The present dissertation work has as the main purpose the reliability and maintenance analysis in unit production of Ammonia in Industry of phosphoric fertilizers production residing in New Karbali- Kavala. The dissertation is constituted by eight chapters. In the first chapter that constitutes the introduction are reported the purpose of the dissertation, the source of data, the structure and the methodological approach that will be followed. In the second chapter is reported concisely the theory of reliability and the mathematical approach for its analysis. Similarly in the third chapter the significance and the theory of maintenance and the basic quantitative measures for the approach are presented. In the fourth chapter are presented concisely previous empirical researches and studies that have been written in the reliability and maintenance theory and applications for various cases and is various branches. In the fifth chapter is presented a description for the structure and the operation of Ammonia unit production and its sub systems, by which it is constituted. In the sixth chapter are presented the numerical data, which are used for the application of the statistical analysis of reliability. The parametric Weibull distribution is selected, the finding of success and failure probability in each subsystem separately and in whole unit as well are reported. Also in the same chapter a Pareto analysis is been made for the of failure type frequency in order to be explicit which type of failure lead to dysfunction and participate at a higher percentage in the production loss. Then a bootstrapping simulation is applied in order to confirm the results that have been found initially. Next, the methodology of neural networks is proposed, which present a great success and augmentative tendency in the application in many sciences and specifically three models are presents and applied. Finally, we apply a neuro-fuzzy model to estimate the reliability of Ammonia production unit. In the seventh chapter are presented the numerical data on the for maintenance analysis. In the eighth chapter Cox proportional hazard models are analyzed and estimated for the preventive maintenance. In the ninth chapter predictive maintenance is analyzed and specifically multinomial Logit models are estimated to predict the probabilities for failure kinds. In the last chapter the conclusions are presented.
- ISBN (eBook)
- 3 MB
- Institution / Hochschule
- Ελληνικό Ανοικτό Πανεπιστήμιο (Fernuniversität Patras) – School of Science and Technology
- Reliability Weibull Neural Networks Neuro-fuzzy bootstrapping Logit Cox hazard