Abstract: For the problem that insufficient lubrication or abnormal abrasion easily occurs during shear gearbox operating, a fault diagnosis method of shearer gearbox based on partial least squares regression was proposed. The Fe content, viscosity, acid value and water content of lubricating oil in shearer gearbox were taken as detection indexes, and a partial least squares regression model was established on the basis of data initialization process and principal component extraction, which could express abrasion of shear gearbox. The model was tested by the actual operation state diagnosis of shear gearbox in a coal mine. The test result shows that error between actual measured value of Fe content in lubricating oil and the predicted value by the model is smaller under normal abrasion conditions of the shear gearbox, but error between the two value is larger under fault conditions, so as to correctly judge the abrasion and fault state of the gearbox.