الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 144 |  |  |
| | | from | 144 | | |
Abstract
Industrial systems reliability, availability and economic life are very important parameters directly affecting their economic performance. The present work is an attempt to build more realistic models for the evaluation of industrial systems reliability, maintainability, availability and economic life time. Modeling of any phenomenon, including industrial processes, becomes more complicated and intricate as they approach reality. The reality is approached in the present work, mainly through the assumption of imperfect maintenance and repair activities, and the inclusion of a capacity margin and buffers besides the main system components. The provision of buffers between successive processes in a multiple-process-series system results in a remarkable enhancement of performance. The approach overcomes the major problem of series systems which is the drastic decrease of system reliability as the number of components increases. The interaction between buffers, whether before or after, and equipment leads to dividing the mathematical model of optimal reliability and maintainability allocation problem into two sub-models. The first model is for the optimal reliability and maintainability allocation for load-sharing systems where the buffer is summing of all the extra capacity from load-sharing equipment. The second model is for the optimal reliability and maintainability allocation for series systems where the buffers are allocated between the sub-systems. Use of simulation approach could deal with both models simultaneously. In the present work, the problem of optimal reliability and maintainability allocation for systems with buffers under imperfect maintenance and repair is investigated. The main problem is divided into two tasks. One; is to find the optimal reliability and maintainability allocation for load-sharing systems, and the second is to find the optimal reliability and maintainability allocation for series systems. Considering the imperfect maintenance and repair in the developed models, makes the renewal processes non-homogeneous stochastic processes which is quite challenging. The problem is mathematically formulated as optimization models for different systems configurations (Load Sharing and Series). Solutions obtained by the application of Discrete Event Simulation (DES) and meta-heuristic Genetic Algorithm (GA) are thought to contribute to the rationalization of decisions necessary to design and operate such systems. Mainly, optimum margin-increase of system components capacities, size of buffers to be included between different processes, optimum maintainability and reliability parameters of different components are considered. One of the central assignments undertaken in the presented work is building an empirical model for evaluating systems? availability for the different configurations Parameters in these empirical models require further consideration for building meta models formulating the relationship of systems? availability parameters and parameters of system components reliability, maintainability, components capacity margin and necessary and sufficient buffer storage. It has been demonstrated that these buffer storages play the most significant role in enhancing system availability. Artificial Neural Network (ANN), as function builders connecting system outputs (availability parameters) to system inputs (reliability, maintainability parameters and capacity margins),are adequate models when compared to that obtained by second order regression modelling considered in literature. This could be explained by the fact that ANN does not require prior assumption of the order of suggested meta models and accounts for models nonlinearities by the proper selection of their topologies and evaluation of connecting weights. For the purpose of designing ANN properly, a system of pertinent experiments is designed in such a manner that reasonable coverage of values of inputs (reliability, maintainability parameters and capacity margin) is considered. Rational combination of input values resulted in 27 experiments to be performed applying discrete event simulation. ANN has been tested and found to be adequate.Two of three developed models are applied to one of the systems in a leading gulf Oil and Gas company. The results show that the developed models support the fact that equipment could not work for ever and need to be replaced when their performance drops under a certain reliability threshold. Buffers volume is a paramount system configuration parameter in Oil and Gas where it compromises the low reliability of equipment, which has been proved through applications of models. These facts made the developed models superior compared to other existing models when it comes to model system reliability and maintainability for Oil and Gas industry.