This study presents a mechanistic force model for predicting thrust force and torque during the drilling of Zr-based bulk metallic glasses (BMGs), known for their high hardness and poor machinability. The proposed model discretizes the twist drill into elementary cutting tools along the lips and treats the chisel edge separately to account for its distinct plowing-dominated behavior. Cutting force coefficients for both regions were identified using a nonlinear optimization routine based on experimental drilling data, avoiding the need for orthogonal cutting tests. The model was validated across twelve different cutting conditions, demonstrating its ability to capture nonlinear effects of feed rate and spindle speed. The average prediction errors were 5.5% for thrust force and 8.8% for torque. These results confirm the model's robustness and suitability for predictive simulation and force monitoring in precision drilling of hard-to-machine amorphous materials like BMGs.
During the cutting machining of nickel-titanium shape memory alloy (NiTi-SMA), unreasonable machining parameters will lead to severe tool wear and poor surface quality. This article establishes the prediction model relating cutting speed (vc), feed rate (f) and cutting depth (ap) to surface roughness (Ra), material removal rate (MRR) and tool wear (VB) based on the Genetic algorithm-optimized BP neural network (GA-BP) by designing full-factor cutting experiments. Experimental findings indicate that the model's fitting degree is 98.57%, and the maximum relative error is 6.2%. On this basis, the NSGA-II is used for multi-objective optimization and Pareto optimal solution search. The EWM-GRA-TOPSIS method is used to calculate the similarity between the randomly selected 30 sets of Pareto optimal solutions and the ideal solution and sort. The optimal cutting parameters were finally determined to be vc = 79 m/min, f = 0.06 mm/r and ap = 0.12 mm. Using the optimized machining parameters for cutting experiments, the Ra value obtained at a processing length of 40 m is 0.459 mu m, the VB is only 0.049 mm and the MRR reaches 9.48 mm & sup3;/s. The maximum relative errors between the experimental and predicted Ra, VB and MRR of the optimal parameters are 2.7%, 4.1% and 0.6%, respectively.
Post-processing machining of additively manufactured carbon fiber-reinforced polymer (AM-CFRP) composites is essential for achieving dimensional accuracy and functional surface quality; however, their heterogeneous, layer-wise architecture poses significant challenges during cutting. In this study, a three-dimensional finite element (FE) framework is developed to investigate the micro-slot milling behavior of AM-CFRP (Onyx (R)) with varying infill architectures, raster orientations, layer thicknesses and depths of cut. The AM-CFRP material is modeled using a homogenized anisotropic continuum formulation, enabling computationally efficient simulation of cutting forces, stress evolution, temperature rise, chip formation, burr generation and machining-induced damage. Numerical predictions reveal that infill architecture and raster orientation strongly govern machinability. Perforated structures exhibit reduced cutting-force amplitudes compared to solid counterparts due to lower effective material engagement, while border reinforcement improves confinement and mitigates interlayer decohesion and delamination. Alternating raster orientations promote higher stress concentrations, increased burr formation and degraded surface integrity relative to more constrained layouts. Chip formation is dominated by fragmented and powder-like debris, consistent with polymer-matrix shearing and interface-driven fracture mechanisms characteristic of CFRP machining. Temperature evolution, driven primarily by adiabatic plastic deformation, shows pronounced sensitivity to infill topology, with honeycomb and grid structures exhibiting higher localized thermal response than rectilinear perforated patterns. The predicted cutting forces and dominant failure modes compare favorably with available experimental and numerical studies, demonstrating that the proposed FE framework captures the correct order of magnitude and relative trends. Overall, this work provides mechanistic insight into the machining behavior of AM-CFRP and establishes a robust numerical tool for comparative machinability assessment and process optimization of additively manufactured composite structures.
This study focuses on the thermomechanical field modeling and microstructural evolution during cutting AISI9310, which is a high-performance steel widely used in critical gear and shaft applications. At first, the analytical models for predicting mechanical and thermal loads are developed. The mechanical model with a prediction error of 15% for cutting force can be used to calculate the distribution of all the stress components; the thermal model has a prediction error of 20% for temperature. By making use of the mechanical and thermal models, the thermomechanical field beneath the machined surface during the quasi-orthogonal turning of AISI9310 is calculated. Then EBSD and TEM examinations are conducted on the samples taken from the 0 to 40 mu m depth range beneath the machined surface for the characterization of the microstructural evolution. EBSD examinations show that high-density low-angle grain boundaries exist within the 20 & micro;m-depth range, and TEM examinations indicate that high-density dislocations accumulate within the grains as well as at the grain boundaries of this region. Further analytical calculation based on the mechanical and thermal loading models shows that the area within the 20 & micro;m-depth range is characterized by a high level of normal stresses and shear stress, which is higher than the macroscopic yield point as well as the Critical Resolved Shear Stress (CRSS), together with a temperature distribution varying from 70 to 350 degrees C. Such a thermomechanical field is apt to promote the dislocation multiplication and movement.
In metal cutting operations, the precise and swift assessment of the tool wear state is crucial for developing tool replacement strategies. Therefore, a method for recognizing tool wear state based on Inception-ResNeSt-BiGRU is proposed. This method uses a combination of force signals and vibration signals to complement the signal information during processing. First, Inception is introduced to perform multi-scale feature extraction, allowing the network to learn richer feature information. Second, ResNeSt is introduced to improve the recognition accuracy in fuzzy regions. The Split-attention mechanism in ResNeSt can selectively focus on important feature information, improving the ability to process information. Third, BiGRU is implemented to capture bidirectional dependencies within deep feature sequences, enhancing the model's capacity to extract intricate patterns. Ultimately, the model utilizes the signal following wavelet denoising as its input to determine the state of tool wear. The average F1-scores for the initial, normal and severe wear stages of the tool are 0.958, 0.984 and 0.991, respectively, for the proposed method, with an average accuracy of 0.984. This method not only offers technical assistance in resolving the wear state recognition issue but also serves as a reference for the development of scientific and rational tool replacement strategies.
This study investigates the influence of tool path and milling parameters on the surface quality and dimensional accuracy of thin-walled structured thermoplastic polymer acrylonitrile butadiene styrene (ABS). Tests examined the effects of up-milling and down-milling on channels of two depths (10 and 20 mm) using three tool path strategies (zig-single direction, zigzag-double direction and follow periphery-peripheral cutting), five cutting speeds (25 to 125 m/min) and three feed rates (0.05, 0.1 and 0.2 mm/rev). The milling operation was carried out with an uncoated tungsten carbide end mill. Performance was evaluated through cutting forces (full and half immersion), surface roughness (base and side walls) and dimensional accuracy (channel width and 1 mm thin-wall deviations). A multi-objective optimization based on the chip removal rate (CRR), width deviation, wall deviations (left and right) and corresponding surface roughness values was conducted in order to determine the optimum cutting parameters. The results demonstrate that the Zig-Zag tool path combined with a cutting speed of 125 m/min yielded the optimal performance. Moreover, the optimization analysis revealed that the ideal feed rates for these conditions fall within the range of 0.05-0.09 mm/rev.
Machining-induced residual stress significantly affects the fatigue life, geometric stability and service performance of machined components. This study presents a fully coupled thermo-mechanical two-dimensional (2D) Finite Element (FE) model to investigate residual stress distributions induced by High-Speed Milling (HSM) of TC4 titanium alloy. The model's accuracy was validated by comparing the simulated outcomes with experimental data on chip segmentation, cutting forces and temperatures, with relative errors ranging from 6.85% to 12.40%. Beyond model verification, periodic fluctuations of plastic strain and residual stress in the machined surface and subsurface were simulated. Results showed tensile stress dominance at lower cutting speed (150 m/min), while compressive stress prevailed at higher speed (450 m/min), particularly for sigma 11. Subsurface stress profiles exhibited hook-shaped distributions, with compressive stress peaking at depths of 7-10 mu m and diminishing beyond 22 mu m. A good correlation was observed between experimental and simulated results, with subsurface stress consistently compressive and increasing with cutting parameters. The effect of thermo-mechanical loads plays a critical role in shaping residual stress distributions; thermal loads primarily influence surface stress, whereas mechanical loads govern subsurface stress. This research provides a comprehensive understanding of residual stress behavior and its impact on fatigue life and service performance.
The GH4169 nickel-based superalloy, renowned for its exceptional mechanical properties, presents significant challenges in machining due to its inherent hardness and thermal resistance. This study investigates the application of a hybrid nano-lubricant coolant mixture, combining multiwalled carbon nanotube nanofluid-assisted minimum quantity lubrication (MWCNT-MQL) with cryogenic liquid nitrogen (LN2), and sprays it through a single nozzle to improve the machinability of GH4169. A computational fluid dynamics (CFD) analysis of the multiphase, mixed-lubricated coolant spray under turbulent flow conditions was performed using the discrete phase model (DPM) in ANSYS Fluent software. Additionally, to enable precise mixing of the dual-phase coolant, a custom-designed mixing chamber was fabricated. Droplet size distribution was conducted using spray footprint analysis, and end-milling experiments evaluated cutting force, surface hardness, roughness, and tool wear under varying spray pressures (2-6 bar) and flow rates (40-60ml/h). The results revealed that the spray with the mean average droplet size (16.76 mu m at 4 bar and 50ml/h) yielded the most favorable outcomes. This resulted in a cooler workpiece surface, leading to increased surface hardness and cutting force. Additionally, a significant reduction in surface roughness and tool wear was observed. Compared to dry machining, hybrid lubri-cooling enhanced machinability and extended tool life by up to 43.16%, as confirmed by milling experiments and wear analysis.
In single-point diamond turning (SPDT), tool center error (TCE) frequently produces conical or cylindrical residues at the workpiece center, significantly influencing the machined surface quality. Although force-sensing-based online identification methods have been developed, the effectiveness is constrained by variations in cutting mechanisms across different materials. This study systematically investigates the influence of different plastic metal materials on TCE identification. Finite element analysis (FEA) simulations were performed in ABAQUS to evaluate the surface morphology and mechanical responses of various materials under tool-above-center and tool-below-center errors. The results indicate that materials with higher Young's modulus generate larger cutting forces and alter the surface stress distribution. Moreover, a comparison between simulations and experiments demonstrates that the proposed simulation model for TCE in ultra-precision machining (UPM) achieves high consistency with the measured values, with a maximum absolute error of 1.20 mu m. In addition, considering the influence of material properties on TCE, a novel identification model was developed in this study. Compared with previous approaches, the proposed model improves identification accuracy by 5-10%. These findings demonstrate that the proposed extended online identification method is both accurate and reliable, providing a robust foundation for advancing UPM toward higher levels of automation and intelligence.
Near-dry electrical discharge milling (Near-dry EDMilling) technology is a variant of Near-dry electrical discharge machining (Near-dry EDM) technology, and it also employs a gas-liquid atomized medium as the discharge medium for milling. This study adopted green argon as the gas-phase medium Near-dry EDMilling of titanium alloy, conducting comparative experiments with air-medium Near-dry EDMilling (Air-N-EDMiling). Results showed Air-N-EDMiling formed high-melting titanium oxides on the surface, while argon-medium Near-dry EDMilling (Ar-N-EDMiling) reduced surface compounds. Compared to Air-N-EDMiling, Ar-N-EDMiling increased material removal rate (MRR) by 22.5%, decreased surface roughness (Ra) by 18.1%, thinned recast layer by 64.81% and reduced microcracks. Orthogonal experiments were analyzed via signal-to-noise ratio (S/N) and variance methods to explore Ar-N-EDMiling parameters' effects on MRR, Ra and relative electrode wear rate (REWR), determining optimal single-objective parameters. Multi-objective optimization using grey correlation degree identified the best combination: current 15 A, pulse width 300 mu s, pulse interval 100 mu s, rotation speed 400 rpm.
In the machining process, Inconel 718 alloy is typically subjected to various complex factors such as mechanical forces, thermal forces, electromagnetic forces and chemical reactions. The impact of machining process parameters on aspects such as workpiece surface morphology, cutting forces, tool wear and microstructure is intricate and interconnected. Therefore, understanding the influence of the machining process on the mechanical properties of Inconel 718 alloy is a complex issue that requires consideration of multiple factors. This study aims to improve our knowledge regarding how grain size evolution affects the mechanical properties of Inconel 718 alloy. Room temperature tensile tests were conducted on Inconel 718 alloy samples machined under different cutting speed conditions to investigate the effects of grain size evolution and thermo-mechanical loading on mechanical properties. The research findings indicate that as the grain size gradually decreases, the plastic deformation capability of Inconel 718 alloy improves, leading to better toughness and ductility. Furthermore, the yield strength and tensile strength of Inconel 718 alloy increase as the grain size decreases, with dislocation strengthening contributing predominantly to the yield strength.
Surface roughness is a key indicator of machining quality, directly affecting product reliability and service life. Existing deep learning methods for roughness prediction mainly rely on convolutional neural networks combined with attention mechanisms. However, conventional attention structures often exhibit limited generalization and insufficient predictive accuracy, particularly due to their weak ability to effectively integrate cutting parameters with sensor-derived vibration signals. To address these limitations, this study proposes a two-stage prediction framework based on multi-source information fusion and a dual-attention mechanism. Unlike methods employing a single one-dimensional attention structure, the proposed framework jointly models spatial-channel and temporal dependencies to achieve more comprehensive feature extraction. In the first stage, vibration features are captured by a convolutional neural network, and preliminary predictions are generated using a dual-attention module that integrates channel-spatial attention with sparse multi-head temporal attention, supported by a bidirectional long short-term memory network. In the second stage, the preliminary predictions are fused with cutting parameters, and a one-dimensional convolutional neural network produces the final roughness prediction. Experiments conducted on two datasets demonstrate that the proposed method reduces RMSE by 18%, decreases MAPE by 20%, and improves R2 by 0.05 compared with alternative approaches, confirming its superior accuracy and generalization performance.
This study significantly improved the microstructure and mechanical properties of the surface strengthening layer on TC4 titanium alloy by introducing composite powder working media of C + B4C and C + B4C + Al into near-dry EDM (PMND-EDM). Experimental results indicate that under the C + B4C condition, the reinforcement layer mainly forms phases such as TiC, TiB, Ti, Al3Ti and AlTi3; whereas after adding Al powder (C + B4C + Al), it promotes the reaction between the matrix and the medium, generating new reinforcing phases such as Al5Ti, Al3BC and B8C. The introduction of aluminum powder provides a molten environment, significantly reducing porosity and splatter droplets and decreasing the density and width of microcracks. Elemental distribution shows that B, C and Al elements are enriched in the transition zone at the edge of the discharge pit, while the central region is mainly composed of Ti. When processing with C + B4C + Al mixed powder at a peak current of 8.2 A, the microhardness of the reinforcement layer reaches approximately 1300 HV, which is more than four times the hardness of the matrix, attributed to the synergistic strengthening mechanism of multiple phases. Meanwhile, the material removal rate (MRR) is improved after adding Al powder. This work demonstrates the superior potential of multi-component powder design in tailoring surface integrity and mechanical properties of difficult-to-machine alloys.
Fiber metal laminates (FML) are widely used in aerospace and automative industries and often require machining, like drilling and milling to obtain desired geometric features or assembly. The present work encompasses investigation on cutting mechanics, damage and performance analysis of fabricated carbon fiber reinforced aluminum laminates (CARALLs) during drilling at different process parameters to improve machining efficiency and minimize structural damage. Drilling experiments are conducted at different feed rates and spindle speeds to assess the influence on thrust force, hole quality, and damage progression. A detailed meso-scale FE modeling framework is developed, incorporating traction-separation laws to capture interlaminar delamination and interface damage. Aluminum layers are modeled for ductile damage, while the Hashin 3D damage model is applied to carbon fiber layers to predict fiber breakage, matrix cracking, and fiber-matrix failures. Results indicate that thrust force increases by approximately 55-65% with feed rate (0.05 to 0.2 mm/rev) and decreases by about 30-35% with spindle speed (1000-3000 rpm). Thrust force variations exhibit alternate peaks due to drill interaction with the layers of aluminum and carbon fiber prepreg present in the laminate, and the same is also identified by finite element (FE) simulation based on tool-work interaction. Different damage modes such as intralaminar matrix damage, fiber failure and matrix loss at the interface of the aluminum and CFRP layers, are identified by scanning electron microscopy. The machining performance is appraised in terms of hole diameter variations, hole quality, and burr height after drilling of the CARALLs at various cutting conditions. Higher feed rates and rotation speed showed the best results.
