis-information for that technical system and use that information in an integrated operations-support environment. Here we will provide advanced diagnosis- and visualization-techniques to support the user in operating the newly developed technical system. References Abels, T., Deep Diagnosis - Module Description and Design, ESPRIT Project 8399 „WaterCIME“, Document No. T6.3-DASA-FS, Daimler-Benz Aerospace AG, Bremen, 1995 Abels, T., Dannenmann, P., User Guide to Deep Diagnosis Module, ESPRIT-Project 8399 „WaterCIME“, Document No. T6.2-DASA-UG, Daimler-Benz Aerospace AG, Bremen, 1997 Abels, T., Dannenmann, P., BEB Test Report, ESPRIT Project 8399, „WaterCIME“, Document No. T8.6-DASA-TR, Daimler-Benz Aerospace AG, Bremen, 1997 Benefield, L., Molz, F., A model for the activated sludge process which considers wastewater characteristics, floc behaviour and microbial population, in: Biotechnology and Bioengineering, Vol. 26, pp. 352-361, 1984 CGS-Project, CSS User Guide, Document No. MA 1214597 005, Daimler-Benz Aerospace AG, Bremen, 1995 Härtel, L. Modellansätze zur dynamischen Simulation des Belebtschlammverfahrens, Schriftenreihe WAR, Band 47, Darmstadt, 1990 Henze, M., Grady, C.P.L. Jr., Gujer, W. Marais, G.v.R., Matsuo, T. Activated Sludge Model No. 1, IAWQ, London, IAWPRC Scientific and Technical Reports No 1, 1987 Henze, M., Gujer, W., Mino, T., Matsuo, T., Wentzel, M.C., Marais, G.v.R., Activated Sludge Model No. 2, IAWQ, London, IAWQ Scientific and Technical Reports No. 3, 1995 Lunde, R., Graphische Modellierung biologisch-technischer Systeme am Beispiel der Abwasserbehandlung, Master’s Thesis, University of Kaiserslautern, Germany, 1996 Otterpohl, R., Dynamische Simulation zur Unterstützung der Planung und des Betriebes kommunaler Kläranlagen, Schriftreihe GWA, Band 151, Aachen, 1995 Sielaff, Dr. C., COMPASS II Knowledge Base Editor Software User Manual, Technical Note OT111-COMPASS II-92/01-5, Daimler-Benz Aerospace AG, Bremen, 1992 SIMBA Simulation der biologischen Abwasserreinigung, Benutzerhandbuch, Fa. Otterpohl Wasserkonzepte GbR, Lübeck, 1995 Takacs, I., Patry, G.G., Nolasco, D. A dynamic model of the clarification-thickening process, Wat. Res. Vol. 25, No. 10, pp. 1263-1271, 1991
© Copyright 2025 Paperzz