Abstract: The accurate parameters calibration of rock specimen is the preconditions for applying the particle flow program to the resolution of actual engineering problems,and tensile strength is one of the parameters.This paper presents a study on the accurate and efficient back analysis of microscopic parameters of particle flow rock,by comparing and analyzing the Brazilian test and direct tensile test and in combination with the neural network which possesses the ability of non-linear mapping.The conclusions are as follows:Brazilian test,whose back analysis accuracy is only 61%,cannot guarantee the successful operations of the numerous tests,and also cannot provide high-quality samples for the neural network as a result of its demanding requirements;Direct tensile test,which is highly adaptable and has a back analysis accuracy of 81%,can ensure the quantity and quality of the samples for the neural network;BP neural network has the ability of mapping the macro and micro parameters accurately on condition that the quantity and quality are guaranteed,which indicates that it is an effective method for the parameters calibration of particle flow rock material.