Noise cancellation for continuous wave mud pulse signal based on Bi-ConvLSTM

Mud pulse telemetry technique is a common method of data transmission used in measurement while drilling (MWD). During the data upload process, the continuous wave mud pulse signal received on the ground is completely covered by noise due to interference of various noises. Therefore, this paper proposes a continuous wave mud pulse signal noise cancellation method based on bidirectional convolutional long short-term memory neural network (Bi-ConvLSTM), which transforms the signal denoising problem into a time series data regression problem of neural network. First, the sample data set is built using the simulated mud pulse signal, which includes a useful signal without noise and the corresponding noise-containing signal. The established neural network is then trained with the sample data set to obtain a neural network model that has a good denoising effect. Finally, the trained model is validated with the validation set samples with different signals to noise ratio (SNR). After the network model denoising, the SNR of the signal increased by about 18 − 30 dB and the correlation coefficient increased by about 0.65 − 0.89. The trained model can effectively suppress the noise from the continuous wave mud pulse signal, which confirms the effectiveness and feasibility of the method proposed in this paper.