Availability analysis of dense plasma focus equipment through neuronal networks

The search for a virtually unlimited renewable energy has been one of the motivation focuses in nuclear energy research by atomic fusion. However, one of the main problems with fusion energy is being able to keep it stable with enough energy to preserve the chain reaction. It was thought that this could only be achieved by energizing a gas at such level that it becomes plasma and keeps it stable. However, other approaches assume that plasma collapses, and others attempts aim to exploit this fragility. The Dense Plasma Focus allows generating an ascending plasma that lasts hundreds of nanoseconds before collapsing. It is of great interest to study the behavior of this plasma in PF equipment in various aspects. Because of this, the study of plasma behavior is essential, but this requires experiments to determine the restrictions, necessary conditions, and energy emissions that are characteristic of the fourth state of matter. The Chilean Nuclear Energy Commission (CCHEN) possess a department dedicated to the study of the physical properties of plasma, depending on the parameters of the equipment used to generate them. In addition, it is of interest to understand the applications that plasma could have in materials science, biology and waste disposal, among others. However, failures in these devices are common and prevent the development of experiments at the expected time, causing subsequent problems in several ways. This analysis includes a comprehensive study of the operation of the plasma generation equipment system of the Dense Plasma Focus type. To achieve this, this analysis determines critical elements, system diagrams and equipment drawings. In addition to obtaining parameters that allow evidence of an advance in the deterioration of the system, it takes into account the corresponding probability distributions. Finally, this study proposes redesigning of the system, with the aim of improving its reliability and availability. Along with this, the study also proposes maintenance protocols and recommendations on monitoring of key variables in the detection of system failures. This report concludes that a monitoring based on the percentage growth of Kurtosis, Shape Indicator, Clearance Indicator, Crest Indicator and Damping of the characteristic signals of the system, can estimate the number of shots from the beginning of an experiment. On the other hand, the study also concludes that distinguishing the data presented by the Dip instance allows establishing the probability of occurrence as a Weibull distribution. This allows distinguishing data that have Dip from those that cannot be performed by pre-treating the signals and subsequent training of a neural classification network. The classifier allows identifying with an assertiveness of 93; 25% for data of the same type and 83; 3% for different data