A Object-Oriented Implementation of a Chemical Waste Consolidation Expert System
Return to front page
1. Background and Rationale
2. Interface Design
3. System Design
4. Evaluation and Results
5. Future Work
Bibliography and References
Appendix A. Example Drum Report
Appendix B. Chemical Compatibility Testing Data
The Physical and Environmental Sciences (PENS) department of Texas A&M University-Corpus Christi needed a specialized software package to assist with many aspects of their operations. A major priority was the need for a system to assist with the final disposal of the chemical inventory and the byproducts of research and teaching laboratory experiments. It was desirable to minimize these disposal costs and disseminate that information to a select number of the faculty, staff and emergency responders.
The waste consolidation software was designed to determine the compatibility of various combinations of chemical waste. Previous predictions have been made based on prior practice and extensive personal knowledge of chemistry. To achieve minimum cost of disposal, the economics of scale generally apply. If a waste handler is to transport gallon-size glass bottles of waste, for example, the bottle is packed in a can of vermiculite; the waste generator pays a contracted rate per pound for the disposal of the aggregate weight of the packing materials, bottles and chemicals. By contrast, 55 gallon drums eliminate the need for the additional packing materials. The difference in cost for the same amount of waste can range from a few hundred dollars for a consolidated drum to several thousand dollars for multiple glass bottles. EOG Environmental [EOG 2005] is the current contracted waste handler, and as expected, the most economical container size for disposal needs is the standard 55 gallon drum. In order to fill a drum of this dimension, many smaller containers of chemical waste must be combined, often creating a mixture of seventy chemicals or more. The complexities of combining many chemicals can prove problematic or disastrous for individuals with modest chemical knowledge.
A search for software that performs this function has so far yielded no exact match. There is software to assist with continuous waste treatment for pilot and production plants with waste streams significantly larger than the one under discussion [Calvin 2005]. These packages are generally for regulatory compliance [EnviroWare 2007], chemical engineering and on-site remediation companies; as industrial-scale processes, they deal with large volumes of relatively few compounds [Calvin 2005]. Computational chemical modeling software abound in a diverse assortment of products such as Spartan [Spartan 2007] and TINKER [TINKER 2007] to effectively predict the properties of compounds and fairly simple mixtures in extreme detail; however, the staggering amount of time and computing power to produce solutions for very simple systems make its application to this problem impractical or impossible. Modeling software programs like [Amber 2005] and [CHARMM 2005] use molecular dynamics and energy minimization to predict chemical interactions and conformational structures. These software packages typically are used to study the relationship of one molecule surrounded by hundreds or thousands of solvent molecules. Amber, for example, operates by “adjusting” one molecule at a time, then “adjusting” the adjacent molecules in sequence. Its goal is to achieve the minimum energy of the system as a whole and can be very time consuming. A representative model of a waste mixture may exceed hundreds of copies of fifty or more constituents. It would require an immense quantity of computational cycles because each constituent would have to be checked against all possible conditions arising from the other constituents. Using this methodology for chemical compatibility calculations is not practical for a waste consolidation software package.
Knowledge of all the intricate reaction possibilities is not needed; the critical data is chemical compatibilities to avoid a runaway reaction and thermodynamic information to prevent the mixture from boiling or reaching a temperature that would cause significant vaporization or ignition of the more volatile components. It was necessary to create a rules-based program that could make informed suggestions for consolidation of the chemical waste in an expeditious fashion. The software had to be stand-alone because the primary point of operation would be at a remote waste storage site running on a portable computer. It was originally proposed that the software would be authored in a visual language to insure accuracy in data entry, specifically the names of the chemicals. The resultant software is not GUI-based but achieves the same goal, allowing the user to select the chemical by inputting a number from an indexed list.
Knowledge-based expert systems have been used for complex tasks such as molecular structure prediction and medical diagnosis [Russell 1995]. To achieve the required level of problem-solving knowledge, the software was structured as a chemical compatibility system that behaves as an expert system, with the data stored in three distinct class objects. The “rules” are input as pairs of chemical name and type, with a separate list of rules containing each chemical name, incompatibility descriptors and reactivity index. By matching the descriptor field in the first list with the incompatibility field in the second list, two chemicals are determined to be incompatible with a level represented by the reactive index. Figure 1.1 shows a schematic diagram of compatibility testing for two chemicals, Compound 1 and Compound 2.
Figure 1.1. Generalized Chemical Compatibility Rule Matching
This software system tracks the contents of the large drums and advises on the order of addition of each waste transport bottle to minimize reaction and temperature fluctuation. It generates forms and reports listing the waste inventory, estimated costs for the waste generators and associated data. The critical data needed to make these decisions was derived primarily from the National Fire Protection Association [NFPA 2007], the Merck Index [Merck 1989], the CRC Handbook of Chemistry and Physics [CRC 1990] and the Material Safety Data Sheets (MSDS) available with each chemical. These are all easily obtainable standard references, or they could be assumed based on the properties of similar compounds. It can be assumed that no software solution to the current problem exists because of the complexity of the interactions involved.