Research About Strength Properties Of Granite Under True Triaxial Stress State And Prediction Of Strength Based On Neural Network

In recent years,in order to solve the increasingly sharp contradiction between supply and demand of resources,hydropower cause were developed vigorously in our country and the excavation of resource trended gradually to deep underground.All of these results under the high stress cause geological hazards problems increasingly,such as the serious deformation of mining roadway,rock burst,impact pressure and so on.These engineering disasters threat the safety of personnel.At the same time,the mechanical behavior of rocks will change under the condition of high stress.Such as the brittleness performance of the hard rock under low stress state will show the ductility under high stress state,the performance of rheological properties,etc..So the true triaxial test for rock mechanics behavior under high stress is of great significance in theory and engineering.In the study of rock mechanics behavior,rock strength is the important aspect.As the basis of rock mechanics theory,scholars have already put forward many strength criterions to describe the strength of the rock characteristics and destruction law as early as the 18th century.For the high nonlinearity,it has brought great difficulty and inconvenience during the progress of research and application.Recently,the developments in the field of artificial intelligence provided a good path for people to solve the problem about complex rock mechanics.Evolutionary neural network combine with artificial neural network and evolutionary algorithm,with nonlinear system,the global network,massively parallel distributed processing,and many other characteristics.It has been applied by the scholars in the field of rock mechanics and gradually became a hot research topic.In this paper,the research mainly includes the following several aspects:(1)In this paper,we introduced comprehensively the true triaxial apparatus for hard rocks at Northeastern University,including the system structure units and the performance of each unit,etc.(2)For the existence of many issues about volume deformation measurement device during the true triaxial test progress in the past,we improved the device.For the sake of verifying the machine performance,tests about the calibration of the testing system and the stability test were performed,and these works guaranteed the reliability of the test data.(3)We introduced the new type metal spacers used by Northeastern University.And through a lot of trying,we solved the sealing technology problem finally.The metal spacers moved relatively during the loading progress,so that the rock samples would state in the middle position all the time.Through the contrast analysis for conventional type metal spacers and the new type metal spacers,we find that the stress concentration was improved greatly by the improvement.(4)Through the whole stress-strain analysis for hard rock,we tried to research the regularity and mechanism of granite from initial loading to the damage.At the same time,based on the experimental datas,we discussed the hot research problem-intermediate principal stress effect and rock mass dilatancy which was of great significance for the stability evaluation of engineering research.(5)We made full use of the advantages of highly nonlinear mapping ability of neural network and the global optimization characteristics of genetic algorithm.Throuth the the analysis of the predicted results by a variety of strength criterion and neural network in hard rock true triaxial rock strength,we tried to explore the feasibility of the neural network in this respect.