This page gives summaries of project that are under development. More details about each project can be found with links
The projects in the portfolio reflect state-of-the-art database management, i.e. how to promote data lineage, provenance, traceability, transparency, and FAIR, Findable, Accessable, Interoperable and Reusable, data practices . The case studies of these projects deal with my background in physical organic chemistry and combustion. A common denominator among recent work is the use of ontologies for not only data descriptions, but also being the base of the model data driven development. All the recent projects have been implemented as cloud based services.
CHEMCONNECT
The ChemConnect project is the development of a cloud-based intelligent repository of experimental and modeling information. The database’s, there are two versions, go beyond simple repositories of data by viewing data as set of linked concepts. The goal of this viewpoint is:
- No data point is isolated: A single data point, whether it be a raw data point from a measurement, a modeling constant or a final result, is not isolated. It has a history and relationships with other data points. CHEMCONNECT provides a foundation for this inter connectivity.
- Traceability, Accountability and Reproducability: Knowing where a data point came from or where it is used promotes the scientific process and data reliability. This is a software implementation promoting the FAIR concept of data.
There are currently two versions of CHEMCONNECT:
- CHEMCONNECT2017: This first version concentrates on parsing the modeling data from combustion reaction mechanisms and interlinking all the molecular, reaction and thermodynamic information for easier search. The key in this version was the use of the Resource Description Framework used by semantic web of the W3C. Its implementation can be found on http://www.chemicalkinetics.info.
- CHEMCONNECT2019: The second version concentrates on providing a full trace of experimental data from the device and its components, measurements, intermediate calculations, final results and modeling results. The key to this version is the use of ontologies, a software technical tool of the semantic web to provide relationships and concepts to domain data. It’s implementation can be found at http://www.connectedsmartdata.info and further information can be found here.
Resource Desription Framework
with ontologies and NEO4J
ChemConnect stores the database information using the Google Firestore framework, but to facilitate queries an Resource Description Framework (RDF) is added. The RDFs are stored using NEO4J, a graph oriented database framework. The information for the RDFs are defined using information from the individual catalog objects. The specific form of the RDFs are defined using the ontology.
Generative AI and ChemConnect
Retrieval-Augmented-Generation (RAG) Using Ontology Descriptions of Graph-Based Data
This goal of this project is to use Retrieval-Augmented-Generation (RAG) to efficiently search a noSQL (Google Firestore) augmented with Graph-Based (Neo4J) Resource Description Framework (RDF) database. The foundation of the project is based on the fact that each of the database objects and each of the RDF instances that are derived from the database are decribed by an ontology. The information using to create the RAG data is derived from this ontology description.
Respecth ontology
In this case study, the classes of ontology objects correspond to the elements of the ReSpecTh specification of the XML data format of the ReSpecTh database. The relationships describe the children of elements and their attributes.
The goal of this case study is to represent all the information and relationships in the hand-generated specification as an ontology. This includes the additional (informal) specifications that are found in the descriptions of the elements.
There are several advantages of using an ontology representation of the specification. The first is that the hand written specification is made more rigorously formal. The second is that it is not a passive descriptive specification but an active specification where questions can be asked about the specification. The ontology specification can be actively used to verify, generate and automatically interpret data formats which use the specification. The specification gives additional machine interpreted semantics and meta data to the specification.
JthermodynamicsCloud
JThermodynamicsCloud is software service for the combustion research domain to perform thermdynamic calculations using Benson Rules and THERM techniques to calculate thermodynamic properties based on 2D-Graphical representations of species and manage the data needed to make those calculations. The software has evolved from THERGAS, to JTHERGAS, to JThermodynamics and finally this web service.
The JThermodynamicsCloud service can be said to be a model driven application, where the ontology is a platform independent model of the data and operational structures. All the ontology concepts outlined here, from the ontology definition to the utilization of this definition in the application, have been implemented. The ontology, as used by the service, has three distinct purposes: documentation, data structure definition and operational definitions. One goal of the ontology is to place as much of the design and domain specific structures in the ontology rather than in the application code. The calculation itself is highly dependent on the varied types of molecular data found in the database The complete service is a system with three interacting components, a user interface using Angular, a (RESTful) backend written in JAVA (with the JENA API interpreting the ontology) and the Google Firestore noSQL document database and Firebase storage.
Combustion Mechanisms
Automatic Mechanism Generation
From an extensive database of molecular and reaction chemistry, an automatic means of generating oxidation mechanisms, valid for both high and low regimes, for a wide range of species is presented. This work is a key component in developing large hydrocarbon, linear and branched, and aromatic mechanisms for surrogate fuel simulations of gasoline, diesel, jet-a and bio-fuels.
Combustion Mechanism Repository (ChemConnect2019)
ChemConnect2019 goes beyond traditional data repositories in that the data is parsed and analysed with respect to an extensive chemical and combustion knowledge base. The parsed data is then linked allowing for efficient searching and comparison of combustion data. The goal is to have all data associated with experiments, from a device description, to the intermediate data (both computed and measured) and to their associated interpretations and the procedures and methodologies to the final published results and references to be available. Having published data linked to its dependent measurements and constants, devices, subsystems, sensors and even people and laboratories provides an effective accountability and more confidence in the data. Data entry and availability can range from private user, to user defined consortia to general public.
Mechanism Characterization and Reduction
This work involves a multi-strategy machine learning approach to automatically characterize combustion mechanisms. For example through clustering, combustion regimes can be characterized. This characterization in regimes can be used in combination with automatically generated decision trees to produce a viable adaptive chemistry method applied to an engine simulation. A fuzzy logic based machine learning clustering and a decision tree technique is applied. Another characterization of the ignition process is through the finding of generic curves. All are used as the basis of mechanism characterization and reduction.
Computer Aided Organic Synthesis
The foundation of my work in combustion kinetics stems from an background in organic and physical chemistry and using AI techniques for extracting information from chemical databases.
The RETROSYN system, a program (written in LISP) for retrosynthetic planning is introduced. Emphasis is placed on its use of reaction classes. The source of information for the reaction classes is a database of reactions (represented simply as a list of reactants and products), and a means by which this information can be extracted, organized, and then used in RETROSYN is elaborated. The organization of the reaction classes involves establishing a hierarchy of classes and subclasses. Further uses and organizations of reaction classes for use in synthetic analysis are given.
The key to the procedure is the set of reaction patterns. A reaction pattern is defined as two sets of substructures (the product and reactant sides of a reaction) representing the essential structural information needed to adequately describe the reactive center (see refs. 10-14 and references therein for descriptions) of the reaction. In our system, a set of reaction patterns is automatically determined in a preprocessing step from a database of specific synthetic reactions from the literature. The reaction patterns are determined by using a graph difference algorithm between the products and reactants represented as colored graphs. These reaction patterns are used in the runtime search procedure.
English