Search

While modern 5-axis computer numerical control (CNC) systems offer enhanced design flexibility and reduced production time, the dimensional accuracy of the workpiece is significantly compromised by geometric errors, thermal deformations, cutting forces, tool wear, and fixture-related factors. In-situ sensing, in conjunction with machine learning (ML), has recently been implemented on edge devices to synchronously acquire and agilely analyze high-frequency and multifaceted data for the prediction of workpiece quality. However, limited edge computational resources and lack of interpretability in ML models obscure the understanding of key quality-influencing signals. This research introduces InterpHD , a novel graph-based hyperdimensional computing framework that not only assesses workpiece quality in 5-axis CNC on edge, but also characterizes key signals vital for evaluating the quality from in-situ multichannel data. Specifically, a hierarchical graph structure is designed to represent the relationship between channels (e.g., spindle rotation, three linear axes movements, and the rotary A and C axes), parameters (e.g., torque, current, power, and tool speed), and the workpiece dimensional accuracy. Additionally, memory refinement, separability, and parameter significance are proposed to assess the interpretability of the framework. Experimental results on a hybrid 5-axis LASERTEC 65 DED CNC machine indicate that InterpHD not only achieves a 90.7% F1-Score in characterizing a 25.4 mm counterbore feature deviation but also surpasses other ML models with an F1-Score margin of up to 73.0%. The interpretability of the framework reveals that load and torque have 12 times greater impact than power and velocity feed forward for the characterization of geometrical dimensions. InterpHD offers the potential to facilitate causal discovery and provide insights into the relationships between process parameters and part quality in manufacturing. [ABSTRACT FROM AUTHOR]

With the emergence of the Industrial Internet of Things and Industry 4.0, industrial automation has grown as an important vertical in recent years. Smart manufacturing techniques are now becoming essential to keep up with the global industrial competition. Decreasing machine's downtime and increasing tool life are crucial factors in reducing machining process costs. Therefore, introducing complete process automation utilizing an intelligent automation system can enhance the throughput of manufacturing processes. To achieve this, intelligent manufacturing systems can be designed to recognize materials they interact with and autonomously decide what actions to take whenever needed. This paper aims to present a generalized approach for fully automated machining processes to develop an intelligent manufacturing system. As an objective to accomplish this, the presence of workpiece material is automatically detected and identified in the proposed system using a convolutional neural network (CNN) based machine learning (ML) algorithm. Furthermore, the computer numerical control (CNC) lathe's machining toolpath is automatically generated based on workpiece images for a surface finishing operation. Machining process parameters (spindle speed and feed rate) are also autonomously controlled, thus enabling full machining process automation. The implemented system introduces cognitive abilities into a machining system, creating an intelligent manufacturing ecosystem. The improvised system is capable of identifying various materials and generating toolpaths based on the type of workpieces. The accuracy and robustness of the system are also validated with different experimental setups. The presented results demonstrate that the proposed approach can be applied in manufacturing systems without the need for significant modification. [ABSTRACT FROM AUTHOR]

This paper details the design and operation of a testbed to evaluate the concept of autonomous manufacturing to achieve a desired manufactured part performance specification. This testbed, the autonomous manufacturing system for phononic crystals (AMSPnC), is composed of additive manufacturing, material transport, ultrasonic testing, and cognition subsystems. Critically, the AMSPnC exhibits common manufacturing deficiencies such as process operating window limits, process uncertainty, and probabilistic failure. A case study illustrates the AMSPnC function using a standard supervised learning model trained by printing and testing an array of 48 unique designs that span the allowable design space. Using this model, three separate performance specifications are defined and an optimization algorithm is applied to autonomously select three corresponding design sets to achieve the specified performance. Validation manufacturing and testing confirms that two of the three optimal designs, as defined by an objective function, achieve the desired performance, with the third being outside the design window in which a distinct bandpass is achieved in phononic crystals (PnCs). Furthermore, across all samples, there is a marked difference between the observed bandpass characteristics and predictions from finite elements method computation, highlighting the importance of autonomous manufacturing for complex manufacturing objectives. [ABSTRACT FROM AUTHOR]

Human-robot collaboration, which strives to combine the best skills of humans and robots, has shown board application prospects in meeting safe-effective-flexible requirements in various fields. The ideation of much closer interaction between humans and robots has greatly developed the exploration of digital twin to enhance collaboration. By offering high-fidelity models and real-time physical-virtual interaction, the digital twin enables to achieve an accurate reflection of the physical scenario, including not only human-robot conditions but also environmental changes. However, the appearance of unpredictable events may cause an inconsistency between the established schedule and actual execution. To cope with this issue, an environment-adaptive assignment method based on digital twin for human-robot collaboration is formed in this study. The proposed approach consists of a factor-event-act mechanism that analyzes the dynamic events and their impacts from both internal and external perspectives of digital twin and a genetic algorithm-based assignment algorithm to respond to them. Experiments are carried out in the last part, aiming to show the feasibility of the proposed method. [ABSTRACT FROM AUTHOR]

Vision-based robots have been utilized for pick-and-place operations by their ability to find object poses. As they progress into handling a variety of objects with cluttered state, more flexible and lightweight operations have been presented. In this paper, an autonomous robotic bin-picking platform is proposed. It combines human demonstration with a collaborative robot for the flexibility of the objects and YOLOv5 neural network model for faster object localization without prior computer-aided design models or dataset in the training. After a simple human demonstration of which target object to pick and place, the raw color and depth images were refined, and the one on top of the bin was utilized to create synthetic images and annotations for the YOLOv5 model. To pick up the target object, the point cloud was lifted using the depth data corresponding to the result of the trained YOLOv5 model, and the object pose was estimated by matching them with Iterative Closest Points (ICP) algorithm. After picking up the target object, the robot placed it where the user defined it in the previous human demonstration stage. From the result of experiments with four types of objects and four human demonstrations, it took a total of 0.5 s to recognize the target object and estimate the object pose. The success rate of object detection was 95.6%, and the pick-and-place motion of all the found objects was successful. [ABSTRACT FROM AUTHOR]

The integration of mobile robots in material handling in flexible manufacturing systems (FMS) is made possible by the recent advancements in Industry 4.0 and industrial artificial intelligence. However, effectively scheduling these robots in real-time remains a challenge due to the constantly changing, complex, and uncertain nature of the shop floor environment. Therefore, this paper studies the robot scheduling problem for a multiproduct FMS using a mobile robot for loading/unloading parts among machines and buffers. The problem is formulated as a Markov Decision Process, and the Q-learning algorithm is used to find an optimal policy for the robot's movements in handling different product types. The performance of the system is evaluated using a reward function based on permanent production loss and the cost of demand dissatisfaction. The proposed approach is validated through a numerical case study that compares the proposed policy to a similar policy with different reward function and the first-come-first-served policy, showing a significant improvement in production throughput of approximately 23%. [ABSTRACT FROM AUTHOR]

The vigorous development of the human cyber-physical system (HCPS) and the next generation of artificial intelligence provide new ideas for smart manufacturing, where manufacturing quality prediction is an important issue in the manufacturing system. However, the small-scale data from humans in emerging HCPS-enabled manufacturing restrict the development of traditional quality prediction methods. To address this question, a data augmentation-based manufacturing quality prediction approach in human cyber-physical systems is proposed in this paper. Specifically, a Data Augmentation-Gradient Boosting Decision Tree (DA-GBDT) model is developed for quality prediction under the HCPS context. In addition, an adaptive selection algorithm of data augmentation rate is designed to balance the trade-off between the training time of the prediction model and the prediction accuracy. Finally, the experimental results of automobile covering products demonstrate that the proposed method can improve the average prediction error of the model compared with the prevailing quality prediction methods. Moreover, the predicted quality information can provide guidance for product optimization decisions in smart manufacturing systems. [ABSTRACT FROM AUTHOR]

Human-robot collaboration has become a hotspot in smart manufacturing, and it also has shown the potential for surface defect inspection. The robot can release workload, while human collaboration can help to recheck the uncertain defects. However, the human-robot collaboration-based defect inspection can be hardly realized unless some bottlenecks have been solved, and one of them is that the current methods cannot decide which samples to be rechecked, and the workers can only recheck all of the samples to improve inspection results. To overcome this problem and realize the human-robot collaboration-based surface defect inspection, a two-stage Transformer model with focal loss is proposed. The proposed method divides the traditional inspection process into detection and recognition, designs a collaboration rule to allow workers to collaborate and recheck the defects, and introduces the focal loss into the model to improve the recognition results. With these improvements, the proposed method can collaborate with workers by rechecking the defects and improve surface quality. The experimental results on the public dataset have shown the effectiveness of the proposed method, the accuracies are significantly improved by the human collaboration, which are 1.70%~4.18%. Moreover, the proposed method has been implemented into a human-robot collaboration-based prototype to inspect the carton surface defects, and the results also verify the effectiveness. Meanwhile, the proposed method has a good ability for visualization to find the defect area, and it is also conducive to defect analysis and rechecking. [ABSTRACT FROM AUTHOR]

The bill of materials (BOM) runs through all stages of the life cycle of manufacturing products, which is the core of manufacturing enterprises. With increasing complexity of modern manufacturing engineering and widespread using of intelligent manufacturing technology, the BOM data keeps rising and transformation process is increasingly frequent and complicated. In order to improve efficiency of BOM management and ensure the diversity, accuracy and consistency of BOM in the product development, the BOM multi-view integrated management and mapping method for complex products were researched. First, a complex product BOM integrated management framework and the evolution model based on multiple views were established which described the BOM integrated management mechanism and transformation relationship among different BOMs. Subsequently, process of BOM transformation was analyzed, and a BOM transformation model was proposed. Moreover, a rule-based BOM multi-view mapping algorithm was proposed. With the rule definition and mathematical modelling for key components, the complex mapping principle was elaborated. Finally, the BOM multi-view transformation cases and the prototype system were illustrated and discussed, which verified the feasibility and versatility of model and method.

Ensuring quality production is essential in additive manufacturing processes such as selective laser sintering (SLS). Deep learning based automated systems may provide a better way for real time monitoring to ensure quality production. These systems are usually trained in a fully supervised manner which require large amount of labeled data. Obtaining labeled training data in large quantities is a tedious and time consuming process. To overcome this, a novel semi-supervised deep learning approach is proposed in this work, which can be trained using both labeled and unlabeled data, and hence, reducing the manual labeling efforts needed to train the system. Experimental results on a SLS powder bed defect detection dataset show that the proposed approach is the new state-of-the-art, and shows its potential as a standalone real time monitoring system for SLS. In this dataset the proposed approach beats the state-of-the-art accuracy of 98 % with only 25 % of the labeled training data compared to other approaches. In addition, an extensive set of experiments were conducted on three additional public defect inspection datasets (NEU steel surface defects, KolektorSDD surface images of plastic electronics commutators, and surface textures) to show the applicability of the proposed approach on other computer integrated manufacturing environments for quality inspection. In all of these datasets the proposed approach beats the state-of-the-art results with relatively small amounts of training data compared to other approaches, which shows the effectiveness of the proposed approach for real time, automated, and accurate quality inspection. [ABSTRACT FROM AUTHOR]

In the realm of additive manufacturing, the selection of process parameters to avoid over and under deposition entails a time-consuming and resource-intensive trial-and-error approach. Given the distinct characteristics of each part geometry, there is a pressing need for advancing real-time process monitoring and control to ensure consistent and reliable part dimensional accuracy. This research shows that support vector regression (SVR) and convolutional neural network (CNN) models offer a promising solution for real-time process control due to the models' abilities to recognize complex, non-linear patterns with high accuracy. A novel experiment was designed to compare the performance of SVR and CNN models to indirectly detect bead height from a coaxial image of a melt pool from a single-layer, single bead build. The study showed that both SVR and CNN models trained on melt pool data collected from a coaxial optical camera can accurately predict the bead height with a mean absolute percentage error of 3.67% and 3.68%, respectively. [ABSTRACT FROM AUTHOR]

What makes living systems flexible so that they can react quickly and adapt easily to changing environments? This question has not only engaged biologists for decades but is also of great interest to computer scientists and engineers who seek inspiration from nature to increase the flexibility of task-performing systems such as machine learning systems, robots, or manufacturing systems. In this paper, we give a broad overview of design features of living systems that are known to promote flexibility. We call these design features the “elements of flexibility”. Moreover, to facilitate interdisciplinary, bio-inspired research that brings the elements of flexibility to man-made task-performing systems, we introduce a general formalism for system flexibility optimization. The formalism is intended to (i) provide a common language to communicate ideas about system flexibility among researchers with different backgrounds, (ii) help to understand and compare existing research on system flexibility, e.g., in transfer learning or manufacturing flexibility, and (iii) provide a basis for a general theory of system flexibility optimization.

The implementation of precision machine tool thermal error compensation in edge-cloud-fog computing architecture has the potential to control the thermal error. However, the challenges faced by the successful implementation are described as follows: The data collection and transfer efficiency are low, and the control accuracy is not deficient. To address these challenges, a hardware design scheme is proposed for the high-performance intelligent gateway node based on the low-power processor architecture of ARM Cortex-A7. Moreover, a new transformer-improved-gate long short-term memory model is proposed, and then it is embedded into edge-cloud-fog computing architecture. With the implementation of gear profile grinding machine thermal error compensation in edge-cloud-fog computing architecture, the maximum values of the tooth profile tilt deviation are reduced from 17.4 µm to 5.4 µm and from 17.9 µm to 5.8 µm for the left and right tooth flanks, respectively. Moreover, the maximum values of the tooth profile deviation are reduced from 18.9 µm to 6.1 µm and from 18.2 µm to 5.8 µm for the left and right tooth flanks, respectively. Compared with the traditional collection mode, the response delay of the designed intelligent gateway in the acquisition mode is reduced by 40%. [ABSTRACT FROM AUTHOR]

Electric vehicle production is subjected to high manufacturing cost and environmental impact. Disassembling and remanufacturing the lithium-ion power packs can highly promote electric vehicle market penetration by procuring and regrouping reusable modules as stationary energy storage devices and cut life-cycle cost and environmental impact. Disassembly efficiency is crucial for battery remanufacturing companies in reverse supply chains. However, disassembly planning suffers from high computational complexity and inferior solutions. This paper developed a multi-objective mathematical model and presented a novel hybrid genetic-firework algorithm based on the precedence graph for obtaining solutions to disassemble the electric vehicle power pack into module levels in an efficient manner. The objectives for the model include not only smoothness of working stations, cycle time, and economic returns, but also consider operation safety and energy consumption. The proposed hybrid algorithm explored the performance of the novel solution searching mechanism of combining the firework and genetic algorithms. The proposed approach is compared with the commonly used multi-objective evolutionary algorithms in the literature, showing its feasibility and effectiveness. [ABSTRACT FROM AUTHOR]