Cellular Automata BIORemediation system M.C.Baracca, P. Ornelli, G.Clai E-mail: baracca@bologna.enea.it E-mail: ornelli@bologna.enea.it ENEA HPCN Via Martiri di Montesole 4 40129-Bologna, ITALY Abstract We present the CABIOR system developed on SGI Onyx2 for an application of parallel computing to the simulation of the bioremediation interventions on polluted soils. The CABIOR system is devoted to the needs of the bioremediation user which generally is not skilled enough in modelling or in computer science, so it has been designed in order to hide as many computational technical details as possible and at the same time to provide advanced features to analyze and to test the model outcome. 1 Introduction The CABIOR system has been developed on SGI Onyx2 platform by ENEA in the framework of the Esprit HPCN COLOMBO project coordinated by CRA Montecatini. The main objective of this two-years collaborative Project was the application of parallel computing to the simulation of the in situ bioremediation of contaminated soils: a three layers (fluid-dynamic, chemical, biological) bioremediation CA based model, including three phases (water, pollutant, air), has been implemented and tested both on pilot plants and real fields [1]. The theory of CA lends itself well to the bioremediation simulation. The automaton is related to a finite three-dimensional space and consists of a number of cells. Each cell is associates with a set of neighbouring cells. All cells contain a set of qualities, which are known as its substates. For example, each cell has a substate defining it as a piece of ground ( as opposed to being a well ) and others substates indicating its porosity, its depth from the surface, its water content and so on. The cells are assumed to interact to the neighbours through a specified rule that decides the state of each cell in the next time-step based on the current state of the cell in question and that of its neighbours. Bacteria and pollutants interact gradually over time and only within their neighbourhood area. This locality feature also benefits parallel programming since one can decompose the model and map the components to cooperating computing processors with relative easy and acceptable communication overhead. Moreover the flow of the simulation could be affected by a computational steering mechanism which controls particular cell substates values on regions of the model. From the SW point of view, the Project has achieved these results: - the CAMELot environment (built on MPI) used by the modelists for developing, executing in parallel and monitoring CA-based models [2,4]; the CABIOR system providing a graphical interfaces for computational features and complex data visualization, portable across various platforms [3]. 2 CABIOR Overview The CABIOR system is taylored on the needs of the bioremediation user which generally is not skilled enough in modelling or in computer science, so it has been designed in order to hide as many computational technical details as possible and at the same time to provide advanced features to analyze and to test the model outcome. It has been developed by means of AVS/Express [5], a commercial visualization package, portable across several platforms of industrial interest, that provides the graphical primitives allowing a sophisticated data visualization. Since at the beginning of the current year the AVS/Express version for LINUX platform was released, the CABIOR application has been succesfully ported by ENEA on the end-user PC cluster. The CABIOR system consists in a graphical environment allowing to enter the pre-processing and post-processing facilities as well as to run the batch simulation and the parameters optimization algorithms. Fig 1: CABIOR Main Window In the following sections the features and the functionalities provided by the CABIOR system are briefly summarized. More detailed information can be found at the web site: http://eboals.bologna.enea.it/colombo. 3 Pre-processing tool The graphical pre-processing tool assists the user during the preparation of input data since it allows the visualization, while editing, of the binary data files required by the CA-based model simulating the bioremediation intervention. The input data files are of two different kinds: one describing the general automaton characteristcs and several other files representing the automaton substates, that is the variables defining the state of each cell. The general characteristics of the automaton can be set or modify by means of the General Parameters Menu while the Parameters Set Menus allow the separated input of the simulation parameters referred to each of the specific bioremediation layers. Fig 2: Automaton General Parameters Menu Pressing the Edit.cmt button, a further panel (Fig.3) opens for the editing of 1/2/3D automaton substates with the possibility to combine a random or constant initialization of the cell values with the single cell value setting, in order to easily initialize homogeneous cells arrays and discontinuities like surfaces, lateral walls, injection and extraction wells. Relevant to the user is the panel capability to show the numerical cell values of a chosen automaton plane simultaneously with its discrete visualization according to the colormap at the bottom of the window. Fig 3: Substates Editor 4 Simulation tool The aim is to allow the end-user to start the bioremediation parallel simulation in batch mode and to produce the periodic outputs which could be visualized subsequently by means of the post-processing tool. The related window simply requires the user to choose the input files following the CAMELot files name convention, the output files name and the number of processes to handle the task, while hiding the practical issues to obtain the executable and then run it on the underlying parallel architecture. 5 Optimization tool In the model there are some parameters which cannot be directly determined, so that their values will be adjusted by comparing the model outcome with a set of experimental data. The basic idea is that it is possible to tune these parameters using experimental data resulting from small scale tests, and that they could be succesfully applied on a much larger scale simulation. Fig 4: Parameters Tuning The optimization procedure applied to bioremediation simulation results, interacts with the simulation code as sketched below Fig 5: Optimization Overview The optimization sequence panels permit to choose the optimization technique (Genetic Algorithm or Simulated Annealing), to turn on the model parameters suitable for optimization, to set the parameters required by the method itself, to input the experimental data file and finally to run the optimization procedure. Fig 6: Simulated Annealing Settings Window 6 Post-processing tool The aim of the post-processing 3D visualization is to perform a detailed analysis of the data gathered during the simulation in order to obtain meaningful insights into simulation results. Using this tool, a bioremediation user may analyze, study and discover useful features and numerical results that cannot be discovered during the simulation. Its main features can be summarized as follows: - to visualize 3D data sets in a 3 dimensional space, using different geometrical modalities like orthoslices, isosurfaces, isovolumes and volumes of cells whose values range in a chosen interval (Bounded volume feature); - to rotate, to translate, to scale and to zoom the picture; - to show the substate value and the coordinates of a single cell simply clicking on the image in the Visualization space (Probe feature); - to visualize vectorial fields; - to visualize the temporal evolution of a substate, according to a chosen geometrical modality, using a sequence of snapshots taken during the simulation at different steps. The user can activate or deactivate the geometrical visualization modalities and the implemented feature addressed in the Select Visibility panel of the Manage Visibility module simply clicking on the toggles. Fig 7: Orthoslices and Probe feature In Fig.7 the 3D matrix values related to pollutant pression substate are represented by means of three orthoslices and the value of the cell selected by the mouse is shown. The isosurface modality creates a surface of a given constant value level in the selected substate. The isosurface feature provides a graphic depiction of the locations of a particular data value in a 3D field. Moreover, the isosurface modality (Fig.8) has been improved adding the opportunity to map on a substate isosurface the values of another substate of the automaton, in order to give evidence to related cell qualities. Fig 8: Example of oil potential isosurface with porosity cell values mapped on The aim of the isovolume module (Fig.9) is to display the interpolated volume of the cells whose values are greater or less than a specified level: the input mesh of the selected substate is cut using the isolevel value set by means of the slider in the panel. The Above toggle turned on, means that the output values are greater than the cut level. Fig 9: Isovolume Fig 10: Example of vectorial field: water flux superposed to water potential Vectorial fields visualization has been included in order to show in an effective way the substates related to the phase fluxes. The vector field is visualised by arrows representing the flux direction while the module is mapped within a suitable chosen color scale. In Fig.10, a more significant visualization has been obtained by means of the superposition of a phase substate and its relative flux as a vector field: in the example the water flux is superposed on the water potential substate. The visualization of the temporal evolution of substates, using a sequence of snapshots taken during the simulation at different steps has been implemented too. The animation may be performed for each one of the supplied geometrical modality, as well as for the vector fields visualization. 5 Conclusions The CABIOR System has been accomplished taking into account the end-user requirements, the portability demand across various platforms, including PC cluster, and the need of an easy-to-use interface. This graphical tool allows the flexible and user friendly input of the data required by the bioremediation model, the setting and the execution of the optimization procedures and the accurate analysis of the bioremediation simulation outcame. References 1. M.Villani, M.Mazzanti, R.Serra, M.Andretta, S.Di Gregorio "Simulation model implementation description", Deliverable D11 of COLOMBO Project, July 1999. 2. S.D.Telford, G.Smith, M.C.Baracca, A.Longo, P.Ornelli, G.Spezzano, D.Talia " Design for Portable, Parallel CA Software Environment", Deliverable D6 of COLOMBO Project, May 1998. 3. M.C.Baracca, P.Ornelli, G.Spezzano, D.Talia "Functional Requirements and Software Package Design", Deliverable D10 of COLOMBO Project, November 1998. 4. K.Kavoussanakis, S.D.Telford, S.Booth, L.Clarke, G.Smith, A.Trew, A.Simpson, G.Spezzano, D.Talia "CAMELot Implementation and User Guide", Deliverable D9 of COLOMBO Project, May 2000. 5. Advanced Visual System Inc. "Using AVS/Express", July 1998.