Rock abrasivity and hardness are two of the crucial mechanical properties in geological exploration and petroleum engineering. To figure out how the rock mineral composition determines the rock mechanical properties, ninety-six samples from ten provinces of China were collected to carry out tests including mineral contents, mineral particle size, abrasivity and hardness, and testing results indicated there is strong relationship between them. Through data processing of normalization, correlation analysis, and grouping, the raw testing data were used to establish a prediction function with Back-Propagation Artificial Neural Network (short as BP-ANN). With this prediction function, rock abrasivity and hardness can be accurately calculated from input parameters including rock type, mineral contents, and particle size. Besides, the calculation results from this prediction function also revealed the changing trend of abrasivity and hardness on how to be affected by mineral contents and particle size.