Artificial neural network models and hybrid machine-learning techniques generally provide more accurate predictions of tunnel boring machine disc cutter wear and penetration rates than traditional empirical methods, according to a systematic literature review published in Discover Applied Sciences. The review synthesizes current knowledge on how rock mass properties govern cutter consumption and evaluates existing prediction methodologies, placing a dedicated emphasis on computational neural network approaches.

In these severe excavation settings, disc cutter wear represents a critical operational challenge that significantly influences penetration rate, overall operational efficiency, and total project expenditure.

To evaluate how wear prediction tools have evolved, the systematic literature review analyzed a broad spectrum of predictive paradigms, spanning classical empirical, analytical, mechanistic, probabilistic, and data-driven models. The investigation critically examined and compared published research concerning rock mass properties, cutterhead design, operational machine parameters, mechanical wear mechanisms, and neural network prediction methodologies.

As part of this assessment, the authors evaluated established engineering wear frameworks alongside recent machine-learning and deep-learning approaches. These traditional benchmarks included the CSM, NTNU, Gehring, QTBM, and RME models.

The synthesis confirms that rock mass properties serve as the primary factors governing disc cutter wear during mechanized tunneling. Specifically, key geological parameters identified in the literature include uniaxial compressive strength, the Cerchar Abrasivity Index, equivalent quartz content, joint spacing, and overall rock mass structure. High-strength and highly abrasive rock formations are consistently associated with increased rates of cutter wear.

The authors observe that appropriate cutter spacing combined with favorable joint orientations can actively enhance cutting efficiency.

Across the evaluated literature, artificial neural network and hybrid machine-learning models consistently demonstrate superior predictive accuracy for both cutter wear and penetration rates when compared with legacy empirical tools. The authors report that these neural network approaches exhibit significant potential for improving wear prediction accuracy and strengthening predictive maintenance capabilities for mechanized excavation projects.

Despite these demonstrated advantages, the review identifies notable constraints that currently limit the practical implementation of machine-learning models in tunneling. The effectiveness of artificial neural networks remains constrained by limited data availability, inconsistent cutter wear metrics across different projects, and insufficient integration of real-time monitoring systems during excavation operations.

To resolve these ongoing challenges and optimize tunnel boring machine performance, the authors emphasize clear directions for future scientific research. The review concludes that future work should concentrate on establishing standardized wear databases, integrating advanced sensor-based monitoring systems, and developing physics-informed artificial intelligence frameworks to support reliable real-time cutter wear forecasting.

Subject of Research: Predicting tunnel boring machine disc cutter wear using artificial neural networks and rock mass properties

Article Title: A review of the influence of rock-mass properties on TBM disc cutter wear with a focus on artificial neural network approaches

Article References: Paul, S., Raina, A. K., Chattopadhyaya, S., & Dewangan, S. (2026). A review of the influence of rock-mass properties on TBM disc cutter wear with a focus on artificial neural network approaches. Discover Applied Sciences. https://doi.org/10.1007/s42452-026-09716-5

Image Credits: AI Generated

DOI: 10.1007/s42452-026-09716-5

Keywords: Tunnel Boring Machines, disc cutter wear, artificial neural networks, rock mass properties, penetration rate, uniaxial compressive strength, Cerchar Abrasivity Index, equivalent quartz content, joint spacing, rock mass structure, CSM model, NTNU model, Gehring model, QTBM model, RME model, predictive maintenance, cutter spacing, physics-informed AI