Research On The Risk Assessment Model Of Geophysical Exploration Operations Based On The Combination Of Rough Sets And Neural Networks

The premise of extracting oil and natural gas is to determine the location of oil and gas layers, storage and mining difficulty by geophysical Company's related work. However, because of the environment of geophysical prospecting construction work is complex, high-risk job and construction difficulties. There are a lot of risks in the construction process, once these factors result in corresponding accidents, it will cause human, financial, material, environment harmed or damaged, thereby affect the process of exploration of oil and natural gas. In order to reduce the occurrence of accidents during construction, this paper will combine rough set with BP neural network to build a risk assessment model. The role of the model is to ensure the security situation of the construction process effective supervision. This thesis focuses on the following aspects were studied:(1) Analysis the incompleteness of calculating the attribute significance based on attribute dependency or based on information entropy. Combined attribute dependency of an attribute with attribute information entropy to build new attribute significance. As a mature algorithm in the field of modern combinatorial optimization, ant colony optimization algorithm (ACO) has great advantages in solving NP-Hard problems. Thus, regarding the new attribute significance as the heuristic information of ant colony optimization to build an attribute reduction algorithm based on ACO.(2) In this thesis, based on the "safety regulatory early warning system" project of a drilling company developed by our laboratory, according to the illegal and hidden dangers appearing in the geophysical exploration, from the point of view of people and things to build a three-layer index system of influencing the security situation in the process of geophysical construction operations. Build a data set based on the third layer index in the three-layer index system, and use the algorithm based on ACO to reduce the data set.(3) According to the reduction results and the casualty degree of accident to build the input-layer and output-layer of BP neural network, and use the trial and error method to gain the number of neurons in hidden-layer based on the neurons of input-layer and output-layer; Then, using the training samples train the network model based on different hidden-layer neurons, and the number of hidden-layer neurons with the lowest network error is selected; At last, using testing data to validate the network model, ensure the feasibility of the model.(4) Using C# programming language to implement the assessment model we established.