Gears generate various types of damage at different scales throughout the life cycle, making it challenging for existing methods to efficiently and accurately assess gear remanufacturability. This paper proposes a multi-scale defect detection method for gear tooth flanks, expanding the minimum detectable scale of surface defects to below 0.1 mm. Considering the imaging characteristics of involute tooth flanks, a multi-light enhanced image acquisition system is established. We constructed a tooth flank dataset containing 7,038 defect instances, introducing a unified sliding window and global non-maximum suppression approach to effectively address the bottleneck in detecting minor- and medium-scale defects while avoiding redundant detections. A high-quality tooth flank defect generation model is developed based on DDPM-ASTU-Net, which synthesizes defect samples according to the dataset characteristics to augment the training set, effectively mitigating the issue of imbalanced data distribution. We propose a DUD-YOLO detection network to target the typical defect distribution patterns in remanufacturing tooth flank surfaces. A lightweight dynamic feature fusion component is designed to enhance high-dimensional key feature extraction. A mixed local channel attention mechanism is merged to preserve crucial inter-channel correlation information while reducing computational complexity without sacrificing accuracy. Multiple convolutional optimizations significantly decrease the model parameters and inference latency. Extensive experiments validate the effectiveness and superiority of the proposed method. Compared to the baseline detection network, DUD-YOLO achieves simultaneous improvements of 5.4 %, 6.5 %, 8.3 %, and 7.2 % in precision, parameters, FLOPs, and inference speed, respectively. The proposed method provides an efficient and precise solution for gear remanufacturability assessment.

Current research on static defect detection for metal gear tooth surfaces has achieved substantial progress. To meet the iterative demands of industrial rapid inspection, Common challenges such as data sample scarcity, dynamic-static image matching, and annotation consistency need to be solved. This paper proposes a generative involute tooth surface data augmentation method that significantly improves network performance under real-time detection conditions. By establishing spatiotemporal dimensionality-enhancing mapping equations for rotating surfaces and designing grayscale correction mechanisms, we develop a 3D surface sub-pixel rotation simulation model (3D-SSRSM) for involute surfaces. This model performs a high-precision simulation of static images into rotating motion blur images. Furthermore, we propose a defect-density-controllable denoising diffusion implicit model (DDCDM) for text-guided generation of static images, effectively supporting training-set augmentation. Performance experiments using datasets from high-precision rotation acquisition systems validate the effectiveness of the proposed method, achieving an average improvement of 8.0% in the detection task and 37.6% in the semantic segmentation task over the best results of other methods. This method offers an efficient data augmentation solution for the rapid recognition of gear surface defects.

Accurate milling force model is crucial for process parameters optimization, machining quality control, and tool condition monitoring. Although prior studies have recognized the influence of tool geometry and dynamic effects on milling forces, research remains limited in modeling the coupling of multiple geometric features and dynamic effects, particularly for indexable end mills. Based on the micro element milling force model, this paper proposes an enhanced model which simultaneously incorporates tool geometric characteristics (including nose radius, axial mounting inclination, and the trochoidal trajectory of the tool tip) and dynamic effects (such as tool runout and regeneration effect). The classical micro element model is first modified by accounting for the impact of insert axial inclination and nose radius on the tool-workpiece contact behavior. A static-dynamic coupled instantaneous undeformed chip thickness (IUCT) model is then developed by incorporating the static IUCT model for different width-of-cut types, while dynamic displacements due to the regeneration effect are computed using the frequency domain response method. Furthermore, a tool runout parameters identification method combining the time and frequency domains is proposed. Finally, the proposed model is validated through a series of comparative experiments. The results demonstrate that the proposed model improves prediction accuracy of milling forces in the x, y, and z directions compared with the classical milling force model, and better captures the force-response variations associated with tool geometry and dynamic effects.

High-speed dry milling (HSDM) technology is one of the most attractive solutions for improving the cutting performance of difficult-to-machine materials. However, the study of HSDM performance for 30CrMnSiNi2A steel has not been reported. In this article, HSDM for 30CrMnSiNi2A steel is investigated using TiN-coated and AlTiN-coated tools, focusing on cutting force, vibration, surface roughness, residual stress, and tool wear. The effects of cutting parameters on cutting force and vibration are remarkably similar. At a spindle speed of 7500 r min-1, both cutting force and vibration are minimized. TiN-coated and AlTiN-coated tools exhibit distinct performance differences. But, HSDM of 30CrMnSiNi2A steel using TiN-coated and AlTiN-coated tools can achieve low surface roughness. The depth of cut significantly affects surface roughness, which is lowest at the depth of cut (0.8 mm). Most of the workpiece surface shows residual compressive stress. The main wear mechanisms for both coated tools are abrasive wear, adhesive wear, and oxidative wear. Furthermore, the AlTiN-coated tool is more wear resistant on the flank face compared to the TiN-coated tool. Chipping due to crater wear is the leading cause of tool failure for the AlTiN-coated tool. Upon various comparisons, the AlTiN-coated tool is more suitable for HSDM of 30CrMnSiNi2A steel.

High-speed dry milling process for high-strength steel is widely used in the aerospace field, which can achieve good machining quality and accuracy while reducing the use of cutting fluid. However, unsuitable milling parameters can cause excessive vibration of the machine tool and even chattering of the milling system, which in turn affects the machining process. Thus, it is necessary to find the milling parameters boundaries of high-speed dry milling for high-strength steel. In this paper, based on the nonlinear milling force model, the regenerative effect and modal coupling are taken into account to establish a three-dimensional dynamic model for high-speed dry milling of high-strength steel, and the stability lobe diagram (SLD) is obtained by using the compound Simpson integration method (CoSIM), which is proved to be accurate and effective by numerical and experimental verification. The method fits using additional known time points and their responses, resulting in more accurate SLDs. Numerical validation results demonstrate that the method has higher prediction accuracy and convergence speed while maintaining higher computational rates. The study conducts experimental verification through multiple sets of high-speed dry milling experiments, demonstrating that the anticipated SLD findings are largely congruent with experimental outcomes. Consequently, the model and solution approach demonstrate significant potential in the selection of milling parameters.

Cutting parameters significantly influence the surface roughness of high-speed dry milling (HSDM) of high-strength steel materials. Thus, accurate prediction of surface roughness and optimization of cutting parameters are critical for HSDM. This paper proposes a surface roughness prediction method based on Gaussian process regression (GPR), where cutting parameters and cutting forces are used as input variables. For 30CrMnSiNiA steel, HSDM experiments were carried out by considering spindle speed, feed per tooth, depth of cut, and width of cut as factors. Based on the experimental results, a comparison with support vector machine (SVM) and artificial neural network (ANN) is conducted to verify the effectiveness and superiority of the proposed prediction method with R 2 of 0.9818 and 0.9736 on the training and test sets, respectively. In addition, considering the cutting parameters optimization, a GPR-based cutting force feature model is established, and its superiority is verified by comparing it with the traditional exponential model. Then, a multi-objective cutting parameters optimization model with surface roughness and material removal rate as the objectives is built by combining the surface roughness model with the cutting force feature model, and the optimal cutting parameters are solved by the multi-objective Harris hawks optimization (MOHHO) algorithm. Compared with the optimal results obtained from experiments, the use of cutting parameters with high speed, large feed per tooth, small depth of cut, and small width of cut is an effective strategy to balance machining quality and efficiency.

Automatic manufacturing feature recognition (AFR) is a critical technology for realizing CAD/CAPP/CAM integration in the era of intelligent manufacturing. Despite the numerous feature recognition approaches that have been proposed, the recognition of intersecting features remains a challenge. The most important reason is that the boundaries of features and their geometric topological information are altered or destroyed when interacting with other features. To address this problem, this paper proposes a manufacturing feature recognition method based on graph and minimum non-intersection feature volume suppression. Firstly, the geometric and topological information of a part is extracted and represented as an attributed adjacency graph. Then, a subgraph isomorphism algorithm is designed to recognize the features corresponding to each subgraph. After that, a method to construct and suppress the minimum non-intersection volume of the recognized features is introduced to repair the boundaries of intersecting features. On this basis, intersecting features are separated and recognized at different stages. The experimental results indicate that the proposed approach is effective in recognizing intersection features. Moreover, the recognition result of the proposed method incorporates the surface and volume of a feature, providing enriched feature information for engineering applications.

A DOA-GRU-based surface roughness prediction method was developed to address the problems of poor generalization and low accuracy of traditional CNC milling surface roughness prediction models. The static data such as spindle speed and feed speed of CNC milling and dynamic data such as force signal were obtained; the DOA-GRU was obtained by optimizing the network structure parameters of the Gated Recurrent Unit (GRU) neural network using the Dingo Optimization Algorithm (DOA); the static data features were selected by Response Surface Methodology (RSM) and the dynamic data features were extracted by DOA-GRU adaptively, and then the dynamic and static data features were fused by the shallow neural network to build the surface roughness prediction model. The experiment results show that the root mean square error of the prediction model is 0.07 and 0.97 for the coefficient of determination R 2 . It indicates that the surface roughness prediction model based on DOA-GRU is effective.

High-speed dry grinding has the characteristics of high processing efficiency and clean environment. The high-speed dry grinding method meets the requirements of the green and efficient development of the national manufacturing industry. However, inappropriate cutting parameters seriously affect the surface quality of the workpiece and cause the workpiece to be scrapped. Therefore, this paper proposed an optimization method based on the combination of the improved extreme learning machine neural network (ELM) for high-speed dry milling surface roughness prediction model and genetic algorithm (GA). The Taguchi orthogonal experiment results show that the surface roughness of high-speed dry milling can be accurately predicted by the improved ELM and thereafter the optimal cutting parameter combination can be determined by GA.

Carbon fiber reinforced polymer (CFRP) composites have been widely used in high-tech industries due to excellent performances. Because subsequent machining operations, including drilling and trimming, et al., are necessary to meet dimensional or assembly-related requirement, accurate prediction of cutting forces is of great importance to improve tool life and machining quality, which is good for planning and improving machining process of CFRP composites. To explore this issue, the paper gives a detailed review and discussion of the cutting force models, including, mechanistic models; macro-mechanical models; micro-mechanical models and numerical models. Among them, macro-mechanical models are the earliest proposed cutting force models, while mechanistic models are the most studied. This paper predicts and analyzes the future development trend of cutting force models, including variable diversification development for semi-empirical models, study on process-oriented cutting force models, research on cutting mechanisms and study on intelligent manufacturing-oriented cutting force models based on modular development of numerical models.

The surface roughness of carbon fiber-reinforced polymer (CFRP) components is extremely important because of the increasing demand for higher performance, reliability, and longer lifetime in aerospace and other manufacturing industries. However, the cutting mechanisms of CFRP are still unclear, which limits its formation mechanism prediction for surface roughness. Thus far, this study presents three action mechanisms in the CFRP machining process: accumulation bouncing on the matrix, impact effect on carbon fibers, and workpiece selfaction. The workpiece self-action was reflected and calculated using the light rope model, which is new in CFRP machining. Furthermore, the formation mechanisms of surface roughness were first elucidated in the highspeed dry (HSD) milling of CFRP. An accurate surface roughness prediction model was theoretically formulated considering the kinematics, dynamics, and carbon fiber distribution. Surface roughness was expressed as the three-dimensional arithmetic mean height. The surface roughness prediction model was established and confirmed to have a high prediction accuracy of 90.05%, demonstrating that the distribution of carbon fibers was the main influencing factor of the surface roughness. Moreover, nonlinear regression analysis was used to clarify the effects of cutting parameters on the surface roughness, impact effect, and relationship between the surface roughness and impact effect. The study verified the feasibility of HSD milling CFRP, and it also provided guidance for breaking the low-speed machining limits by improving the machining process.

针对高速干切滚齿机床的能耗优化问题,对变工艺参数下的机床能耗分布特性及其预测模型进行了研究.首先,以机床能耗元件为研究对象,建立了高速干切滚齿机床功率模型;然后,基于机床电路分布原理,通过滚齿实验对所建模型进行了分析验证,揭示了变工艺参数与机床不同部位的功率和能耗之间的关系;最后,构建了高速干切滚齿机床能耗预测模型.

Carbon fiber-reinforced polymer (CFRP) composites have been widely used in the aerospace industry due to their excellent mechanical properties. Cutting mechanisms of machining CFRP and its effect on machined surface integrity are still unclear due to inhomogeneity and anisotropy. To this end, this paper studied the cutting mechanisms of CFRP by establishing an impact-based Specific Cutting Energy (SCE) distribution prediction model of high speed dry (HSD) milling CFRP which is be an effective and eco-friendly cutting method. SCE is divided into five sub-SCEs affected by shearing, pressing, impacting, bouncing and delamination, respectively. Among them, sub-SCEs generated by pressing and impacting are influenced by carbon fiber distribution due to size effect and material properties, carbon fiber distribution model was introduced into the SCE prediction model. The proposed model was verified with maximum relative error 7.6%, demonstrating that impact and size effect are of great significance in revealing the cutting mechanism of CFRP. To clarify the effect of SCE distribution on surface integrity, three-dimensional (3D) arithmetic mean height and 3D fractal dimension were used to quantitative characterization of the surface integrity and sub-SCEs are diagnosed for obtaining the way of sub-SCE distribution effecting surface integrity by correlation analysis. According to SCE diagnosis, the milling parameters are optimized by removing sub-SCEs unrelated to surface integrity, founding that HSD milling can reduce machining defects.

As the radius of carbon fibers and cutting edge are in the same order of magnitude, workpiece self-action, which could not be neglected in machining Carbon Fiber Reinforced Polymer (CFRP) considering size effect, has become a new perspective. Specific Cutting Energy (SCE) is a significant indicator for chip formation, cutting forces, tool wear, and machined surface integrity. Based on this, we presented a light springs model to clarify workpiece self-action to establish a specific cutting energy (SCE) prediction model of high speed dry (HSD) milling CFRP. The light springs model, which was also a direct reflection of the size effect, reflected the interaction of carbon fibers, interfaces and the matrix. The carbon fibers distribution in the chip was counted to confirm the light springs deformation at different position as carbon fiber and interface are consisted of the light spring. First of all, based on workpiece self-action and elastic-plastic theory, the cutting mechanisms were clarified to calculate SCE with high prediction accuracy (maximum relative error 7.7%). Furthermore, SCE distribution was obtained based on different cutting mechanisms, including bending, stretching, delamination and pressing bouncing. To improve machined surface integrity, three-dimensional (3D) arithmetic mean height and 3D fractal dimension were used for quantitative characterization of the surface integrity. The paper defined integrated evaluation indexes SCEDS and SCESa to reflect effective SCE to optimize milling parameters based on correlation analysis of SCE distribution and surface quality, founding that HSD milling operation could reduce machining defects.

在分析国内外机加工生产线监控系统的研究与应用基础上,提出了一种汽车壳体类零件加工生产线运行状态监控系统总体方案,对生产线数据采集方法、机床异常状态识别方法、质量预测方法等进行了研究,并在此基础上开发了一套汽车壳体类零件加工生产线运行状态监控系统.本系统有利于汽车壳体类零件加工生产线的转型升级,对汽车壳体类零件加工生产效率和加工质量的提升具有重要意义.

For process parameter optimization in high-speed dry hobbing, an optimization decision method is proposed in this study based on the multi-objective whale optimization algorithm (MOWOA). Firstly, according to the characteristics of high-speed dry hobbing, the processing time and processing error models are constructed; considering the carbon quota energy conservation and environmental protection policy, the processing cost model is established. Founded on the above models, the fitness function of the multi-objective optimization model is proposed. Afterward, on the basis of the traditional single-objective whale algorithm, the non-dominant set and crowding calculation method is introduced to establish a multi-objective optimization whale algorithm model. On this basis, the Pareto solution set is obtained. Finally, the actual decision case is compared to verify the effectiveness of the proposed method. Furthermore, the optimization data of MOWOA and several commonly multi-objective algorithms are compared to analyze the characteristics of the optimization solution set, thus verifying the superiority of the process parameter optimization method based on MOWOA.

针对柔性作业车间动态调度问题构建以平均延期惩罚、能耗、偏差度为目标的动态调度优化模型,提出一种基于深度Q学习神经网络的量子遗传算法.首先搭建基于动态事件扰动和周期性重调度的学习环境,利用深度Q学习神经网络算法,建立环境行为评价神经网络模型作为优化模型的适应度函数.然后利用改进的量子遗传算法求解动态调度优化模型.该算法设计了基于工序编码和设备编码的多层编码解码方案;制定了基于适应度的动态调整旋转角策略,提高了种群的收敛速度;结合基于T ent映射的混沌搜索算法,以跳出局部最优解.最后通过测试算例验证了环境-行为评价神经网络模型的鲁棒性和对环境的适应性,以及优化算法的有效性.

The geometric error of high-speed dry cutting gear hobbing is affected by many factors, such as process parameters, gear hobbing machine tools and hobs, and it is difficult to predict. This paper uses the vibration of the hob spindle to predict the geometric error, and builds a geometric error prediction model for high-speed dry cutting gear hobbing based on multiple population genetic algorithms and BP neural network. This model comprehensively considers the influence of the hob spindle speed, feed rate and hob spindle vibration on the geometric errors of high-speed dry cutting gear hobbing, which can provide a useful reference for the improvement of gear machining geometric accuracy. Through the experiment, it is concluded that adding the hob spindle vibration during the machining process as the model input can obtain more accurate prediction results of the geometric error of the hobbing machining. Compared with the traditional BP neural network prediction model, this model has better prediction ability.

Dry hobbing has received extensive attention for its environmentally friendly processing pattern.Due to the absence of lubricants,hobbing process is highly dependent on process parameters combination since using unreason-able parameters tends to affect the machining performance.Besides,the consideration of tool life is frequently ignored in gear hobbing.Thus,to settle the above issues,a multi-objective parameters decision approach considering tool life is developed.Firstly,detailed quantitative analysis between process parameters and hobbing performance,i.e.,machining time,production cost and tool life is introduced.Secondly,a multi-objective parameters decision-making model is constructed in search for optimum cutting parameters (cutting velocity v,axial feed rate fa) and hob parameters (hob diameter d0,threads z0).Thirdly,a novel algorithm named multi-objective multi-verse optimizer(MOMVO) is utilized to solve the presented model.A case study is exhibited to show the feasibility and reliability of the proposed approach.The results reveal that (i) a balance can be achieved among machining time,production cost and tool life via appropriate process parameters determi-nation;(ii) optimizing cutting parameters and hob param-eters simultaneously contributes to optimal objectives;(iii)considering tool life provides usage precautions support and process parameters guidance for practical machining.

Optimum process parameters play an important role in improving manufacturing process which have a vital influence on the energy consumption and production cost. Considering the fact that hobbing process is sensitive to process parameters, an integrated multi-objective process parameters optimization method for gear hobbing is proposed to reduce energy consumption and production cost. Thus, this paper firstly analyzes the hobbing process parameters and establishes a description of hobbing process parameters problem. Then a multi-objective optimization model of hobbing process parameters is introduced, with energy consumption and production cost to be optimized. An improved multi-objective ant lion optimizer (IMOALO) is designed to solve multi-objective optimization problem. Finally, a case study is presented in detail to verify the optimization model. The results show that energy consumption and production cost can be optimized simultaneously by determining appropriate process parameters based on proposed method. It has potential in providing favorable support and assistance for technical operators in the practical parametric decision.