Date of Award
Doctor of Philosophy (PhD)
Graduate School of Computer and Information Sciences
Michael J. Laszlo
William L. Hafner
This study investigated the feasibility of developing a rule-based expert system to support nuclear plant operators in the mitigation of accidents. Human error in the use of Emergency Operating Procedures (EOPs) was the problem to be addressed. The goal was to determine if the Nuclear Plant Information System (NPIS) at one specific power plant would support development of an EOP operator aid to reduce the number of human errors.
A review of the literature confirmed that human error is a recurring problem. This is particularly evident during high stress situations, such as accidents. A rule-based operator aid can reduce human performance errors. This made the research significant enough to justify its undertaking. Previous studies have developed rule-based operator aids for new plant design, but ignored nuclear plants constructed before the availability of computers.
The data required for performance of each Emergency Operating Procedure step was identified. Individual step data requirements were compared to data presently available from the Nuclear Plant Information System. For needed data not directly available, alternative parameters were selected where possible. The alternative parameters provide the same plant information to the operator as that described in the Emergency Operating Procedure step. Parameters that cannot be monitored directly or by an alternative method are identified. This was documented in an EOP data matrix. Using the EOPs, a set of rules was developed for a 10, 000-gpm large break loss of coolant accident (LBLOCA). A plant specific simulator scenario for a LBLOCA was conducted. Data collection was performed during this scenario. A comparison of the collected data to the EOP rules determined all required information is not available. This was documented in an EOP rules matrix. A separate matrix compiled all parameters not directly on NPIS and without an acceptable alternative. The effect of each missing parameter on the mitigation strategy is documented.
An EOP expert performed an independent verification of the rules matrix. This verifies correctness of the rules and the required data. A NPIS database expert performed an additional independent verification of unavailable NPIS data. This verified that the matrix data identified as unavailable from NPIS was correct. The study resulted in a set of recommendations for improving plant process computers, EOP procedures, and EOP development techniques. This approach may be used in other nuclear plants. Additional research avenues were identified.
Jesse E. Smith. 2000. The Feasibility of Developing a Computer Based Expert System to Mitigate Nuclear Power Plant Accidents. Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, Graduate School of Computer and Information Sciences. (847)