Institute for Systems Research Technical Reports

Permanent URI for this collectionhttp://hdl.handle.net/1903/4376

This archive contains a collection of reports generated by the faculty and students of the Institute for Systems Research (ISR), a permanent, interdisciplinary research unit in the A. James Clark School of Engineering at the University of Maryland. ISR-based projects are conducted through partnerships with industry and government, bringing together faculty and students from multiple academic departments and colleges across the university.

Browse

Filters

1989 - 198931990 - 19916

Settings

Author: search.filters.author.Mavrovouniotis, Micheal L.

Search Results

Now showing 1 - 9 of 9
  • Thumbnail Image
    Item
    Synthesis of Direct Mechanisms for Chemical Systems
    (1991) Mavrovouniotis, Micheal L.; ISR
    A chemical system consists of intermediate species, terminal species, and mechanism steps. Understanding the behavior of a chemical system can be significantly aided by the identification of mechanisms responsible for overall reactions which do not involve net consumption or production of reaction intermediates. Issues arising in the definition and identification of direct mechanisms, which are the shortest possible mechanisms, are discussed. In the context of examples of catalytic synthesis of ammonia and methanol, an alternative approach for the construction of mechanisms from steps is presented. An algorithm for the construction of direct mechanisms is then formally stated; the algorithm is based on successive processing and elimination of reaction intermediates which should not appear in the overall stoichiometry of the reactions accomplished by the mechanisms. Throughout the operation of the algorithm, irreversible steps are used only in their permitted direction. The basic algorithm may construct indirect or duplicate mechanisms, but variations of the algorithm are proposed which discard such redundant mechanisms. A number of hypothetical chemical systems illustrate the differences between the proposed algorithm and other approaches.
  • Thumbnail Image
    Item
    Computer-Manipulation of Conjugate Forms in Proper Estimation
    (1991) Mavrovouniotis, Micheal L.; Constantinou, Leonidas A.; ISR
    Physical and chemical properties of pure compounds and mixtures are essential for the analysis and design of chemical processing systems. A method for the estimation of properties of organic compounds from their molecular structure is presented, based on the contributions of Atoms and Bonds in the properties of Conjugate forms of a molecular structure (ABC). A real chemical compound can be considered the hybrid of a number of conjugates, which are alternative formal arrangements of the valence electrons of the molecule. The property-estimation method generates all conjugate forms of the molecule and assigns properties to each conjugate, simply by summing contributions from atoms and bonds in the particular electronic arrangement of the conjugate. The properties of the actual compound are then derived from the properties of the conjugates. The generation and analysis of conjugates is based on symbolic computation and Object-Oriented Programming (OOP). Atoms, bonds, molecules, electron pairs, and other entities can be represented as interconnected objects within OOP. The generation, comparison, and analysis of conjugates can be carried out through computer- based manipulation of the objects and their interconnections. One needs to encode operators which generate the conjugates, as well a rules for pruning the generation so that only the most important conjugates are considered. The nature and connectivity of atoms within a molecule determine the physical and chemical properties of the molecule. Traditional group-contribution methods eliminate much of the detailed molecular-structure information at an early stage of the property-estimation effort. Through symbolic computation and the concept of conjugation, the ABC approach aims to use molecular structure information more effectively.
  • Thumbnail Image
    Item
    Hierarchical Neural Networks
    (1991) Mavrovouniotis, Micheal L.; ISR
    With the common three-layer neural network architectures, networks lack internal structure; as a consequence, it is very difficult to discern characteristics of the knowledge acquired by a network, in order to evaluate its reliability and applicability. An alternative neural-network architecture is presented, based on a hierarchical organization. Hierarchical networks consist of a number of loosely-coupled subnets, arranged in layers. Each subnet is intended to capture specific aspects of the input data. A subnet models a particular subset of the input variables, but the exact patterns and relationships among variables are determined by training the network as a whole. However, the hierarchy of subnets gives the network hints to look for patterns in the most promising directions. Their modular organization makes hierarchical neural networks easier to analyze, because one can focus on the analysis of one subnet at a time, rather than attempt to decipher the whole network at once.
  • Thumbnail Image
    Item
    Vector Notation for Partial Molar Properties in Single-Phase Simple Systems
    (1990) Mavrovouniotis, Micheal L.; ISR
    Conventional coordinate systems used in thermodynamic analysis of mixtures use either mole numbers or mole fractions as independent variables. The definition of quantities and the derivation of relationships in these systems entail formidable algebra. The introduction of a position vector for the amount and composition of a multi-component mixture allows the expression of useful derivative quantities in vector and tensor notation, independently of the coordinate system used.
  • Thumbnail Image
    Item
    Estimation of Properties from Conjugate Forms of Molecular Structures: The ABC Approach
    (1990) Mavrovouniotis, Micheal L.; ISR
    The ABC technique for estimating properties of organic compounds from their molecular structures is presented, based on the contributions of Atoms and Bonds to the properties of Conjugates. Conjugates are alternative formal arrangements of valence electrons; a real chemical compound is a hybrid of conjugates. In ABC conjugates of a molecule are generated, and thermodynamic and quantummechanical properties are assigned to each conjugate, by summing contributions from atoms and bonds of the conjugate; the properties of the compound are then obtained as combinations of the properties of all the conjugates. In a simple application, ABC estimates the heat of formation of alkanes within 2.8kJ/mol, compared to 4.7kJ/mol for group contributions. The ultimate goal is the estimation of fractional charges on individual atoms of a compound and electron densities of bonds, because these are related to intermolecular interactions and chemical properties.
  • Thumbnail Image
    Item
    Computer-Aided Synthesis of Biochemical Pathways
    (1990) Mavrovouniotis, Micheal L.; ISR
    The synthesis of biochemical pathways satisfying stoichiometric constraints is discussed. Stoichiometric constraints arise primarily from designating compounds as required or allowed reactants, and required or allowed products of the pathways; they also arise from similar restrictions on intermediate metabolites and bioreactions participating in the pathways. An algorithm for the complete and correct solution of the problem is presented; the algorithm satisfies each constraint by recursively transforming a base-set of pathways. The algorithm is applied to the problem of lysine synthesis from glucose and ammonia. In addition to the established synthesis routes, the algorithm constructs several alternative pathways that bypass key enzymes, such as malate dehydrogenase and pyruvate dehydrogenase. Apart from the construction of pathways with desired characteristics, the systematic synthesis of pathways can also uncover fundamental constraints in a particular problem, by demonstrating that no pathways exist to meet certain sets of specifications. In the case of lysine, the algorithm shows that oxaloacetate is a necessary intermediate in all pathways leading to lysine from glucose, and that the yield of lysine over glucose cannot exceet 67% in the absence of enzymatic recovery of carbon dioxide.
  • Thumbnail Image
    Item
    Hierarchical Neural Networks for Monitoring Complex Dynamic Systems.
    (1989) Mavrovouniotis, Micheal L.; ISR
    With the common three-layer neural network architectures, the processing of a large number of signals requires an enormously large neural network. Such a network is very difficult to train and may not have sufficient speed for real-time applications. The computational complexity also prevents the use of additional hidden layers, potentially leading to total inability of the network to capture essential complex patterns in the signals. Furthermore, the lack of internal structure in a large network makes it very difficult to discern characteristics of the knowledge acquired by the network, in order to evaluate its reliability and applicability. We will investigate alternative neural network structures that contain much fewer connections and are organized in a hierarchical fashion. Our hierarchical networks consist of a number of loosely-coupled subnets, arranged in layers each subnet is intended to capture specific aspects of the input data. At the bottom layer, each subnet operates directly on a particular subset of the input variables. In the intermediate layers, each subnet receives its inputs from subnets of the previous layer and sends its outputs to subnets in the next higher layer. Each subnet is expected to model and summarize in its output the important characteristics of a particular set of related input variables. In order to construct the subnets we start from the set of inputs and identify all its subsets which, based on our a priori knowledge of the structure and behavior of the system being modelled, consist of related inputs. We call these subsets input clusters. In general, the clusters will overlap, and there will even be clusters that are fully contained in (i.e., are subsets of) other clusters. This defines the hierarchy that will be used in the construction of the network Whenever a larger cluster is equal to the union of smaller clusters, the subnet that corresponds to the larger cluster will not receive its inputs directly from the input set, but rather from the outputs of other subnets that correspond to the smaller clusters. This should only take place if the rationale for establishing the cluster can be viewed as a combination of the rationales of the smaller cluster, i.e., if the set-subset relation of the cluster is not coinddental. The complexity of each subnet (i.e., its number of hidden and output nodes) can be adjusted based on the complexity of the relationship that led to the establishment of the cluster.
  • Thumbnail Image
    Item
    Symbolic Computing in the Prediction of Properties of Organic Compounds.
    (1989) Mavrovouniotis, Micheal L.; ISR
    A completely new approach is under development for estimating properties of chemical compounds from their molecular structure. The approach is based on viewing each compound as a hybrid of several different molecular structures. The molecular structures are generated and manipulated by symbolic computation methods. The properties to be estimated include not only physical and thermodynamic properties, but also chemical properties of compounds which, currently, can not be predicted by any method.
  • Thumbnail Image
    Item
    Constraint Propagation with N-ary Semiquantitative Relations.
    (1989) Mavrovouniotis, Micheal L.; ISR
    An approach for semiquantitative constraint propagation using both simple and complex nodes is presented. Each node has a label consisting of the union of a negative interval and a positive interval. Compared to simple interval labels, this representation provides significant increase in expressiveness, with only a moderate increase in complexity. In addition to simple nodes (variables), there are complex nodes, representing dimensionless products of variables. Previous efforts have focused on reasoning only with independent complex nodes; other nodes can be expressed as suitable algebraic expressions of independent nodes. The approach followed here involves the use of all irreducible complex nodes; these nodes are the simplest possible, in that they cannot be broken into smaller complex nodes. Since irreducible nodes are not necessarily independent, they are related by implicit constraints. The number of nodes and the number of implicit constraints are polynomial in the size of the problem. The coexisting layers of simple and complex nodes can be manipulated to limit the propagation: Labels on complex nodes are only propagated if they contain information that is not already provided by the simple nodes. This effectively reveals those complex nodes that bear interesting labels. These representation and reasoning choices are suited to engineering domains in which many dimensional kinds of variables are present and dimensionless ratios of variables are significant in defining the state of a system.