A gene expression programming-based method for real-time wear estimation of disc cutter on TBM cutterhead

Frequent entry in the tunnel boring machine cutterhead for disc cutter wear inspection is a risky, time-consuming, and labor-intensive activity. Existing disc cutter wear prediction models mainly focus on cutter consumption before construction, and it is impossible to estimate the wear of a single cutter when they are applied to on-site construction. To solve this problem, this research presents a method for estimating the wear of each disc cutter on the cutterhead in real time by only using several monitored machine parameters. Firstly, a novel health index that can characterize the wear of each disc cutter is constructed, and the field parameters that have greater impact on the health index are selected. Then, the explicit mathematical expression between the selected parameters and the health index is established based on genetic expression programming. Finally, the on-site data collected from an Indian subway tunnel were used to validate the effectiveness and superiority of the proposed method. The results show that the proposed method can estimate the wear of each disc cutter in real time only by monitoring the rotational speed of cutterhead and tunneling speed. Its average accuracies on the validation set and test set are 90.6% and 85.9%, respectively. Compared with the ridge regression, decision tree, support vector regression and k-nearest neighbors, its accuracy on the test set is 13.0%, 12.7%, 11.2%, and 15.4% higher, respectively. Therefore, the proposed method can greatly reduce the cost for cutter inspection, and the explicit model can be easily deployed to the construction site.