Bearings usually work in the complex noise environment, which makes the time-domain vibration signal easy to be polluted by various noises, thus misleading the diagnosis results.To solve this problem, a method combining a one-dimensional (1-D) denoising convolutional autoencoder (DCAE) and a (1-D) convolutional neural network (CNN) is proposed. In order to simulate the real noise environment, Gaussian noise with different signal-to-noise ratio was added to the original vibration signal. 1D-DCAE was used to denoise the original signals, and then the denoised signal was used for 1D-CNN fault dignosis.1D-DCAE was built based on full convolution network (FCN) and the pooling layer was discarded to improve the anti-noise capability. The results show that FCN-based 1D-DCAE has better noise reduction effect, and the improved model can adaptively diagnose faults of various noise environments.