In the strategic action of ore prospecting breakthrough， the drilling depth is constantly increasing， and intelligent working condition recognition has become a necessary measure for safe， efficient and low-cost drilling. Its steps include signal acquisition， feature extraction， state recognition and diagnosis decision， among which the establishment of accurate and reliable in-hole accident discrimination criterion is the difficulty and core of state recognition and diagnosis decision. Firstly， based on the theoretical knowledge and engineering practice experience of in-hole accidents， this paper divides the six types of accidents in drilling holes into buried drilling， stuck drilling， burned drilling， broken drilling， hole leakage and hole overflow， and summarizes their occurrence symptoms. The characteristics and change trend expression of real-time drilling signal parameters are analyzed in detail. On this basis， the relative mean， standard deviation， mean slope， standard deviation slope and their thresholds of six characteristic parameters， including pump pressure， drilling speed， torque， rotation speed， overhead weight and outlet flow， are proposed to characterize the change of each parameter. Finally， the accident discrimination criterion is established in the form of matrix. The discrimination criterion has strong theoretical and practical basis， and can divide the detailed drilling accident categories in actual use， which can provide a discrimination basis for subsequent software development.