Theoretical And Experimental Research On Damage Characters Of Concrete Due To Triaxial Compression

The wider utilization of concrete has necessitated a better understanding of its damage characteristics and constitutive behavior under various loading patterns. The aim of this dissertation is to shed some light on these subjects. First, systematic experimental study and theoretical analysis have been performed to study the damage of concrete due to loading history. In addition, an artificial neural network approach is developed to study the constitutive relations of materials, and some constitutive models of concrete under various loading history have been established. The main contents of this thesis are as follows:The first part comprises first three chapters of thesis, in which, some preparatory research on damage and constitutive behaviors of concrete are presented. The theory of continuum damage mechanics and the basic principle about neural network are introduced. First, the developments of damage variables are studied, and the new definition of damage and the technical routes to study are presented. The combination of lateral stress and maximum vertical strain denotes the loading history and the reduction of tensile strength and compressive strength denotes the damage of concrete. This work is the theoretical foundation of study on the damage, of concrete due to loading history. Second, basic knowledge about artificial neural network and BP network model are studied in this part, the disadvantages of BP model are pointed out, and the improved algorithm is presented. This method has unique learning capabilities, which can be used in learning complex nonlinear relationships, and offers a fundamentally different approach in modeling of constitutive behavior of concrete.The second part comprises 4* to 7th chapter of thesis, which is the most important part of thesis. It is concentrated on the investigation of the laws of damage initiation and development, under the conditions of monotone loading, cyclic loading, varies confinement pressures and triaxial equi-compression. First, the damage trial of concrete cube is introduced. The specimen size is 100X100X100", and the number is about 700. Sufficient test data were acquired for investigating the damage and constitutive behavior of concrete. Second, according to the fitting of experimental data, the evolution equation of damage is obtained. The main factors affecting the damage development are examined and the causes are expounded. Finally, The relationship between the damage and the descent of ultrasonic velocity due to triaxial compressive loading history was investigated. It is concluded that the ultrasonic pulse method is convenient and applicable to estimate the damage of concrete due to loading history in engineering practice.The third part comprises 8th and 9th chapter of thesis, which mainly studies the damage behavior and constitutive model of concrete using neural network method. In the beginning, study on the damage characteristic of concrete due to loading history based on BP network, has been carried out for the first time. According to the simulative results, the laws of damage are investigated using Mier's viewpoint that the cracks within concrete can be divided into two types-initiative one and passive one. Then, the techniques of neural networks are also applied to the constitutive modeling of concrete. The main benefits in using a neural-network approach are that all behavior can be represented within a unified environment of a neural network and that the network is built directly from experimental data, i.e., the network is presented with experimental data and learns the relationships between stresses and strains. Such a modeling strategy has important implications for modeling the behavior of concrete. Consequently, the behaviors of concrete under triaxial compressive loading, including monotonic proportionately loading, cyclic proportionately loading, axial monotonic and cyclic loading with invariable confining pressures, are modeled with a back-propagation neural network. The preliminary results of using neural networks to...