The geological radar method has found widespread use in evaluating the quality of tunnel lining. However, relying on manual experience to interpret geological radar data may cause identification errors and reduce efficiency when dealing with large numbers of data. This paper proposes a method for identifying internal quality defects in tunnel lining using deep learning and transfer learning techniques. An experimental physical model for detecting the quality of tunnel lining radars was developed to identify the typical radar image features of internal quality defects. Using the geological radar method, a large volume of lining quality detection radar image data was collected, in conjunction with several examples of tunnel engineering. The preprocessing of geological radar data was performed, including gain and normalization, and a set of data samples exhibiting typical lining quality defects was prepared with 6236 detection targets in 4246 images. The intelligent recognition models for tunnel lining quality defects were established using a combination of geological radar image datasets and transfer learning concepts, based on the SSD and YOLOv4 models. The accuracy of the SSD algorithm for cavity defect recognition is 86.58%, with the YOLOv4 algorithm achieving slightly lower accuracy at 86.05%. For steel bar missing recognition, the SSD algorithm has an accuracy of 97.7%, compared to 98.18% accuracy for the YOLOv4 algorithm. This indicates that deep learning-based models are practical for tunnel quality defect detection.

A number of 1/5 scale normal concrete (NC), reinforced concrete (RC), Engineered cementitious composites (ECC) and reinforced ECC (R/ECC) tunnel lining specimens were tested under vertical loading applied at the mid-top of the crown and the mechanical behaviours were analysed. Four lining FE models were established and compared in detail with the test data. Subsequently, numerical analysis of NC lining strengthening methods for different damage levels was performed. The NC linings exhibiting different damage levels were strengthened by either RC arch, ECC arch, R/ECC arch, steel plate and carbon fibre reinforced polymer (CFRP) sheet. The mechanical behaviour of the retrofitted NC linings was analysed, and the strengthening effects were thoroughly evaluated. The results showed that the load-bearing capacity and ductility of NC linings retrofitted by R/ECC arch were better than NC linings strengthened by ECC arch or RC arch. The steel plate and CFRP sheet retrofitted NC linings resulted in lower load-carrying capacities compared with linings strengthened by arches. For the plate strengthening methods, when the damage level of the strengthened NC lining is lower, more pronounced strengthening effects could be attained compared with the highly damaged linings.

The presence of symmetric and asymmetric voids directly affects the mechanical behaviors of tunnel linings and further induces tunnel diseases among influence factors. In this paper, 1:5-scale model tests were carried out to study the mechanical behaviors of reinforced concrete (RC) linings considering the voids located at the crown and at the spandrel. Based on the experimental results and concrete plastic damage (CDP) model, the effects of void (i.e., location and size), subgrade stiffness, and lining size on bearing capacity of RC lining were investigated using numerical simulation. The results of model test and parametric analysis showed that the existence of voids significantly affected the mechanical behavior of the lining during inelastic deformation period. The lining with a larger void size corresponded to low bearing capacity and larger deformation around the void, thus increasing the damage possibility of linings. The influence of voids on the bearing capacity of linings varied with the void location, load direction (especially under horizontal symmetrical loads), and subgrade stiffness. High soil stiffness corresponded to a great influence of the void size on the lining bearing capacity. In addition, the lining strength increased inconsistently with the increase of model size. On the basis of parameter sensitivity analysis, the Levenberg–Marquardt (L-M) optimization algorithm and Logistic model were used to establish the equation of lining bearing capacity loss rate considering the void effect.

Engineered cementitious composites (ECC) have higher ultimate tensile strains than normal concrete. The mechanical properties of ECC strongly depend on raw materials and the mix proportions. The uniform design test method and alternating conditional expectation, which is a nonparametric regression analysis method, were used to design the ECC mix proportion. According to the regression analysis, the optimized W/B, S/B, and F/B ranges could be obtained as 0.35–0.42, 0.25–0.3, and 0.02, respectively. The tested proportions for validation were randomly adopted within the range of W/B, S/B, and F/B. The uniaxial compression, tension, and four‐point bending tests were conducted to verify the material behaviour of the designed ECC. Results showed that all the specimens had large ultimate tensile strains and high fracture energy capacities, and strain hardening was also observed. The fibers were found to be uniformly distributed in the specimens by using a scanning electron microscope. This paper may provide theoretical and practical guidance for the ECC and other cement‐based material mix proportion design.

The mechanical properties of high‐toughness engineering cementitious composites (ECC) were tested, and a damage constitutive model of the materials was constructed. A new aseismic composite structure was then built on the basis of this model by combining aseismic joints, damping layers, traditional reinforced concrete linings, and ECC linings. A series of 3D dynamic‐response numerical models considering the composite structure‐surrounding rock‐fault interaction were established to explore the seismic response characteristics and aseismic performance of the composite structures. The adaptability of the structures to the seismic intensity and direction was also discussed. Results showed that the ECC material displays excellent tensile and compressive toughness, with respective peak tensile and compressive strains of approximately 300‐ and 3‐fold greater than those of ordinary concrete at the same strength grade. The seismic response law of the new composite lining structure was similar to that of the conventional composite structure. The lining in the fault zone and adjacent area showed obvious acceleration amplification responses, and the stress and displacement responses were fairly large. The lining in the fault zone was the weak part of the composite structures. Compared with the conventional aseismic composite structure, the new composite lining structure effectively reduced the acceleration amplification and displacement responses in the fault area. The damage degree of the new composite structure was notably reduced and the damage area was smaller compared with those of the conventional composite structure; these findings demonstrate that the former shows better aseismic effects than the latter. The intensity and direction of seismic waves influenced the damage of the composite structures to some extent, and the applicability of the new composite structure to lateral seismic waves is significantly better than that to axial waves. More importantly, under the action of different seismic intensities and directions, the damage degree and distribution area of the new composite structure were significantly smaller than those of the conventional composite lining structure.