Prediction of Disc Cutter Replacement of Tunnel Boring Machine Using Denoising Auto Encoder

While a tunnel boring machine (TBM) is working, rocks are crushed into pieces by disc cutters which often fail during construction. To replace disabled cutters timely, the condition of the cutters needs to be checked regularly. However, this is time-consuming and uneconomical. In this paper, a denoising auto encoder (DAE) model is proposed to judge whether TBM disc cutters need to be replaced. First of all, the field data related to dis cutter status are selected as inputs. Then, the cutter conditions can be learned automatically base on a DAE model. Test result on a water transport tunnel shows that the proposed model can obtain an average accuracy of 99.7% and an average F1 score of 99.4% on field data prediction. Compared with other machine learning and deep learning models, proposed method reduces the need of manual data denoising and feature extraction.