Vibration and oil debris analysis are widely used in gearbox condition monitoring as the typical indirect and direct sensing techniques. However, they have their own advantages and disadvantages. To better utilize the sensing information and overcome its shortcomings, this paper presents a virtual sensing technique based on artificial intelligence by fusing low-cost online vibration measurements to derive a gearbox condition indictor, and its performance is comparable to the costly offline oil debris measurements. Firstly, the representative features are extracted from the noisy vibration measurements to characterize the gearbox degradation conditions. However, the extracted features of high dimensionality present nonlinearity and uncertainty in the machinery degradation process. A new nonlinear feature selection and fusion method,named kernel factor analysis, is proposed to mitigate the aforementioned challenge. Then the virtual sensing model is constructed by incorporating the fused vibration features and offline oil debris measurements based on support vector regression. The developed virtual sensing technique is experimentally evaluated in spiral bevel gear wear tests,and the results show that the developed kernel factor analysis method outperforms the state-of-the-art featureselection techniques in terms of virtual sensing model accuracy. Vibration and oil debris analysis are the widely used in gearbox monitoring as the typical indirect and direct sensing techniques. However, they have their own advantages and disadvantages. on artificial intelligence by fusing low-cost online vibration measurements to derive a gearbox condition indictor, and its performance is comparable to the costly offline oil debris condition. However, the extracted features of high dimensionality present nonlinearity and uncertainty in the machinery degradation process. A new nonlinear feature selection and fusion method, named kernel factor analysis, is proposed to mitigate the said challenge. Then the virtual sensing model is constructed by incorporating the fused vibration features and offline oil debris measurements based on support vector regression. The developed virtual sensing technique is experimentally evaluated as spiral bevel gear wear tests, and the results show that the developed kernel factor analysis method outperforms the state-of-the-art feature selection techniques in terms of virtual sensing model accuracy.