![]() | Only 14 pages are availabe for public view |
Abstract In spite of the various classifications of nuclear reactors either according to their purpose, structure or the energy distribution of neutrons, they all possess the same bases of calculational tools. The design, construction and operation follow up for any nuclear reactor require specific criteria to maintain its safety. These criteria require the knowledge of the parameters affecting the reactor operation like power and flux distribution, control rod worth, effective multiplication factor and fuel burnup. All these parameters must be determined throughout the reactor cycle. Special computer codes have been developed to determine the values of all the reactor parameters. They vary according to the applied model, computational methods and the calculation procedure. These computer codes are called neutronic calculation codes. These codes are validated and maintained for application in a variety of nuclear energy research programs. Researchers have investigated the potential applications of artificial neural network in enhancing the safety and efficiency of nuclear reactors. The areas under investigation are: diagnosis of specific abnormal conditions, detection of the change of mode of operation, signal validation, monitoring of check valves, modeling of the plant thermodynamics, monitoring of plant parameters, and analysis of plant vibrations. The power level of a reactor depends on the macroscopic fission cross section and the neutron flux. Over a short time interval, the cross section remains essentially constant, although it may not have the same value at all locations in the core. Hence, the power level at any instant can be considered proportional to the neutron flux. In most situations a reactor is controlled by varying the neutron flux. Among the general methods available for changing the neutron flux in a reactor is the insertion and withdrawal of the control rods. The present study proposes an Artificial Neural Network (ANN) modeling technique that predicts the control rods positions in a nuclear research reactor. The neutron flux in the core of the reactor is used as the training data for the neural network model. The data used to train and validate the network are obtained by modeling the reactor core with the validated neutronic calculation code; CITVAP. The type of the network used in this study is the feedforward multilayer neural network with the backpropagation algorithm. The results achieved show the potential effectiveness of neural network in determining the control rods positions knowing the neutron flux at each fuel element for a material and test reactor; MTR. The results also show the ability of the neural network to find out functional relationship between the neutron flux (inputs) and the control rods positions (outputs), where we don’t know how to describe the functional relationship in advance, but we do know examples of correct mapping. This is the power of the neural network to discover its own algorithms. The method proposed in the study can be used to predict critical control rods positions to be used for reactor operation in the process of replacing burned fuel by fresh fuel. |