Edge Machine Research Articles

Establishing Local Diagnostic Reference Levels for Head Computed Tomography Examinations

Background/Objectives: Head Computed Tomography (CT) is an essential diagnostic tool for identifying brain pathologies and visualizing blood vessels. However, CT exposes patients to ionizing radiation, making it necessary to establish local diagnostic reference levels (DRLs) to ensure patient safety. This study aimed to establish DRLs for head CT scans and assess the influence of patient characteristics on radiation dose. Methods: A retrospective analysis was conducted on 2043 non-contrast and 488 contrast-enhanced head CT scans performed between 1 July 2023 and 31 March 2024 using a SIEMENS SOMATOM Definition Edge machine. Computed Tomography Dose Index (CTDIvol) and Dose-Length Product (DLP) values were analyzed, with DRLs set at the 75th percentile. The influence of gender, height, and weight on radiation dose was also evaluated. Results: The DRL for both non-contrast and contrast-enhanced scans was 58.18 mGy for CTDIvol and 1018.11 mGy·cm for DLP per acquisition. Total DLP was 2046.09 mGy·cm for contrast-enhanced and 1027.99 mGy·cm for non-contrast scans. No significant correlation was found between patient characteristics and radiation dose, allowing for a uniform DRL to be established. Conclusions: Uniform DRLs were successfully established for head CT scans, ensuring safe radiation doses for both non-contrast and contrast-enhanced studies. The lack of correlation between patient-specific factors and dose supports the use of standardized DRLs, contributing to optimized radiation safety in head CT diagnostics.

Experimental study on manufacturing of vacuum-brazed grinding tools with diamond grains

AbstractThis work provides an experimental study on manufacturing methods for vacuum-brazed diamond grinding for edge machining of carbon fiber-reinforced polymers (CFRP). Therefore, various methods for applying active brazing alloys and diamond grains were investigated. The application methods were evaluated regarding practicability and the possibility of automation. Additionally, the samples were examined after vacuum brazing and brazing results were compared concerning method, amount and type of brazing alloy, diamond grains and binding agent. It was found that the best results could be achieved by using fluidized bed coating for diamond application and a brazing alloy containing a 15% binding agent. The method and brazing alloy were then used to produce cylindrical mounted points for grinding tests. It was possible to show that vacuum-brazed tools performed as good as commercially available electroplated tools.

A Survey of Machine Learning in Edge Computing: Techniques, Frameworks, Applications, Issues, and Research Directions

Internet of Things (IoT) devices often operate with limited resources while interacting with users and their environment, generating a wealth of data. Machine learning models interpret such sensor data, enabling accurate predictions and informed decisions. However, the sheer volume of data from billions of devices can overwhelm networks, making traditional cloud data processing inefficient for IoT applications. This paper presents a comprehensive survey of recent advances in models, architectures, hardware, and design requirements for deploying machine learning on low-resource devices at the edge and in cloud networks. Prominent IoT devices tailored to integrate edge intelligence include Raspberry Pi, NVIDIA’s Jetson, Arduino Nano 33 BLE Sense, STM32 Microcontrollers, SparkFun Edge, Google Coral Dev Board, and Beaglebone AI. These devices are boosted with custom AI frameworks, such as TensorFlow Lite, OpenEI, Core ML, Caffe2, and MXNet, to empower ML and DL tasks (e.g., object detection and gesture recognition). Both traditional machine learning (e.g., random forest, logistic regression) and deep learning methods (e.g., ResNet-50, YOLOv4, LSTM) are deployed on devices, distributed edge, and distributed cloud computing. Moreover, we analyzed 1000 recent publications on “ML in IoT” from IEEE Xplore using support vector machine, random forest, and decision tree classifiers to identify emerging topics and application domains. Hot topics included big data, cloud, edge, multimedia, security, privacy, QoS, and activity recognition, while critical domains included industry, healthcare, agriculture, transportation, smart homes and cities, and assisted living. The major challenges hindering the implementation of edge machine learning include encrypting sensitive user data for security and privacy on edge devices, efficiently managing resources of edge nodes through distributed learning architectures, and balancing the energy limitations of edge devices and the energy demands of machine learning.

Edge Machine Learning-Enabled Predictive Fault Detection System for Conveyor Belt Maintenance Optimization in Industrial Settings

Digital transformation is essential for industrial and manufacturing sectors, especially in conveyor belt systems. On the other hand, Innovative electronic components and software applications offer opportunities for advanced fault prediction, minimizing operational disruptions, reducing inefficiencies, and enhancing overall industrial efficiency. Efficient conveyor belt systems are crucial for industrial operations, requiring early fault detection to prevent costly downtime and ensure safety. The utilization of embedded machine learning via the Edge Impulse platform enhances data collection and spectral feature extraction using wavelet approximation. This article presents a novel methodology for predictive fault detection in conveyor belt systems employing edge machine learning. Leveraging the ESP32 microcontroller and accelerometer sensor for real-time vibration data capture, a four-layer Artificial Neural Network (ANN) model, trained on 336 feature samples, is deployed via Edge Impulse and TensorFlow scripts. TensorFlow Lite compresses the model for microcontroller integration. The ESP32 microcontroller, acting as an edge device, achieves real-time predictions with 97.5% accuracy and 0.20 minimal loss. This work promises significant strides in predictive maintenance, offering cost savings, operational reliability, and efficiency gains in industrial settings, revolutionizing maintenance strategies.

Special Issue on Advanced Metal Cutting Technologies

Cutting technologies are utilized in many industrial sectors, such as automobile, aircraft, and medical-device manufacturing as well as dies and molds. Thus, the requirements for higher geometric accuracy, better surface integrity, and longer component service lifetimes have substantially increased. In addition, maintaining high machining efficiency while simultaneously reducing power consumption and CO2 emissions during machining is important for sustainable development. This special issue features 11 papers on the most recent advances in cutting technologies for metals and composite materials. - Reverse finishing characteristics of drilling surfaces - Machining error simulation for end milling - Visualization of cutting phenomena using a single-crystal diamond tool - Milling of TiB2 particle-reinforced high-modulus steel - Electrodischarge-assisted turning of carbon fiber-reinforced polymers - Suppression of chatter vibration during double-insert turning - Prediction of surface roughness components during turning - Microtome blades for high-precision tissue sectioning - Boiling of coolant near the cutting edge in high-speed machining - Residual stress during drilling of aluminum alloy - Effect of strain hardening on burr control during drilling This issue provides an understanding of recent developments in cutting technologies, aiming to inspire further research in this field. We deeply appreciate the careful work of all the authors and thank the reviewers for their diligent efforts.

Boiling of Coolant Near the Cutting Edge in High Speed Machining of Difficult-to-Cut Materials

This study investigates film boiling of coolant as a cooling inhibitor in a narrow wedge-shaped space between the tool flank face and the machined surface of a workpiece, observed during high-speed turning of stainless steel SUS304 and nickel-based superalloy Inconel 718. The boiling, likely triggered by high surface temperatures at both the face and surface close to the cutting edge, impedes coolant access to the tool tip area and efficient cooling. Therefore, the impact of coolant pressure on the boiling zone size was initially explored across pressures ranging from 0.1 to 20 MPa. A burn mark band due to coolant boiling, distinctly visible on the flank face of an insert with a yellow hard coating, expanded over cutting time. The film boiling area width, or the distance from the flank wear area to the band, decreased with increasing coolant pressure, reflecting the enhanced cooling ability and tool life with high-pressure coolant. Applying Boyle–Charles’ law to film boiling indicated that vapor pressure was directly related to coolant velocity rather than pressure. In contrast, the boiling area width increased with increasing cutting speed.

Airfare Forecasting using Machine Learning to Predict Prices

Abstract: The issue of predicting ticket 111 costs is the focus of this essay. With the assumption that these characteristics have an impact on the cost of an airline ticket, a set of features typical of a normal flight is determined for this purpose. Eight cutting edge machine learning (ML) models using the characteristics are trained to forecast the cost of airline tickets, and the models' output is contrasted with one another. This work examines how the feature set used to represent an airline affects accuracy as well as the prediction accuracy of each model. To train each machinelearning model for the trials, a unique dataset including 1814 Aegean Airlines data flights for a particular foreign destination (from Thessaloniki to Stuttgart) was created. Key Words: Airfare price prediction, Machine learning models, Feature dependency, Regression accuracy.

Decision Support System for Optimizing Tactics and Strategies of Sports Competition Using Reinforcement Learning Algorithm

This paper presented an augmented reality secure edge machine learning dynamic optimization model for detecting volleyball movement recognition using a decision support system. The proposed Edge Augmented Decision Support System Blockchain (EdgeAu-DSSBC) model aims to address the limitations of centralized machine learning models, such as high latency, network congestion, and data privacy concerns, by utilizing edge computing and dynamic optimization techniques. The proposed EdgeAu-DSSBC system consists of two main components: an augmented reality interface that allows users to interact with the system and an edge machine learning algorithm that performs the recognition task. The system uses dynamic optimization techniques to optimize the parameters of the machine learning algorithm in real time based on the feedback received from the users. The decision support system provides additional guidance to the users and helps them to make informed decisions based on the recognition results. The EdgeAu-DSSBC model was evaluated using a dataset of volleyball movement videos and compared to existing centralized and edge-based machine-learning models. The experimental results demonstrate that the EdgeAu-DSSBC exhibits improved performance with an accuracy of 99%.

Choosing parameters of a magnetic abrasive polishing for machining of edges of aircraft body parts from aluminum alloys

Choosing parameters of a magnetic abrasive polishing for machining of edges of aircraft body parts from aluminum alloys

Моделирование скругления острых кромок наклонных охлаждающих отверстий лопаток турбин методом электроэрозионного фрезерования на многоосевых станках с ЧПУ

The article views efficiency improving issues, quality and accuracy of sharp edge smoothing for numerous angular fans in the turbine blades of gas turbine engines made of heat-resistant nickel alloys and used in modern growing passenger aviation transport based on the development of ED milling new technology for modern multi-axis CNC machines with a rotating hollow electrode and a high-pressure fluidsupply of a special flush through the electrode body. An automated process for sharp edge machining has been developed for the production of more than 200 differently placed angular fans with a diameter of 0,45…0,6 mm in one blade instead of labor-intensive manual benchwork, where various needles, abrasive and metal rotary files are used. It is found that modeling and the development of special software resulted in a uniform radius of smoothing with high accuracy due to automatic ED milling operation along the entire trailing edge of each angular hole in the turbine blades at a multi-axis machining electro discharge center. The use of a rounded edge of the outlet ports provides more efficient cooling of the turbine blades and higher performance of gas turbine engines. At the same time, it is found that with a significant efficiency improving in blade sharp edge machining, a higher quality of the rounded edge surfaces in the holes is ensured. The results of the conducted research are recommended for the adoption of technologies in the aircraft engine industry.

A detailed study on Algorithms for Predictive Maintenance in Smart Manufacturing: Chip Form Classification using Edge Machine Learning

A detailed study on Algorithms for Predictive Maintenance in Smart Manufacturing: Chip Form Classification using Edge Machine Learning

Design Space Exploration for Edge Machine Learning Featured by MathWorks FPGA DL Processor: A Survey

Design Space Exploration for Edge Machine Learning Featured by MathWorks FPGA DL Processor: A Survey

High-productivity, field-relevant assessment of relative fracture resistance for hard materials – The ball mill edge chipping test (BMECT)

Abrasion-resistant materials for rock-facing applications typically display low fracture toughness. For materials selection, available fracture data are limited, due to the cost of testing. To address this shortage, a new type of fracture test, denoted the ball mill edge chipping test (BMECT), has been designed. The device is derived from the ball mill abrasion test (BMAT) [37]. It uses simple block or ball specimens with machined edges. Specimens of multiple materials are tested simultaneously. Collisions promote multiple small-scale fracture events at the exposed edges, causing progressive volume loss. The reciprocal of volume loss indicates relative fracture resistance.This paper analyses the BMECT’s relevance to service, and explores its capacity to generate data at low cost. Several data sets are presented, illustrating its application to investigate factors affecting fracture resistance among common abrasion-resistant materials. The results demonstrate multiple advantages over traditional fracture tests — including low cost, high sensitivity to differentiate between materials, and close replication of in-service damage modes. It represents a fundamentally new approach to fracture resistance assessment. Whereas traditional tests are regarded as deterministic, the BMECT is stochastic, providing potential for superior statistical data quality.

Task Placement and Resource Allocation for Edge Machine Learning: A GNN-Based Multi-Agent Reinforcement Learning Paradigm

Task Placement and Resource Allocation for Edge Machine Learning: A GNN-Based Multi-Agent Reinforcement Learning Paradigm

Exploring machine learning to hardware implementations for large data rate x-ray instrumentation

Over the past decade, innovations in radiation and photonic detectors considerably improved their resolution, pixel density, sensitivity, and sampling rate, which all contribute to increased data generation rates. This huge data increases the amount of storage required, as well as the cabling between the source and the storage units. To overcome this problem, edge machine learning (EdgeML) proposes to move computation units near the detectors, utilizing machine learning (ML) models to emulate non-linear mathematical relationships between detector’s output data. ML algorithms can be implemented in digital circuits, such as application-specific integrated circuits and field-programmable gate arrays, which support both parallelization and pipelining. EdgeML has both the benefits of edge computing and ML models to compress data near the detectors. This paper explores the currently available tool-flows designed to translate software ML algorithms to digital circuits near the edge. The main focus is on tool-flows that provide a diverse range of supported models, optimization techniques, and compression methods. We compare their accessibility, performance, and ease of use, and compare them for two high data-rate instrumentation applications: (1) CookieBox, and (2) billion-pixel camera.

Effects of Machining Parameters and Tool Reconditioning on Cutting Force, Tool Wear, Surface Roughness and Burr Formation in Nickel-Based Alloy Milling.

Nickel-based superalloys are among the most difficult materials to machine because they have high thermal strength, they are prone to hardening, carbides severely abrade the tool, and they have very poor thermal conductivity. Slot milling is a specific issue as it is characterized by rapid tool wear and frequent tool breakages. This is why reconditioned tools are frequently employed in industrial environments, as they can considerably decrease the expenses associated with tools. The chosen machining strategy also plays a crucial role in the tool's lifespan and the quality of the machined surface, making it essential to select the appropriate strategy. Hence, the authors have opted for two conventional trochoidal strategies, namely the circular and swinging toolpath, along with a contemporary toolpath known as the Autodesk Inventor HSM Adaptive strategy. The authors investigated the effects of technological parameters and toolpaths on cutting forces, tool wear, surface roughness and burr formation on machined edges. The results show that lower cutting parameters and adaptive strategies lead to the smallest tool loads, tool wear, the best quality of surface roughness and burr formation on machined edges.

A future location prediction method based on lightweight LSTM with hyperparamater optimization

In this study, we presented a method for future location prediction based on machine learning over geopositioning data sets. There are large amounts of geopositioning data sets collected by mobile devices mainly due to modern geopositioning systems such as GPS, GLONASS and Galileo. Based on these geopositioning data sets, it is possible to have a wide variety of location-based services. These data sets can be used for future location prediction of objects, especially humans. Additionally, they have a high possibility for further applications. The purpose of this research is to present a simple and lightweight method that can be applicable to devices with lower computing capability devices, such as AIoT (Artificial Intelligence of Things) or EdgeML (Edge Machine Learning) devices. We introduced a basic LSTM (Long Short Term Memory) model with hyperparameter optimization, especially on window size of continuous geopositioning data, using limited previous geopositioning data for location prediction purposes. We found that the results of using our method for future location prediction are considerably fast and accurate compared with existing neural network-model-based approaches. We also applied our method to non-continuous geopositioning data sets and found it to be equally effective.

Current opinion: advances in machine perfusion and preservation of vascularized composite allografts - will time still matter?

A major hurdle hindering more widespread application of reconstructive transplantation is the very limited cold ischemia time (CIT) of vascularized composite allografts (VCAs). In this review, we discuss cutting edge machine perfusion protocols and preservation strategies to overcome this limitation. Several preclinical machine perfusion studies have demonstrated the multifactorial utility of this technology to extend preservation windows, assess graft viability prior to transplantation and salvage damaged tissue, yet there are currently no clinically approved machine perfusion protocols for reconstructive transplantation. Thus, machine perfusion remains an open challenge in VCA due to the complexity of the various tissue types. In addition, multiple other promising avenues to prolong preservation of composite allografts have emerged. These include cryopreservation, high subzero preservation, vitrification and nanowarming. Despite several studies demonstrating extended preservation windows, there are several limitations that must be overcome prior to clinical translation. As both machine perfusion and subzero preservation protocols have rapidly advanced in the past few years, special consideration should be given to their potential complementary utilization. Current and emerging machine perfusion and preservation technologies in VCA have great promise to transform the field of reconstructive transplantation, as every extra hour of CIT helps ease the complexities of the peri-transplant workflow. Amongst the many advantages, longer preservation windows may allow for elective procedures, improved matching, establishment of novel immunomodulatory protocols and global transport of grafts, ultimately enabling us the ability to offer this life changing procedure to more patients.

Wearable edge machine learning with synthetic photoplethysmograms

Strict privacy regulations pose challenges to the development of machine learning (ML) in the field of health technology where data is particularly sensitive. Gathering and using robust, bias-free, and suitably anonymized datasets required by ML models is difficult, time-consuming, and thus expensive. Parametric synthetic data offers a solution by mimicking real-world processes with easily adjustable parameters that shape the information content of the data as desired. This article presents a system demonstrating how synthetic data can be used in conjunction with wearable edge devices. Importantly, the system preserves privacy as there is no risk of leaking sensitive information from the model or during the use of the wearable device. The system consists of (1) a synthetic photoplethysmogram (PPG) model, (2) convolutional neural network (CNN) models trained with the synthetic signals, (3) a wearable edge device that computes heart rate from real-time PPG signals using the developed CNN models, and (4) an accompanying mobile phone application receiving the results. The synthetic model produces realistic PPG signals together with labels that can be used in CNN model training. The quality of the synthetic data is sufficient to train even a tiny CNN model with only two convolutional layers and 28 parameters to detect PPG waveform feet. The developed wearable device is able to run the model smoothly and the performance of the model is on par with the more complex models and other foot detection algorithms.

A Comprehensive Review and a Taxonomy of Edge Machine Learning: Requirements, Paradigms, and Techniques

The union of Edge Computing (EC) and Artificial Intelligence (AI) has brought forward the Edge AI concept to provide intelligent solutions close to the end-user environment, for privacy preservation, low latency to real-time performance, and resource optimization. Machine Learning (ML), as the most advanced branch of AI in the past few years, has shown encouraging results and applications in the edge environment. Nevertheless, edge-powered ML solutions are more complex to realize due to the joint constraints from both edge computing and AI domains, and the corresponding solutions are expected to be efficient and adapted in technologies such as data processing, model compression, distributed inference, and advanced learning paradigms for Edge ML requirements. Despite the fact that a great deal of the attention garnered by Edge ML is gained in both the academic and industrial communities, we noticed the lack of a complete survey on existing Edge ML technologies to provide a common understanding of this concept. To tackle this, this paper aims at providing a comprehensive taxonomy and a systematic review of Edge ML techniques, focusing on the soft computing aspects of existing paradigms and techniques. We start by identifying the Edge ML requirements driven by the joint constraints. We then extensively survey more than twenty paradigms and techniques along with their representative work, covering two main parts: edge inference, and edge learning. In particular, we analyze how each technique fits into Edge ML by meeting a subset of the identified requirements. We also summarize Edge ML frameworks and open issues to shed light on future directions for Edge ML.