When diagnosing compound faults of rolling bearings，the signal processing method of separation is usually used before diagnosis. Because of mutual coupling or interference of fault characteristic signals，it is easy to misdiagnose or miss diagnosis.Aiming at this problem，a compound fault diagnosis method based on Autogram resonance demodulation and 1.5-dimensional spectrum is proposed，which can identify fault types without separating compound fault signals. The original vibration signal is denoised by using variational mode decomposition（VMD），and a comprehensive index Z is proposed to select the effective components of VMD for signal reconstruction to improve the signal to noise ratio. The Autogram algorithm is used to determine the center frequency and bandwidth of the resonance frequency band，and the envelope demodulation of the resonance signal is carried out to obtain the 1.5-dimensional spectrum of the envelope signal. The types of rolling bearing composite faults are identified according to the fault features in the 1.5-dimensional spectrum. The feasibility of the proposed method is verified by the measured signals of rolling bearings with three different types of compound faults. The experimental results show that the proposed method can improve the accuracy and intuition of composite fault identification，and has certain application value in engineering practice.