Prediction Of Rockbursts Based On Genetic Algorithms And BP Neural Network

More and more waterpower and mineral resource need to be exploited for the high-speed development of economy and society in recent years, therefor numbers of water and hydropower engineering projects and deep exploiting projects are going to be carried out in succession. As a result, more and more engineering of new projects will be built in underground caverns which are very long, deep and dense. High geostresses is the typical characteristic in these areas. As a basical kind of geological disaster, rockbursts just occure in high geostresses zones usually. It is a pity that there is no perfect prediction theory or mechanism so far. So it is meaningful to predict and prevent rockbursts. So the GA-BP Neural Network was built in this paper to predict whether the rockburst would happen and its intensity. Some actual data of rockbusts in former engineering were collected in this text to test prediction effect. The main work in this paper is generalized as following.1. On the basis of colleting numbers of achievements in home and abroad, the text introduced the research actuality of rockburst, such as rockburst type and intensity classification, common characteristics, formation condition and mechanism, methods of prediction.2. The losing circumstances brought by some rockbursts in home and abroad were collected stated, and the paper put forward the constructing trend of water and hydropower engineering projects and deep exploiting projects, quoted some important hydraulic engineering, pointed out the importance and imminence of rockbursts prediction.3. On the basic of synthetically analyzing two aspects, theory and engineering application, the paper pointed out that tensile strength, compression strength, shear stress of surrounding rock and elastic energy index were main factors which afected the occuring of rockburst chiefly, so three representative indexes, brittleness index, elastic energy index and stress intensity ratio, were selected to be the main affecting index and to be input parameters of BP network. 4. Artificial Neural Network(ANN), especially BP Network, including its basic theory, application, advantages and shortcomings, and the characteristic of Genetic Algorithms(GA) were introduced in detail in the paper. Application of GA in design of ANN, viz. optimizing ANN with GA, was put forward too in the paper, which can help to solve the congenital limitation of ANN.5. ANN and GA were failed together according to their characteristics, forming GA-BP network. Some initial data were selected as training stylebook. First, the best original weights and threshold,which would be the original data of BP network, were got by the running of GA. Then stylebook were trained by BP network to establish the potential relationship between stylebook and factual result of rockbursts. This process was called study of network. In the end, project cases were fetched in to validate the actual predictiong effect of GA-BP network. The result indicated that the prediction was consistent with the fact.6. Compared the prediction result of GA-BP network with the prediction result of BP network and common criterion of rockbuest respectively. The result indicated the prediction effect of GA-BP network was better than the latter two methods. It can be used to predict the occuring of rockburst.In the finality, some points and the limitations of prediction of BP network, along with further studies are discussed.