Cutter wear is one of the key factors influencing construction efficiency during shield tunnelling. Prediction of cutter wear can improve construction efficiency by reducing the times of cutter inspections in engineering practice. Evaluation of cutter life is vital for cutter wear prediction, however, existing cutter life indices can only estimate the health condition of all cutters on cutterhead on a holistic basis. A new index was proposed to evaluate cutter wear located at a specific installation position on cutterhead. A deep learning model integrating the index was developed for the estimation of accumulated cutter wear during real time shield tunnelling. The new index can be obtained by monitored field parameters and can predict cutter wear with historical wear patterns. The input and output data samples were reshaped for multi-step prediction. A shield tunnelling section in Guangzhou weathered granite was used for validation. The proposed method can help reduce the cost of cutter replacement by reducing the times of machine interventions. The method article is a companion paper to the original article [1].•Proposed index for prediction of cutter wear rate.•Deep learning model of 1D-CNN and GRU.•Multi-step cutter wear prediction.
Keywords: 1D-CNN; Integrated method based on a new index and CNN-GRU; Wear rate index, Gated recurrent unit, Wear prediction.
© 2023 The Authors. Published by Elsevier B.V.