Proposition of New Structures to the “Don’t Know” Classification Applied to the Problem of Diagnostic of Abnormal Situations in Nuclear Power Plants

Machine learning, a field that has grown in the past years, includes applications in the Nuclear Engineering domain. The problem of diagnostic of abnormal situations (PDAS) can be seen as an application of the separation problem of machine learning to help operators understand the signals in the control room in nuclear power plants. To avoid returning wrong responses that could worsen the judgment of the operator, a methodology called “Don’t know” has been developed and applied to the PDAS problem. This article introduces new geometrical structures with the objective of upgrading the Don’t know model proposed in the literature, improving its accuracy, precision, recall, and specificity using a new proposition of algorithm H-QEA, based on the quantum-inspired evolutionary algorithm, to optimize the classification of events.