This research consists of two main parts: development of a rock expert system (RES) as an easy-to-use and effective tool for evaluating rock properties, and modification and utilization of the three-dimensional Particle Flow Code (PFC3D) to analyze rock behavior. Because of different reasons, it is often difficult to obtain the rock property values directly. As an alternative, typical values and empirical correlations are often used to evaluate the rock property values. However, the typical values and empirical correlations come in various forms and are scattered in different sources. It is often difficult, time-consuming or even impossible for an engineer to find appropriate information to estimate the required rock properties. So in the first part of the research, the RES was developed as an easy-to-use and effective tool for evaluating rock properties by conducting detailed review and evaluation of well determined values and empirical correlations of rock properties in the published literature, and developing a central database and data application tools. The study of RES demonstrates the storage of rock property values and correlations is strongly applicable and the web based data application tool is effective to use and easy expandable. Considering its granular nature, the discrete element method (DEM) has been widely adopted to analyze the mechanical behavior of rock. The Particle Flow Code (PFC) is one of the most popular DEM softwares. The basic idea of PFC is to treat rock as an assembly of bonded particles that follow the law of motion and consider the model behavior dominated by the formation and interaction of micro cracks developed within the particle-particle cement (bond). Unlike the continuum methods, PFC can deal with the natural process from micro cracking to macro failure, without predefining a failure criterion for the rock. However, there are still issues related to the application of PFC to analyze different rock problems. For example, so far, most of the studies use PFC2D although many of the problems are three dimensional and should be better simulated with PFC3D. It is also found that the simulations using the default PFC parallel bond model extremely underestimate the ratio of unconfined compressive strength to tensile strength (UCS/T). So in the second part of the research, the important aspects related to the application of PFC3D, including model scale, particle size distribution and contact model, were studied, a new contact model was developed for addressing the limitation of the default PFC3D on obtaining unrealistically low UCS/T ratios, and finally the new contact model was used to investigate rock fracture initiation and propagation.