Tunnel Defects Research Articles

Probing finite element modelling of defects in friction stir welding by tailoring mass scaling factor

Friction stir welding (FSW), a solid-state welding process primarily used to join the low melting point materials, often encounters surface and subsurface defects. Modelling surface texture and internal volumetric defects in the FSW process using any numerical method is computationally expensive. The analysis of thermo-mechanical behavior of the FSW process using coupled Eulerian-Lagrangian (CEL) method is sensitive to mass scaling factor (MSF). Tailoring this factor for a computationally efficient model is a promising approach to predict defect in FSW process. The influence of MSF on the computational framework of the CEL finite element method (FEM) in the context of defect formation for two different types of materials (difference in densities) is addressed in the present work. The progress of different energies associated with the CEL approach is indicative of the solution availability by the application of MSF. The variation of these energies is significantly high in FSW process when encountering a defect and dictates the optimum conditions of MSF. An optimum range of MSF is deliberated on account of an efficiently model of FSW process. The simulated surface and subsurface defects, and the discrete temperature data are compared with experimental observation. The maximum difference between the predicted and experimentally measured tunnel defect is ∼ 0.75 mm, whereas the predicted temperature is within 5% of experimental result. It establishes the reliability of the developed numerical model and directs to trigger the model parameters for minimizing the computational time by tailoring the MSF. The optimum range of MSF between 106 and 107 strikes a balance between the result accuracy and computational feasibility for welding of AZ31B. Here, the distribution of strain and strain rate is representative of predicting the defect formation, such as, void, no material zone, crack, surface texture etc. The strain rate is almost double on the retreating side (RS) compared to advancing side (AS) due to the presence of the tunnel defect on AS.

Variation Mechanism of Three-Dimensional Force and Force-Based Defect Detection in Friction Stir Welding of Aluminum Alloys.

As a direct reflection of the interaction between the stirring tool and the base metal in the friction stir welding process, the force signal is an important means to characterize welding quality. In this paper, the variation mechanism of three-dimensional force and its relation with welding quality were explored. The acquired signals were subject to interference from high-frequency noise, so mean filtering and variational mode decomposition were applied to obtain the real signals. The denoised signals were analyzed and the results showed that the traverse force was ahead of the lateral force by a ratio of π /4, while the phase difference between the axial force and the other two forces changed with the process parameters. Through application of the least square method and polynomial fitting, the empirical formulas of three-dimensional force were obtained, and these were applicable regardless of tunnel defects. The minimum value of the lateral force increased several times more than that of traverse force when the welding speed increased from 80 mm/min to 240 mm/min. When the pole radiuses of most data points had a value greater than 4, tunnel defects were highly likely to generate. In order to predict welding quality more accurately, a prediction model based on long short-term memory was constructed. The model recognized the various modes of good welds and tunnel defects with 100% accuracy. The identification ability for large and small defects was relatively poor, and the average accuracy of classifying the three categories of welding quality was 84.67%.

Deep learning based distortion restoration and defect segmentation from linear scanning camera images

With long-term operation of tunnels, the surface of tunnel lining is prone to crack, spalling, water leakage and other defects. Therefore, the detection of tunnel surface defects has become an important process to ensure the safety of tunnel operation. Linear scan cameras have become the main way of tunnel surface defect detection because of their ultrahigh resolution and ultrahigh scanning frequency. However, in the process of collecting tunnel surface images at high speed, the linear scan cameras are easy to cause images distortion, which leads to reduction of tunnel defect identification accuracy. Thus, this paper proposes a tunnel surface distortion image restoration method based on a supervised generative adversarial network, which introduces an attention mechanism to guide the method to calibrate the weights of feature channels. It solves the problem of distortion in the process of images acquisition of the linear scan cameras. Then, to deal with the difficulty in identifying small-size tunnel defects in complex tunnel environments, this paper combines a backbone network with a multi-scale feature fusion module based on the You Only Look At CoefficienTs (YOLACT) network, which enhances the feature extraction ability of the network and improves the accuracy of tunnel surface defect identification.

Enhancing the coherence of superconducting quantum bits with electric fields

In the endeavor to make quantum computers a reality, integrated superconducting circuits have become a promising architecture. A major challenge of this approach is decoherence originating from spurious atomic tunneling defects at the interfaces of qubit electrodes, which may resonantly absorb energy from the qubit’s oscillating electric field and reduce the qubit’s energy relaxation time T1. Here, we show that qubit coherence can be improved by tuning dominating defects away from the qubit resonance using an applied DC-electric field. We demonstrate a method that optimizes the applied field bias and enhances the average qubit T1 time by 23%. We also discuss how local gate electrodes can be implemented in superconducting quantum processors to enable simultaneous in situ coherence optimization of individual qubits.

Investigation of Magnesium and Chromium Fillers FSW Dissimilar Joint of AA6063 and AA 5154

Friction stir welding (FSW) is a solid-state metal joining process. There is no melting and recasting of metal while welding is used. Some of the defects commonly encountered in FSW are tunnel defect, bond, cracks, pin holes, and pipping defects. The defects occur because of improper metal mixing and less heat input in the weld nugget zone. In the fusion welding process, a filler rod is employed to form a quality weld with superior mechanical properties. In this work magnesium and chromium powders are used as filler materials. The purpose of this study is to ascertain whether filler materials and manufacturing processes have an impact on the weld nugget zone spot weld joint formation as well as the mechanical and abrasive properties of welded joints. This study's FSW filler materials mixing ratio and process parameters were improved by using the Central Composite Design (CCD) idea, which is discussed in more detail below Response Surface Methodology (RSM). The best empirical relationship between the parameters was provided by the CCD. The mathematical relationships were established to forecast the maximum tensile strength, maximum weld nugget hardness, and minimum corrosion rate by incorporating filler materials with process parameters. The optimal processing factors combination is predicted by conducting the validation test. The optimum parameters were the tool rotatory speed 600–1000 revolution per minute, welding speed 60 to 180 mm/min, plunge depth of 0.05 to 0.25 mm, center distance between the sample is 0–4 mm, as well as powder mixing ratio of 90 : 10, 92.5 : 7.5, 95 : 5, 97.5 : 2.5, and 100 : 0, the tensile test, microhardness, and corrosion rate analysis were conducted on the weld specimen. The welded test specimen provides better joint strength, weld nugget hardness, and enhanced corrosion resistance properties. The microstructure analysis shows the fine grain structure and homogeneous distribution of filler material with the base metal in the welded area.

Motion control and positioning system of multi-sensor tunnel defect inspection robot: from methodology to application

As the mileage of subway is increasing rapidly, there is an urgent need for automatic subway tunnel inspection equipment to ensure the efficiency and frequency of daily tunnel inspection. The subway tunnel environment is complex, it cannot receive GPS and other satellite signals, a variety of positioning sensors cannot be used. Besides, there are random interference, wheel and rail idling and creep. All the above results in poor performance of conventional speed tracking and positioning methods. In this paper, a multi-sensor motion control system is proposed for the subway tunnel inspection robot. At the same time, a trapezoidal speed planning and a speed tracking algorithm based on MPC (Model Predictive Control) are proposed, which simplify longitudinal dynamics model to overcome the complex and variable nonlinear problems in the operation of the maintenance robot. The optimal function of speed, acceleration and jerk constraint is designed to make the tunnel inspection robot achieve efficient and stable speed control in the subway tunnel environment. In this paper, the "INS (inertial navigation system) + Odometer" positioning method is proposed. The difference between the displacement measured by the inertial navigation system and the displacement calculated by the odometer is taken as the measurement value, which reduces the dimension of the conventional algorithm. The closed-loop Kalman filter is used to establish the combined positioning model, and the system error can be corrected in real time with higher accuracy. The algorithms were verified on the test line. The displacement target was set to be 1 km and the limit speed was 60 km/h. The overshooting error of the speed tracking algorithm based on trapezoidal velocity planning and MPC was 0.89%, and the stability error was 0.32%. It improved the accuracy and stability of the speed following, and was much better than the PID speed tracking algorithm. At the speed of 40 km/h, the maximum positioning error of the robot within 2 km is 0.15%, and the average error is 0.08%. It is verified that the multi-sensor fusion positioning algorithm has significantly improved the accuracy compared with the single-odometer positioning algorithm, and can effectively make up for the position error caused by wheel-rail creep and sensor error.

Microscopic observation of two-level systems in a metallic glass model.

The low-temperature quasi-universal behavior of amorphous solids has been attributed to the existence of spatially localized tunneling defects found in the low-energy regions of the potential energy landscape. Computational models of glasses can be studied to elucidate the microscopic nature of these defects. Recent simulation work has demonstrated the means of generating stable glassy configurations for models that mimic metallic glasses using the swap Monte Carlo algorithm. Building on these studies, we present an extensive exploration of the glassy metabasins of the potential energy landscape of a variant of the most widely used model of metallic glasses. We carefully identify tunneling defects and reveal their depletion with increased glass stability. The density of tunneling defects near the experimental glass transition temperature appears to be in good agreement with experimental measurements.

Material Defects in Friction Stir Welding through Thermo–Mechanical Simulation: Dissimilar Materials with Tool Wear Consideration

Despite the remarkable capabilities of friction stir welding (FSW) in joining dissimilar materials, the numerical simulation of FSW is predominantly limited to the joining of similar materials. The material mixing and defects' prediction in FSW of dissimilar materials through numerical simulation have not been thoroughly studied. The role of progressive tool wear is another aspect of practical importance that has not received due consideration in numerical simulation. As such, we contribute to the body of knowledge with a numerical study of FSW of dissimilar materials in the context of defect prediction and tool wear. We numerically simulated material mixing and defects (surface and subsurface tunnel, exit hole, and flash formation) using a coupled Eulerian-Lagrangian approach. The model predictions are validated with the experimental results on FSW of the candidate pair AA6061 and AZ31B. The influence of tool wear on tool dimensions is experimentally investigated for several sets of tool rotations and traverse speeds and incorporated in the numerical simulation to predict the weld defects. The developed model successfully predicted subsurface tunnel defects, surface tunnels, excessive flash formations, and exit holes with a maximum deviation of 1.2 mm. The simulation revealed the substantial impact of the plate position, on either the advancing or retreating side, on the defect formation; for instance, when AZ31B was placed on the AS, the surface tunnel reached about 50% of the workpiece thickness. The numerical model successfully captured defect formation due to the wear-induced changes in tool dimensions, e.g., the pin length decreased up to 30% after welding at higher tool rotations and traverse speeds, leading to surface tunnel defects.

Inspection of Friction Stir Weld Joint on Metal Matrix Composite

The study examines the quality of friction stir welded joints made with butt joints on a metal matrix composite. In the composite, AA6061 is the matrix and silicon carbide is the reinforcement. Joints were made with three different pin profiles: circular, triangular and square. Macroscopic and microscopic images were used to qualitatively assess the joints. With the circular pin-profiled tool, the joints had distinctive onion ring patterns on their surface. Plasticized metal flows uniformly from the advancing side to the retreating side. While using triangular and square pin tools, weld spatters were formed along the retreating side of the joint. During the joining process, the plasticized material intermittently flowed. Inspection of the weld cross-section showed a major tunnel defect. The circular pin could not transfer materials that were not sufficiently plasticized. Plasticized material below the triangular pin tool was splashed by the tool pin edges. The bottom of the joint got enough weldment due to this. As a result, there were bind defects along the advancing and retreating sides of the joint. A square pin tool did not cause any defects in the weld joint. Because the tool pin had more edges, the weldment was deposited uniformly throughout the joint. Various zones of the joint showed grain restructuring. Using the circular pin profile tool, a noticeable plane separated thermo-mechanically and heat-affected zones. When using triangular and square pin tools, the three zones transitioned gradually. When using the square pin tool, the grain size was smaller along the stir zone and the thermo-mechanically affected zone. A square pin profile permitted greater frictional heat, which led to plasticization of the weldment and the pin. Based on comparisons with the other two counterparts, the friction stir welded joint made using the square pin tool had superior quality.

Effect of friction stir welding parameters on microstructure-based defect formation and mechanical properties of Tailor Welded Blanks

This maiden work on friction stir welding (FSW) of unequal thickness dissimilar (UETD) materials with plate thickness ratio (PTR) of 2.54 presents a comprehensive analysis of the effect of FSW process parameters on mechanical properties and microstructure-based defect formation of tailor welded blanks (TWBs). In contrast with equal thickness butt joints, the thickness mismatch in TWBs creates various challenges in the process requisite, influencing the process mechanics leading to inadequate heat input and improper mixing of base materials, making the joint prone to various defects. Defects such as tunnel defect, kissing bond defect, voids, and incomplete root penetration (IRP) defect observed and discussed comprehensively in this study are largely attributed to the complexities in the process requisites and mechanics, selection of unfavourable combinations of FSW process parameters, variation of plate thickness, and occurrence of surface oxide layers at the abutting surfaces. The present work included and discussed microstructural features around defects, temperature history, and mechanical properties of joints such as tensile strength and hardness to understand the effect of FSW process parameters on failure mechanism of TWBs in detail. Based on the results obtained from mechanical properties of joints discussed in this study, it was revealed that IRP defect deteriorated the mechanical properties the most followed by tunnel defect, kissing bond defect and voids. Figure: Illustrates (a) FSW set up (b) macrostructure of welded joint (c) tensile specimen and (d) specimen indicating positions of microhardness readings

Simultaneous tunnel defects and lining thickness identification based on multi-tasks deep neural network from ground penetrating radar images

The overall assessment of tunnel lining, including shapes, categories, and depths of tunnel internal defects as well as the thickness of tunnel linings is vital to the safe operation of tunnels. We proposed a method comprising a multi-task deep neural network and curve fitting post-processing operation for simultaneously identifying the shapes, categories, and depths of tunnel defects as well as lining thicknesses from ground penetrating radar (GPR) images. The multi-task deep neural network, denoted as M-YOLACT, was designed to identify defects, lining profiles, and hyperbola shapes simultaneously. We introduced a curve-fitting post-processing operation to calculate the dielectric constant automatically based on the hyperbola shapes and evaluated the defect depths and lining thicknesses. The method was validated by numerical simulations, sandbox, and field tests. The method effectively identified the shapes and classes of tunnel defects as well as the thickness profiles from GPR B-Scan images.

Convolutional networks and transformers for intelligent road tunnel investigations

Visual inspections do not provide a reliable and objective assessment of the conservation state of road tunnels. Although direct tests might represent a valid survey approach, they would often lead to prohibitive costs if performed extensively. Therefore, indirect techniques, such as ground-penetrating radar (GPR), have become fundamental to supporting limited direct tests. The analysis of the GPR tunnel linings profiles is mainly hand-operated. It permits the detection of various tunnel linings defects, characterizing a tunnel’s global health state. In the present work, the authors developed an artificial intelligence (AI) based automatic road tunnel defects hierarchical classification framework to improve the efficiency of this powerful indirect surveying method. Adopting the most recent tools in image processing provided by the deep learning (DL) community, the authors proposed a convolutional neural network (CNN) with the acknowledged ResNet-50 architecture, initialized through the transfer learning method. For the sake of comparisons, the authors also adopted the state-of-art convolutional EfficientNet architecture. To further improve the proposed framework, the authors investigated how the bidimensional Fourier transform applied as a preprocessing procedure could affect the classification performances of the ResNet-50 model. Finally, to further enhance the classification performance, the state-of-art neural vision transformer (ViT) architecture has been adopted with the transfer learning approach to the currently proposed defects classification framework.

A Novel Detection and Assessment Method for Operational Defects of Pipe Jacking Tunnel Based on 3D Longitudinal Deformation Curve: A Case Study.

Adjacent tunnel construction and environmental disturbances can lead to longitudinal deformation in pipe-jacking tunnels. The longitudinal deformation of the tunnel is closely related to the occurrence of joint dislocation, joint opening, and other defects. In view of the difficulty of obtaining 3D longitudinal deformation curves, a method is proposed to obtain 3D longitudinal deformation curves based on a large number of 3D point cloud data with high spatial resolution and large spatial dimensions. Combined with the mechanism of defects occurrence, a theoretical basis for tunnel defects assessment based on tunnel longitudinal deformation is proposed. Taking one pipe jacking tunnel as an example, the longitudinal settlement curve and the 3D longitudinal deformation curve are compared. The correlation between the 3D longitudinal deformation curve and defects such as mud leakage, cracks, and differential deformation is illustrated from the perspective of three indexes: deformation amount, bending deformation, and shearing deformation. The accuracy and reliability of the 3D longitudinal deformation curve in tunnel defects detection and assessment are verified.

Heat generation, plastic deformation and residual stresses in friction stir welding of aluminium alloy

The interactions among thermal history, plastic deformation and residual stresses in the friction stir welding (FSW) process under different welding parameters have been widely considered a crucial issue and still not fully understood. In the present study, a novel three-dimensional fully coupled thermo-mechanical finite element (FE) model based on Coupled Eulerian-Lagrangian approach (CEL) has been developed to simulate the FSW process of aluminium alloy AA 6082-T6 and to analyse the thermo-mechanical interaction mechanisms under different welding conditions. The numerical model successfully simulates the plunge, dwell, and welding steps in FSW and captures the evolution of temperature, plastic deformation, and residual stresses in the welded joint. The obtained results were validated by experimental testing with observed cross-weld thermal history, optical macrography and residual stress measurement using the neutron diffraction technique. The results reveal that the tool rotation speed governs the temperature evolution; the peak temperature increased from 740 to 850 °K when the tool rotation speed rose from 800 to 1100 rpm. The rotational speed also affected the plastic deformation, material flow, and the volume of material being stirred during the welding process. Higher plastic deformation is formed in the stirring zone by increasing the tool angular velocity. This behaviour led to an increase in the stirring effect of the welding tool, reduction of the tunnel defect size and enhancing the quality of weldments. The distribution of residual stresses in different zones of the FSW joints has been found to have an M-shaped profile. A significant tensile residual stress is characterised in the edge of the nugget zone in both longitudinal and transverse directions, balanced by compressive stresses in the thermo-mechanically affected zone, heat-affected zone and base metal. The presented FE modelling provides a reliable insight into the effects of the welding parameters on the weld quality of FSW joints and process optimisation with minimised experimental trials.

Influence of degradation disease on tunnel stability based on a time-dependent constitutive model

The surrounding rock and lining of tunnels deteriorate with increasing operation time, which causes changes in the mechanical parameters of the lining material and then affects the stability of tunnels. According to the time degradation law of the lining proposed in this paper, the mechanical parameters of lining materials corresponding to different operation years of the tunnel can be obtained. Model tests of the lining without degradation and degradation for 10, 30, and 50 years were carried out. Based on the realistic failure process analysis (RFPA2D) system, the time-dependent failure constitutive model is introduced to simulate the degradation of the surrounding rock and lining. Based on the model test and numerical analysis, the influence of degradation on the mechanics and failure features of tunnel were studied. It was found that the degradation would impact the structural stress and the stability of the surrounding rock. The failure mode of the model and the loading stress were different with the degradation time. The longer the degradation time, the more serious the deformation of the vault, and the larger the damage area of the surrounding rock. The research results play an important theoretical guiding role in analysing the progressive failure process of the surrounding rock and lining and have great value in engineering applications for reinforcing and treating tunnel defects.

Automatic Tunnel Crack Inspection Using an Efficient Mobile Imaging Module and a Lightweight CNN

Cracks in tunnel linings are the most common tunnel defects. As early indicators of structural deterioration, cracks represent critical problems for the safety of tunnels. Several mobile tunnel inspection systems (MTISs) have been developed for tunnel crack inspection. However, due to the weak signals of cracks, these MTISs require considerable exposure time to capture high-quality tunnel images, necessitating a low travel speed. Meanwhile, traditional crack detection methods encounter difficulties in processing tunnel crack images because of their low contrast and poor continuity. To overcome these challenges, this study presents a new MTIS for fast tunnel crack inspection that consists of a novel mobile imaging module and an automatic crack detection module. The imaging module is composed of an array of high-resolution charge-coupled device (CCD) cameras, a mobile laser scanner, and a lighting array. The core of the crack detection module is a novel lightweight convolutional neural network (CNN) designed for efficient tunnel crack detection, with an effective spatial constraint strategy to guarantee crack continuity. We collected a new tunnel crack dataset consisting of 1,218 images using our mobile imaging module at a driving speed of 80 km/h. Comprehensive experiments were conducted on this dataset to evaluate the performance of our proposed network. The results demonstrate that the presented CNN can effectively detect tunnel cracks with state-of-the-art performance, achieving an F1-score greater than 0.88 and an inference speed of 17 FPS with only 3.4M model parameters. The code and data are available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/urban-informatics/LinkCrack</uri> .

Defect monitoring in dissimilar friction stir welding of aluminum alloys using Coupled Eulerian-Lagrangian (CEL) finite element model

ABSTRACTThis article uses coupled Eulerian–Lagrangian finite element algorithm to conduct a three-dimensional thermomechanical study to capture the shape and characteristics of defect type generated while achieving the dissimilar friction stir welding of aluminium alloys. The volume-of-fluid method is used to model the Eulerian region and predict the localised formation of process defects. Three different tool shapes are utilised to achieve the dissimilar friction stir welding joining between AA 2024-T3 on the advancing side and AA 6061-T6 on the retreating side. Process parameter effects such as rotational tool speed, traverse tool speed and tool tilt angle are also investigated. The finite element model results are validated by comparing with the results of a previous experimental study. The results showed the augmentation of the traverse welding speed from 40 to 80 mm/min is a key factor in causing process imperfections such as void and tunnel defects. The lower tilt angle value of 1° resulted in long tunnel defects when high rotational speeds are applied. Also, the combination of high rotational and low transverse speeds promotes the production of a free-defect friction stir welding joint.

Exploratory Research on Drainage Structure of Highway Tunnel Based on Reducing the Risk of Crystallization Blockage

Crystal blockages of tunnel drainage systems severely undermine the tunnel lining structure and operation safety. In order to reduce the risk of crystal blockages of tunnel drainage systems, the distribution of highway tunnel defects was identified through a field survey, indoor test, and literature analysis, and an optimization method of tunnel drainage structures was proposed. The research suggested the following: (1) Lining water leakage and construction joint water leakage were the most common defects in the tunnel drainage system of Renhua–Xinfeng Expressway and Yingde–Huaiji Expressway in Guangdong Province, accounting for 60% and 32% of total defects, respectively. The number of defects that occurred in the drainage system of the tunnel was larger in the granite formation, with the number of road seepage and inspection chamber crystallization incidents reaching 2.5/km and 2.8/km, respectively. (2) The groundwater was mainly alkaline with a pH value of 8~12, Ca2+ (107 mg/L) was the cation with the largest ion concentration, and HCO3− (165 mg/L) was the anion with the largest ion concentration. The crystals in the tunnel drainage system were predominantly square, spindle, and rhombic calcite and aragonite composed of CaCO3, mixed with a small amount of sediment. (3) To reduce the risk of crystal blockages of the tunnel drainage system and ensure tunnel lining structure safety, a threefold optimization measure was proposed, namely, setting one-directional drainage pipes between the cable trench and the roadside blind drainage ditch, applying “π” type anti-crystallization drainage water-stop belts at the circular construction joints in the secondary lining, and both increasing the slope of the transverse drainage pipe and using an anti-crystallization drainage pipe. The research results will play an important role in guiding the design, construction, and maintenance of highway tunnel drainage systems in China.

Tunnel lining water leakage image Segmentation based on improved BlendMask

Water leakage of tunnel is a common disease in all kinds of existing tunnels. Automatic, timely, and accurate detection of water leakages is of great significance to the safe operation and maintenance for all kinds of tunnels. Due to the complicated background of tunnel surfaces, it is difficult for traditional tunnel defect detection algorithms to detect such defects quickly and accurately. To address these problems, this article presents an improved BlendMask image segmentation model with top-down and bottom-up architecture which accurately extract tunnel’s water leakage area from tunnel lining images. The proposed method is validated by an experimental study, and the results are compared with those obtained by the tunnel defect segmentation methods including Mask R-CNN and DeepCrack and other deep learning methods including CondInst, SoloV2, fully convolutional network and UNet. The Recall, Precision, F1-score, Dice and mean intersection over union (mIoU) for the proposed method are superior than those by the other methods above with respect to on test images. The subjective analysis on predicted results by different methods also shows the presented method can perform well.

Force data-driven machine learning for defects in friction stir welding

This study proposes a strategy for developing force-data-driven machine learning models to precisely predict defects and their types in friction stir welding (FSW). The characteristics of the three component forces in FSW, including traverse force (Fx), lateral force (Fy), and plunge force (FZ) are studied. The change in the force wave corresponded well with the variation in the defect. Fyavg had the best correlation with the characteristics of tunnel defects, whereas some other time-frequency features had negligible effects on the defect variation. The machine learning models built with the input of 15 force features could detect defects with an accuracy of 95.8% and classify them into tunnels and porosities with an accuracy of 98.0%. The abnormal increase in Fyavg, caused by the buildup of redundant material transported to the retreating side, was the main characteristic of force change when a defect was formed.