Tool wear monitoring is essential for ensuring machining efficiency and product quality, particularly for difficult-to-machine materials such as Inconel 718 (IN718). Traditional deep learning models, such as Conventional Convolutional Neural Networks (CNNs), often struggle to capture complex wear patterns and lack accuracy across varying machining conditions while developing image-based tool wear identification systems. To address these limitations, this paper presents a Vision Transformer (ViT) model for identifying tool-wear categories during end-milling of IN718. The performance of the ViT-based model is systematically compared with a CNN-based EfficientNet-b0 model. The robustness and generalization of the ViT-based model are validated on two previously unseen image datasets: one with conditions similar to those of the training data and another acquired under varying lighting conditions. The results indicate that the ViT model outperforms the EfficientNet-b0 model in terms of classification accuracy and computational efficiency. The ViT model achieves higher accuracy with fewer training epochs and faster convergence. Furthermore, it exhibits strong generalization across different lighting conditions, demonstrating robustness to variations in the machining environment. The findings presented in this work clearly demonstrate ViT’s effectiveness in tool wear classification and its potential as a reliable, efficient algorithm for developing tool wear monitoring systems for practical machining applications.

The concurrently existing wear mechanisms during the machining of hard materials necessitate conducting numerous experiments to generate a comprehensive dataset of worn cutting tool images while developing on-machine tool wear monitoring systems. The higher costs and time associated with image dataset preparation limit the number of experiments that can be conducted during the model development, resulting in limited exposure to practical machining conditions. This paper presents a systematic approach for generating diverse images of worn cutting tools by selectively applying image augmentation techniques. The augmentation techniques are critically evaluated to effectively capture the variabilities during image acquisition in a real-time machining environment. The techniques were chosen to simulate the presence of coolant, dust on the camera lens, lighting variations, differences in tool size and shape, and faulty camera setup. The initial dataset of 200 images labeled across four tool wear categories was generated by conducting machining experiments. A larger and balanced dataset of 28,000 images was created by applying these augmentation techniques to 80

Automotive manufacturing plants are increasingly making use of data collected in real-time to understand the state of their operations. Examples of such efforts to date have included direct monitoring of process parameters, location tracking within manufacturing facilities, and wear monitoring for failure prediction and predictive maintenance. However, consideration of real-time monitoring of human assembly workers' condition - particularly with respect to the workload they bear - is notably lacking. This contributes to negative outcomes for individual workers and manufacturing organizations alike. To build a foundation for experimentation and validation concerning real-time estimation techniques, this work first motivates the investigation of workload monitoring through consideration of the present state of assembly workers in the automotive industry, describing workload estimation and the shortcomings of existing techniques for such an application. A method for manipulating the workload associated with a manual assembly task is then proposed, and an experiment to assess its efficacy is described. Preliminary results from experimentation are presented, with only manipulations of temporal demand yielding significant differences in reported values of its corresponding workload sub-scale. Modifications to the manipulations of other scales are considered, before a brief discussion of these results' implications for enabling future work to develop and validate robust, unobtrusive methods for workload estimation is offered, and directions for further investigation are described.

Implementing tool wear monitoring approaches is often challenging due to the requirements of directly observing the wear state or integrating sensor-based instrumentation. This work proposes identifying wear stages of a turning tool by analyzing the machined surface quality. An indirect tool wear classification approach is presented to categorize tool wear into three classes-initial wear, steady wear stages, and catastrophic wear during the turning operation. The machined surface images were captured over diverse process parameters to realize labeled datasets for these three wear classes. A pre-trained Convolutional Neural Network (CNN), EfficientNet-b0, was fine-tuned using transfer learning to classify the surface images and predict tool wear stages subsequently. The proposed approach demonstrated the potential to offer an alternative solution to on-machine tool wear monitoring. Although the primary results showed the utility of the proposed approach in predicting tool wear stages, the analysis of misclassifications using confidence scores and heatmaps revealed some discrepancies. It highlighted the need for further research to enhance surface image features that can realize a robust and reliable indirect tool wear classification model. (c) 2025 The Authors. Published by ELSEVIER Ltd. This is an open-access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the NAMRI/SME.

Machine Vision (MV) systems integrated with Artificial Intelligence (AI) based technologies like Convolutional Neural Network (CNN) have shown promising results in classification and sorting applications. One such application is the classification of threaded fasteners like screws, bolts, nuts, etc, during the assembly process to ensure secured assembly of the components. However, increasing penetration of the internet into manufacturing organizations introduces cybersecurity concerns that can impact the performance of AI-based MV systems. Cyberattacks such as data poisoning, and model evasion targeted towards AI systems can manipulate the input data or the model itself, leading to misclassification and potential operational failures. With an increase in MV systems being deployed in automated manufacturing, it is essential to safeguard the AI models against cyberattacks. Hence this research is motivated by the need to analyze the impact of one such class of cyberattacks, i.e. model evasion attack on an AI-based MV system. The focus of this research will be to analyze the impact of this attack on the accuracy and confidence of the model as applied to fastener classification. Model evasion attacks can compromise a model's ability to classify images correctly, resulting in the model's reduced confidence levels and/or inaccurate image classification. This paper utilizes the Fast Gradient Sign Method (FGSM) to craft adversarial images of threaded fasteners, which are then used to perform model evasion attacks on the MV system for fastener classification. FGSM is particularly effective in white-box attacks, where the attacker has full knowledge of the model's architecture and parameters. FGSM generates adversarial images of threaded fasteners quickly and efficiently by introducing subtle and nearly imperceptible modifications to the input images. It leverages the gradients of the model's loss function to identify the optimal direction and magnitude of perturbations, ultimately deceiving the model into making incorrect predictions. The research questions answered in this study are: a) How does FGSM-generated adversarial images impact the accuracy and confidence of the classification results from the developed Efficient-Net-b0 based MV system? b) Is the XAI system implemented as part of the MV system capable of detecting FGSM-generated adversarial attacks? c) Can human-in-the-loop be a solution to mitigate the impact of FGSM-generated adversarial images? The study shows that FGSM-generated adversarial images reduced the accuracy and confidence of the classification results. The XAI system detected adversarial images by displaying skewed heat maps that highlight the image regions that were the most influential in contributing to the model's decision. Lastly, when humans are presented with FGSM-generated adversarial images, they are still able to identify the images with high accuracy and confidence. In conclusion, the paper evaluates the impact of a cyberattack on an AI-based MV system used in manufacturing, and also validates a potential solution to mitigate the impact of such an attack.

The paper systematically realizes a vision-based on-machine Tool Wear Monitoring (TWM) system for integration with a CNC milling machine to identify tool wear states during machining hard materials such as Inconel 718 (IN718). The proposed TWM system consists of a microscope-based image acquisition setup mounted inside the machine and pre-defined programmed motions to capture high-resolution images of worn side cutting edges. The pre-trained Convolutional Neural Network (CNN) model, Efficient-Net-b0, was developed using transfer learning to identify tool wear states utilizing labeled image datasets generated in the machining environment. The labeled datasets were generated systematically by intermittently capturing images during IN718 machining at varying surface speeds. The present study considered four tool wear states, Flank, Flank+BUE, Flank+Face, and Chipping, representing combinations of abrasion, adhesion, diffusion, and fracture wear mechanisms. The effectiveness of the proposed TWM system was evaluated by identifying the wear state for previously unseen test datasets. The results showed that the TWM system can identify tool wear states with an accuracy of 94.11%. Furthermore, the study analyzes reasons for misclassifications using feature maps and classification probability scores to achieve better prediction abilities.

Electrically assisted machining involves the application of an electric current through the workpiece-tool interface to enhance machining performance. Literature suggests that applying electric current during machining can reduce machining forces, improve surface finish, and increase efficiency, potentially making the process more effective and economical. These positive effects are often attributed to the electro-plastic effect, where the electric current allegedly induces plasticity in the material and describes the non-thermal effect of the electric current. In the present work, electrically assisted milling was performed with the application of pulsed electric current while the cutting tool engaged with the material. Important parameters such as current amplitude and machining forces were recorded to investigate the effect of electrical and mechanical factors during the process. Additionally, numerical simulations were conducted to estimate the current density distribution and the temperature profile within the workpiece. The findings contradict previous literature assumptions, revealing that the electroplastic effect may not be responsible for the observed improvements. Instead, it appears that any reductions in force or enhancements in surface quality are more likely due to localized heating in the tool rather than changes in the material properties of the workpiece itself. This insight suggests a shift in understanding electrically assisted machining mechanisms, emphasizing the role of thermal effects at the tool rather than electro plastic changes in the workpiece.

Vision-based tool wear monitoring systems augmented with Artificial Intelligence (AI)-based algorithms can effectively identify tool wear states. However, inconsistent image quality due to varying lighting conditions on manufacturing shop floors often obscures the scalability and reliability of these systems for practical applications. This study presents an on-machine vision-based tool wear monitoring system capable of handling varying lighting conditions using human guidance and an eXplainable AI (XAI) approach. The present study captured tool wear images under two lighting conditions, L1 and L2, using a microscope-based on-machine image acquisition system. The images were classified into four tool wear states: Flank, Flank + BUE, Flank + Face, and Chipping, commonly observed while machining Inconel 718 (IN718). Tool wear images captured under the L1 lighting condition were used to train the Convolutional Neural Network-based Efficient-Net-b0 model. The model was integrated subsequently with the Human Guided-XAI (HG-XAI) approach to predict tool wear states for images captured under the L2 lighting condition. The performance of the HG-XAI approach was evaluated using metrics such as Accuracy, Matthews Correlation Coefficient (MCC), and F1-Score and compared with the standalone Efficient-Net-b0 model. The results show that the HG-XAI approach achieved an accuracy of 96%, MCC of 0.96, and F1-Score of 0.97, demonstrating significant improvements over the standalone Efficient-Net-b0 model. The findings of this paper substantiate the scalability and reliability of the integrated HG-XAI approach under varying lighting conditions.

Machining of Nickel-Based Superalloys (NBSAs) is characterized by rapid work hardening, low thermal conductivity, and extreme abrasive behavior. Due to such characteristics of NBSAs, the machining of materials like Inconel 718 (IN718) results in extensive tool wear and high cutting forces. The present study employs nanobubble-based cutting fluid during machining of IN718 and investigates its effect on tool wear and cutting forces. The trochoidal milling experiments were conducted on an IN718 workpiece with nanobubble-based and normal cutting fluid at varying surface speeds. The evolution of tool flank wear area and peak resultant cutting force were recorded and compared to evaluate the effectiveness of nanobubble-based cutting fluid over normal cutting fluid. The results showed that nanobubble-based cutting fluid lowers the tool wear rate by enhancing heat dissipation and reducing abrasion wear. However, higher cutting forces were observed with nanobubble-based cutting fluid due to increase in hardness of the workpiece. Also, it is realized that using nanobubble-based cutting fluid can significantly improve the lifespan of the tool, leading to sustainable manufacturing. (c) 2024 The Authors. Published by ELSEVIER Ltd.

The machining of free-form components by ball-end milling inherently produces surface error in the form of scallops. The objective of any free-form toolpath strategy is to balance productivity while minimizing scallop height to reduce surface error. Conventional machining strategies produce repeatable material patterns (constant scallop height) that may limit workpiece function in areas such as lubricity, directional anisotropy, and aesthetic appearance. These strategies also involve steady-state cuts, which allow accumulation of temperature, restrict the permissible depth and speed. In the present paper a novel complex stochastic toolpath strategy has been proposed that comprises pseudo-random circular contours concatenated into a smooth path. The approach enables continuous variation of chip load, force, and direction, and avoids conditions of continuous, periodic high cutting loads and heat accumulation. Based on initial testing, it has been observed that stochastic toolpaths are longer than conventional toolpaths. However, a decrease in average cutting loads enable reduction in cutting time with feed optimization. Additionally, the proposed strategy resulted in lower scallop height than conventional machining, thereby improving surface condition.

Small and Medium-sized Enterprises (SMEs) can benefit significantly by exploiting and successfully commercializing technological innovations in academic institutions. The primary challenge in realizing this requirement is the non-availability of systematic repositories and lower Technology Readiness Levels (TRL) with systems or products available at academic institutions. This paper presents Research2Market Connect to aggregate innovations within academic institutions and a listing of the repository on a cloud-based platform for offering shared access to SMEs. As academic research may not be readily suited to market needs, the platform facilitates enhancing the TRL level (from TRL 4 to TRL 7) through online collaborations and resource provisioning using the service-oriented model. The platform elements an academic profile, an industry profile, and a mechanism for improving the TRL level using the Design-as-a-Service (DaaS) framework, are conceived and discussed in the paper. An illustrative case study of the rapid development and deployment of a face shield during the COVID-19 pandemic is presented to showcase the importance of collaborative innovation using a cloud-based platform. It has been shown that cloud-based platforms similar to Research2Market Connect can enable SMEs to leverage research in academic institutions with human expertise and computational resources to overcome resource barriers and swiftly respond to changing market conditions.

Identifying tool wear state is essential for machine operators as it assists in informed decisions for timely tool replacement and subsequent machining operations. As each wear state corresponds to a unique mitigation strategy, timely identification is vital while implementing solutions to minimize tool wear. The paper presents a novel Human Guided-eXplainable Artificial Intelligence (HG-XAI) approach for identifying the tool wear state by integrating human intelligence and eXplainable AI with a pre-trained Convolutional Neural Network (CNN), Efficient-Net-b0 model. The tool wear states were identified based on different wear mechanisms during the machining of IN718. The study considers four distinct tool wear states, i.e., Flank, Flank+BUE, Flank+Face, and Chipping, representing abrasion, adhesion, diffusion, and fracture wear mechanisms. The image-based datasets were created to depict various tool wear states by machining IN718 at varying surface speeds. The effectiveness of the proposed HG-XAI approach was evaluated by comparing its prediction accuracy with a standalone Efficient-Net-b0 model lacking human intelligence and XAI. Further, the scalability of the HG-XAI approach was examined by predicting wear states from images acquired at different cutting parameters. The results from the present study showed that the HG-XAI approach can predict the tool wear state with an accuracy of 93.08

Dissimilar material joining is essential for improving the strength-to-weight ratio of materials for various applications. Friction element welding (FEW) is a promising solution for joining highly dissimilar materials that vary in strength and thickness. However, the influence of the process parameters on the material's resultant microstructure and mechanical properties remains unclear. In this study, the relationship between microstructure and microhardness distribution of the welded specimen is experimentally studied, and the effects of temperature and stress evolution are revealed by a thermal-mechanical finite element model. It is found that the microhardness can be improved by over 50% in the central region due to microstructural change and grain refinement. The beneficial microstructural change can be achieved by inducing either a high peak temperature (over the austenitization temperature) or a high peak stress (over the hardening factor) during the FEW process, which can be obtained by controlling the endload and rotational speed of the friction element. The size of the region with improved hardness is observed to vary with the depth of deformation in the steel layer. For the transverse shear strength (TSS), it is observed that irrespective of the temperature levels reached, TSS increases with increasing stress in the steel layer. Temperature plays a crucial role when the steel layer's temperature is higher than the austenitization start temperature wherein TSS increases with the temperature. (c) 2024 The Authors. Published by ELSEVIER Ltd. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc- nd/4.0)

Machine vision in quality control and sorting applications enhance organizational efficiency. Machine vision systems are coupled with a pre-trained Convolutional Neural Network (CNN) to enhance the capability of the system for classification and identification tasks. The overarching research goal of this study is a) to understand how a CNN decides on classifying threaded fasteners, and b) how well does the CNN's decision making compare with that of a human. In order to answer the first research question, an image-based fastener identification model augmented with a pre-trained CNN was deployed. The CNN used is called Efficient-Net-b0, that can perform a wide range of image classification tasks. The training set provided to the EfficientNet-b0 model consisted of labeled images of 12 types of threaded fasteners. The data set was enlarged by using image augmentation techniques such as varying the brightness, contrast, and orientation of the captured images. The results produced by the CNN classifier were then parsed through Gradient-weighted Class Activation Mapping (Grad-CAM). This technique produce visual explanations of the decisions made by CNNs. This is the XAI component of this research. It provides transparency in the reasons for the identification and classification done by the Efficient-Net-b0, thereby providing context for the key feature of the threaded fastener that was used to classify and identify it. In order to answer the second research question, a user study was conducted. The participants of this study are novice and experienced mechanical engineers enrolled in a Bachelor's and a Master's program at two universities in the United States. The aim of this study was to answer three research sub-questions, each of which was compared to the results from the Efficient-Netb0 as explained by Grad-CAM. The three questions are: i) Can human subjects distinguish between fasteners within the same category?, ii) What features do human subjects look at when distinguishing between fasteners in the same category as compared to XAI?, and iii) Can human subjects identify the same fastener when it is placed in different orientations? The study was conducted via an online Google form that shows the participant two pictures of threaded fasteners simultaneously. There is a total of 40 pairs of images. Each set of pictures comes with two questions. The first question is a multiple-choice question that asks the user whether the image they are looking at represents the same component. The second question is a short answer question asking the user to describe their choice for the multiple-choice question. Each user answers a total of 80 questions, and the total number of participants 43, resulting in a sizable dataset. In addition to providing a comparison between the accuracy rates of both human and AI in classifying fastners the results of the experiments also gave us a better insights into the features/categories utilized by both XAI and participants to achieve the goal of fastner classification.

Manufacturing plants are pursuing a more profound knowledge of movement within the plant, including processes, products, and people. Varied methods are used today to track machines, parts, and bins, including computer vision, Bluetooth, Radio-Frequency Identification (RFID), Wi-Fi, 5G, and Ultra-wideband. Prior studies have examined the capabilities of these systems in the laboratory, and limited studies have analyzed performance in the real world. This work describes the testing methodology to be used in the first study that examines the performance of multiple commercial Ultra-wideband systems within the same environment that closely simulates a real-world automotive production facility to understand the capabilities of modern system performance in a series of functional tests. This work details the tests developed and limited results in cooperation with a large automotive OEM to go beyond the specifications available from each commercial vendor. Future work will expand upon the resulting data output and compare additional types of motion-tracking systems, such as computer vision. (c) 2024 The Authors. Published by ELSEVIER Ltd.

Nickel-Based Superalloys (NBSAs) are widely used for components subjected to high-temperature applications due to their excellent mechanical strength, toughness, and corrosion resistance. Despite favorable properties, NBSAs work-harden during machining, resulting in acute temperature rise at the cutting edge, severe plastic deformation, and rapid tool wear. The lower thermal conductivity, intense friction at the chip-tool interface, chemical affinity with tool material, and temperature gradients typically lead to abrupt crater formation or cutting-edge chipping in addition to rapid flank wear. Three distinct phenomena characterize tool wear during end milling of NBSAs; rapid flank wear, abrupt crater formation, and cutting-edge chipping. The continued use of worn or damaged cutting tools leads to poor surface finish and, eventually, catastrophic failures, resulting in significant machine downtime. As each tool wear condition has a unique mitigation strategy, timely identification and classification are imperative to implement solutions that minimize wear and guide tool replacement. In recent years, the augmentation of vision-based systems with pre-trained Convolutional Neural Networks (CNNs) has shown great promise in failure identification and classification tasks. The present work develops an image-based classification model using a pre-trained CNN, Efficient-Net-b3, for identifying three tool wear conditions during end milling of Inconel 718 (IN718). The network training uses labeled image datasets that capture various tool wear characteristics generated using end-milling experiments. The extensive training dataset requirement of the CNN was met using image augmentation techniques by varying the brightness, contrast, and orientation of the captured images. The prediction abilities of the algorithm were corroborated by validating the model on a validation dataset and further testing on new unseen datasets. It has been shown that Efficient-Net-b3 demonstrates robust prediction accuracy for all three tool wear conditions. The proposed classification model can be further employed for developing an on-machine vision-based tool wear classification system.

Modern manufacturing enterprises must be agile to cope with sudden demand changes arising from increased global competition, geopolitical factors, and unforeseen circumstances such as the Covid-19 pandemic. Small- and Medium-Sized Enterprises (SMEs) in the manufacturing sector lack agility due to lower penetration of Information Technology (IT) and Operational Technology (OT), the inability to employ highly skilled human capital, and the absence of a formal innovation ecosystem for new products or solutions. In recent years, Cloud-based Design and Manufacturing (CBDM) has emerged as an enabler for product realization by integrating various service-based models. However, the existing framework does not thoroughly support the innovation ecosystem from concept to product realization by formally addressing economic challenges and human skillset requirements. The present work considers the augmentation of the Design-as-a-Service (DaaS) model into the existing CBDM framework for enabling systematic product innovations. The DaaS model proposes to connect skilled human resources with enterprises interested in transforming an idea into a product or solution through the CBDM framework. The model presents an approach for integrating human resources with various CBDM elements and end-users through a service-based model. The challenges associated with successfully implementing the proposed model are also discussed. It is established that the DaaS has the potential for rapid and economical product discovery and can be readily accessible to SMEs or independent individuals.

Bearings are critical components for transferring load and motion between subsystems with reduced friction for rotational equipment. Manufacturers implement condition monitoring technology to prevent failures of bearings by using different data acquisition methods, such as vibration, acoustics, temperature, motor current, and ultrasonic sensing, to monitor and predict changes that could indicate early equipment degradation. However, the data quality and availability varies depending on the application, and low data availability from physical environments can lead to poorly trained models. This paper explores how to transfer data and information about failure propensity between bearings of different sizes using a combined physics and data-driven approach. Though the approach is exemplified with roller bearings, the method is extensible to other types of failures in different-sized equipment. Data are generated using an experimental test stand measuring failure of one component, with the intent to scale findings to represent a real-world system; findings are verified with a physics-based model. Data from three different bearing sizes, i.e. 6205, 6206, and 6207, are used to train the algorithms to identify similarity among the sets. Classifiers trained with the raw data provide over 90% accuracy, leading to the conclusion that the data classes are separable based on bearing size. 6205 and 6207 data were scaled to simulate 6206 and tested to see if a classifier could differentiate between the true 6206 bearing data and the simulated bearing data derived from the 6205 and 6207 data. In the simulated data, the classifier accuracy for each algorithm dropped below 90% to as low as 50% (Naive Bayes case). The lower accuracy implied a greater overlap in the vibration features, increasing the data similarity between the different bearing sizes. Future work will further investigate defining dimensionless numbers as scaling parameters for bearing data.

Electrically assisted heat treatment is the process of applying an electric current to a sample during heat treatment. Literature has generally shown there to be a difference in the resulting effects of direct current (DC) current and highly transient current (i.e. electropulsing). However, these differences are poorly characterized. In situ transmission electron microscopy (TEM) observation of an AA7075 sample while DC and pulsed current were passed through it was performed herein to explore the effects of an electric current on precipitate development. Numerical simulation results indicate that the thermal response of the samples was very rapid, causing the sample to reach steady-state temperatures almost instantly. There does not appear to be any significant difference between the results of pulsed current application and DC current. Additionally, the failure mechanism of an electrical biasing TEM sample is explored.

Tool wear plays a decisive role in achieving the required surface quality and dimensional accuracy during the machining of Inconel 718-based products. The highly stochastic phenomenon of tool wear, particularly in later stages, results in difficulty in predicting the failure point of the tool. The present research work aims to study this late-stage wear of the tool by generating consistent wear conditions and thereby decoupling the late-stage wear from the wear history. To do so, a multi-axis grinding operation is employed to create artificial tool wear that replicates the topology of natural wear occurring in the process. In order to evaluate the imitating ability of the proposed methodology, microscopic images in different wear states of naturally and contrived worn tools were analyzed. The methodology was validated by comparing the resulting process forces measured during end milling with the natural and contrived worn tool for different path strategies. Finally, a qualitative finite element (FE) analysis was conducted, and specific force coefficients for worn tool segments were determined through simulation.