Architecture Model Research Articles

Pseudo Replay-based Class Continual Learning for Online New Category Anomaly Detection in Advanced Manufacturing

The incorporation of advanced sensors and machine learning techniques has enabled modern manufacturing enterprises to perform data-driven classification-based anomaly detection based on the sensor data collected in manufacturing processes. However, one critical challenge is that newly presented defect category may manifest as the manufacturing process continues, resulting in monitoring performance deterioration of previously trained machine learning models. Hence, there is an increasing need for empowering machine learning models to learn continually. Among all continual learning methods, memory-based continual learning has the best performance but faces the constraints of data storage capacity. To address this issue, this paper develops a novel pseudo replay-based continual learning framework by integrating class incremental learning and oversampling-based data generation. Without storing all the data, the developed framework could generate high-quality data representing previous classes to train machine learning model incrementally when new category anomaly occurs. In addition, it could even enhance the monitoring performance since it also effectively improves the data quality. The effectiveness of the proposed framework is validated in three cases studies, which leverages supervised classification problem for anomaly detection. The experimental results show that the developed method is very promising in detecting novel anomaly while maintaining a good performance on the previous task and brings up more flexibility in model architecture.

A Reconfigurable Architecture for Industrial Control Systems: Overview and Challenges

The closed architecture and stand-alone operation model of traditional industrial control systems limit their ability to leverage ubiquitous infrastructure resources for more flexible and intelligent development. This restriction hinders their ability to rapidly, economically, and sustainably respond to mass customization demands. Existing proposals for open and networked architectures have failed to break the vicious cycle of closed architectures and stand-alone operation models because they do not address the core issue: the tight coupling among the control, infrastructure, and actuator domains. This paper proposes a reconfigurable architecture that decouples these domains, structuring the control system across three planes: control, infrastructure, and actuator. The computer numerical control (CNC) system serves as a primary example to illustrate this reconfigurable architecture. After reviewing open and networked architectures and discussing the characteristics of this reconfigurable architecture, this paper identifies three key challenges: deterministic control functionality, the decoupling of control modules from infrastructures, and the management of control modules, infrastructures, and actuators. Each challenge is examined in detail, and potential solutions are proposed based on emerging technologies.

Stock price prediction based on hybrid CNN-LSTM model

Abstract. Stock price prediction is of great significance but faces numerous challenges. In this study, a hybrid CNN-LSTM model was utilized. The stock data were processed and feature engineering was conducted. The model architecture and training strategies were expounded. Experimental results demonstrated that this model outperformed traditional methods and benchmark models on the test set, featuring strong capabilities in feature extraction and handling long-term dependencies. Key roles of data preprocessing, hyperparameter adjustment, and model fusion were also summarized, providing valuable references for stock price prediction.

Advancements in Target Detection: A Comparative Analysis of Deep Learning Models for Real-Time Applications

Abstract. Target detection is a vital field within computer vision, playing an essential role in applications. This paper investigates the advancements and efficiencies of contemporary target detection methodologies, focusing on deep learning frameworks. Through a systematic review and evaluation of these models' architectures and performances using the Common Objects in Context (COCO) dataset, the study highlights their operational effectiveness and practical implications in real-world scenarios. A detailed comparative analysis is conducted, assessing the models based on mean Average Precision (mAP) and input dimensions to determine their suitability across various detection tasks. You Only Look Once version 3 (YOLOv3), in particular, is recognized for its ability to combine high-speed detection with significant accuracy, effectively addressing real-time processing challenges. The results confirm that YOLOv3, despite its smaller input size, performs comparably to more complex systems, demonstrating notable enhancements in design efficiency and processing speed. This research underscores the potential for future optimizations in model architectures to bridge the gap between high-speed and high-accuracy detection tasks, potentially transforming real-time applications across multiple sectors. The practical significance of this work lies in its ability to guide future developments in object detection, benefiting both academic research and industrial applications.

Generative AI for Predictive Maintenance: Predicting Equipment Failures and Optimizing Maintenance Schedules Using AI

Predictive maintenance has emerged as a transformative approach to managing equipment health, reducing unplanned downtime, and extending asset lifespan. Leveraging advancements in generative artificial intelligence (AI), this paper explores the role of AI-driven predictive maintenance in predicting equipment failures and optimizing maintenance schedules. Traditional maintenance strategies, such as reactive and preventive approaches, often lead to inefficiencies, increased operational costs, and unexpected breakdowns. Predictive maintenance, powered by AI, offers a proactive alternative that not only anticipates failures but also enhances scheduling efficiency, maximizing equipment uptime and reducing maintenance costs. Generative AI models, including techniques such as Generative Adversarial Networks (GANs) and reinforcement learning, have shown immense promise in learning complex patterns from historical data and simulating potential equipment failure scenarios. These AI-driven models can analyze vast and diverse data sources—including sensor readings, maintenance logs, environmental conditions, and historical failures—to provide accurate, real-time insights into equipment health. This paper details the architecture and functioning of generative AI models in predictive maintenance, emphasizing their role in both anomaly detection and failure prediction. A systematic comparison of reactive, preventive, and predictive maintenance is provided, underscoring the unique benefits and challenges of predictive maintenance. We discuss the types of data essential for predictive maintenance and present sample data structures used in model training and deployment. Additionally, this paper demonstrates how generative AI models predict equipment failures by identifying anomalous behaviors before they escalate, enabling preemptive actions. A failure probability model is presented to illustrate how failure risks evolve over time, alongside tables showcasing the critical data points in predictive maintenance. The paper also explores the optimization of maintenance schedules using generative AI, where models simulate and compare different maintenance timing strategies, ultimately minimizing downtime and maximizing productivity. However, we also acknowledge the current limitations of generative AI in this domain, including data privacy concerns, computational intensity, and the challenges of model interpretability for practical implementation. Looking forward, we examine future trends such as the integration of Internet of Things (IoT) devices and the emergence of more sophisticated AI models that will likely enhance predictive maintenance applications. This paper concludes by highlighting the transformative potential of generative AI for predictive maintenance, offering insights for industries seeking to innovate their maintenance practices and achieve superior operational resilience.

Application of convolutional neural networks in image classification and applications of improved convolutional neural networks

Abstract. This paper reviews the application and improvement of convolutional neural networks (CNNs) in image classification. Firstly, a shallow CNN for interstitial lung disease image classification is presented. This model suppresses overfitting through a unique network architecture and optimisation algorithm. Next, the improved VGG16 architecture and MIDNet18 model are discussed and their superior performance in brain tumour image classification is demonstrated. Subsequently, a CNN-CapsNet model for cervical cancer image classification and its improvement are presented and the customised model is compared with the conventional VGG-16 CNN architecture in the paper. Next, the application of sparse convolutional kernels and hybrid sparse convolutional kernels (HDCs) in solving the problem of computational resource consumption is presented. Subsequently, methods for solving the problem of limited training data through transfer learning and network data augmentation techniques are discussed, as well as GAN-generated datasets for solving the overfitting problem. Finally, the effect of degraded images on the classification effectiveness of CNNs is explored. The results show that the improved CNN architecture and algorithms have significant effects in solving the problems of overfitting and computational resource consumption, and can significantly improve the accuracy and efficiency of image classification. And degraded images do adversely affect the accuracy of CNN for image classification.

Enhancing Lung Cancer Screening with Bidirectional LSTM and GRU Models

Abstract. This study aims to enhance lung cancer patient screening by developing and evaluating bidirectional Long Short-Term Memory (LSTM) and bidirectional Gated Recurrent Unit (GRU) models using the Lung Cancer dataset from Kaggle. The dataset includes 16 features related to patient symptoms and lung cancer status, providing a broad spectrum of symptoms to improve model accuracy. The research advances Artificial Intelligence (AI)-driven healthcare by integrating these sophisticated machine learning techniques into diagnostic processes. The methodology involves four main steps: preprocessing the dataset for model compatibility, defining the model architecture with bidirectional LSTM and GRU layers and evaluating its performance. The results show an overall accuracy of 52.17%, with accuracy, recall, and F1 scores for both cancerous and non-cancerous categories around 50%. Despite the hybrid model's average performance, it establishes a basis for future enhancements. Optimizing model parameters and exploring additional other techniques to improve prediction accuracy and clinical applicability will be done in the future.

Influence Reasoning Capabilities of Large Language Models in Social Environments

We ask whether state-of-the-art large language models can provide a viable alternative to human annotators for detecting and explaining behavioural influence online. Working with a large corpus of online interactions retrieved from the social media platform Mastodon, we cross-examine a dataset containing 11,000 LLM influence labels and explanations across nine state-of-the-art large language models from 312 scenarios. We use a range of resolution categories and four stages of shot prompting to further measure the importance of context to language model performance. We also consider the impact of model architecture, and how social media content and features from the explanation impact model labelling accuracy. Our experiment shows that whilst most large language models struggle to identify the correct framing of influence from an interaction, at lower label resolutions, models like Flan and GPT-4 Turbo perform with an accuracy of 70%-80%, demonstrating encouraging potential for future social influence identification and explanation, and contributing to our understanding of the general social reasoning capabilities of large language models.

Enterprise Architecture Business Model Planning Using EAP Framework (Case Study: PT. Gempita Cahaya Makmur)

The advancement of technology in a company has an impact on improving business quality. From this observation, architectural planning is a part that is used to build alignment between business strategy and information technology. Architecture within the business domain illustrates how a company conducts business activities and functions to achieve the company's goals. Therefore, the company's business model architecture depicts the current state of architecture by identifying business needs and activities. From this study, the business role of PT. GEMPITA CAHAYA MAKMUR, a company engaged in the procurement of goods and services, especially in the field of wholesale office stationery, printing, photocopier sales, and photocopier and laptop rentals, which has customers from medium-sized companies, large companies, both private and government. The use of the Enterprise Architecture Planning or EAP framework focuses on business architecture. The purpose of this research is expected to produce a blueprint proposal that will be beneficial for PT. GEMPITA CAHAYA MAKMUR to plan the business model architecture that will become the foundation for the design phase of application architecture.

Integrating spoken instructions into flight trajectory prediction to optimize automation in air traffic control

The booming air transportation industry inevitably burdens air traffic controllers’ workload, causing unexpected human factor-related incidents. Current air traffic control systems fail to consider spoken instructions for traffic prediction, bringing significant challenges in detecting human errors during real-time traffic operations. Here, we present an automation paradigm integrating controlling intent into the information processing loop through the spoken instruction-aware flight trajectory prediction framework. A 3-stage progressive multi-modal learning paradigm is proposed to address the modality gap between the trajectory and spoken instructions, as well as minimize the data requirements. Experiments on a real-world dataset show the proposed framework achieves flight trajectory prediction with high predictability and timeliness, obtaining over 20% relative reduction in mean deviation error. Moreover, the generalizability of the proposed framework is also confirmed by various model architectures. The proposed framework can formulate full-automated information processing in real-world air traffic applications, supporting human error detection and enhancing aviation safety.

Impact of fluvial meander-belt sedimentary heterogeneity on the efficiency of low-enthalpy geothermal doublets: heat-transport simulations of forward stratigraphic models

In the present study, different geocellular models of meander-belt stratigraphic architectures were produced that are representative of the sedimentary products of sand-bed meandering rivers and their petrophysical heterogeneity. The static models were created combining a rule-based stratigraphic modelling approach with geostatistical modelling, and were applied in groundwater-flow and heat-transport simulations using MODFLOW-2005 and MT3D-USGS. Overall, histories of injected cold-water plume propagation were examined considering (i) three architectural frameworks as representative of different river morphodynamics, (ii) four scenarios of facies architecture, and (iii) alternative well layouts. The presence, size and spatial distribution of sedimentary heterogeneities related to river hydrodynamics, channel-form abandonment, or to modes of meander-transformation are seen to control the shape of the thermal plume, thereby affecting well-doublet performance. The considered scenarios of facies make-up for point-bar deposits have a modest influence on temperature decline near the abstraction well. The presence of sandstone beds in the lower heterolithic parts of abandoned-channel fill does not facilitate significant thermal-plume expansion beyond mud-plug. The effect of basal lags made of open-framework conglomerates on heat advection depends on their geometry, but is effectively negligible when it makes up less than 1% of deposits. Relatively thin mud drapes lying on point-bar accretion surfaces are seen to act as baffles to flow, but their impact is minimal in view of their small number. The study provides useful and novel insight on the potential impact of sedimentary heterogeneity in fluvial reservoirs, which can be applied to the design of well doublets, and highlights areas deserving further investigations.

DREAM: Debugging and Repairing AutoML Pipelines

Deep Learning models have become an integrated component of modern software systems. In response to the challenge of model design, researchers proposed Automated Machine Learning (AutoML) systems, which automatically search for model architecture and hyperparameters for a given task. Like other software systems, existing AutoML systems have shortcomings in their design. We identify two common and severe shortcomings in AutoML, performance issue (i.e., searching for the desired model takes an unreasonably long time) and ineffective search issue (i.e., AutoML systems are not able to find an accurate enough model). After analyzing the workflow of AutoML, we observe that existing AutoML systems overlook potential opportunities in search space, search method, and search feedback, which results in performance and ineffective search issues. Based on our analysis, we design and implement DREAM, an automatic and general-purpose tool to alleviate and repair the shortcomings of AutoML pipelines and conduct effective model searches for diverse tasks. It monitors the process of AutoML to collect detailed feedback and automatically repairs shortcomings by expanding search space and leveraging a feedback-driven search strategy. Our evaluation results show that DREAM can be applied on two state-of-the-art AutoML pipelines and effectively and efficiently repair their shortcomings.

Liquid plugs in narrow tubes: application to airway occlusion

We study liquid plugs in the pulmonary airways based on the two-phase axisymmetric weighted residual integral boundary-layer model of Dietze et al. (J. Fluid Mech., vol. 894, 2020, A17), which was originally developed to study liquid films coating the inner surface of a cylindrical tube in interaction with a core gas flow. The augmented form of this model, which was never applied beyond a proof of concept, allows for the representation of liquid pseudo-plugs. Here, we demonstrate its predictive power vs experiments and direct numerical simulations, in terms of the dynamics of plug formation and the characteristics of developed liquid plugs, such as their shape, flow field, speed and length, as well as the associated wall stresses and their spatial derivatives. In particular, we show that the augmented model allows us to establish a direct continuation path from travelling-wave solutions (TWS) to travelling-plug solutions (TPS). We then apply the model to predict mucus plugs in the conducting zone of the tracheobronchial tree, based on the lung architecture model of Weibel. We proceed by numerical continuation of travelling-state solutions in terms of the airway generation, whereby we impose the wavelength of the linearly most-amplified convective instability (CI) mode or that of the absolute instability (AI) mode. We identify the critical airway generation for liquid-plug formation (TWS/TPS transition), maximum potential for wall-stress-induced epithelial cell damage and CI/AI transition, and investigate how these phenomena are affected by the main control parameters, i.e. airway orientation vs gravity, air flow rate, mucus properties and airway size.

A Software Architecture Model to Manage Heterogeneous Data using Edge Computing in 5G Environment

With the improvement of 5G communication service, the need for edge computing has become more apparent. The Mobile Edge Computing service (MEC) has become one of the best solutions for mobile User Equipment (UE) data processing. MEC provides more context-aware local processing of UE's data. Meanwhile, the MEC needs to manage the heterogeneous nature of UEs and the type of data format they provide. Therefore, a Heterogeneous Data Service Bus is developed to manage different types of data formats and organize them in efficient manner. We propose a software architecture model to address the management of data heterogeneity within computing edges. The architecture discussion showcases its applicability and proper handling of heterogeneous data in the Edge Computing Environment.

AdaGIN: Adaptive Graph Interaction Network for Click-Through Rate Prediction

The goal of click-through rate (CTR) prediction in recommender systems is to effectively work with input features. However, existing CTR prediction models face three main issues. First, many models use a basic approach for feature combinations, leading to noise and reduced accuracy. Second, there is no consideration for the varying importance of features in different interaction orders, affecting model performance. Third, current model architectures struggle to capture different interaction signals from various semantic spaces, leading to sub-optimal performance. To address these issues, we propose the Adaptive Graph Interaction Network (AdaGIN) with the Graph Neural Networks-based Feature Interaction Module (GFIM), the Multi-semantic Feature Interaction Module (MFIM), and the Negative Feedback-based Search (NFS) algorithm. GFIM explicitly aggregates information between features and assesses their importance, while MFIM captures information from different semantic spaces. NFS uses negative feedback to optimize model complexity. Experimental results show AdaGIN outperforms existing models on large-scale public benchmark datasets.

Remediation of LWD data lag with hybrid real-time data using self-attention-based encoder-decoder model

This study aims to address the lag issue in Logging While Drilling (LWD) data, which is crucial for real-time decision-making in subsurface resource exploration. The primary objective is to enhance the accuracy of LWD measurements, which suffer from a positional discrepancy due to the tools being positioned several meters behind the drill bit. This lag can lead to delayed responses and misinterpretations during drilling operations. To achieve this, we introduce a novel Self-Attention-Based Encoder-Decoder (SABED) model that compensates for the time/depth lag by utilizing hybrid real-time data, including drilling engineering and LWD data. Our methodology involves training and validating the SABED model using data from the Volve field in the North Sea. The model's architecture is designed to effectively capture the complex relationships between drilling data and LWD measurements. Experimental results demonstrate that the SABED model can predict gamma ray values up to 45 m ahead with a Mean Relative Error (MRE) of less than 1.5% in the primary test well (F7), outperforming conventional sequential deep learning models. Further evaluations on an auxiliary test well (F10) indicate robust performance and generalization capabilities, even with noisy drilling data. The model also meets real-time operational requirements, processing predictions in approximately 4.5 s per 300-step interval (30 m) on both CPU and GPU. Notably, the SABED model maintains predictive accuracy despite data loss, using linear interpolation for missing segments. These findings underscore the SABED model's effectiveness in mitigating LWD data lag and its potential as a valuable tool for real-time geological information prediction. This research contributes novel insights to the field by providing an advanced methodology for improving data accuracy in LWD operations, thereby enhancing decision-making in the petroleum industry.

Computational modeling of Cs3Sb2I9-based novel architecture under WLED illumination for indoor photovoltaic applications

Computational modeling of Cs3Sb2I9-based novel architecture under WLED illumination for indoor photovoltaic applications

Textile Fabric Defect Detection Using Enhanced Deep Convolutional Neural Network with Safe Human–Robot Collaborative Interaction

The emergence of modern robotic technology and artificial intelligence (AI) enables a transformation in the textile sector. Manual fabric defect inspection is time-consuming, error-prone, and labor-intensive. This offers a great possibility for applying more AI-trained automated processes with safe human–robot interaction (HRI) to reduce risks of work accidents and occupational illnesses and enhance the environmental sustainability of the processes. In this experimental study, we developed, implemented, and tested a novel algorithm that detects fabric defects by utilizing enhanced deep convolutional neural networks (DCNNs). The proposed method integrates advanced DCNN architectures to automatically classify and detect 13 different types of fabric defects, such as double-ends, holes, broken ends, etc., ensuring high accuracy and efficiency in the inspection process. The dataset is created through augmentation techniques and a model is fine-tuned on a large dataset of annotated images using transfer learning approaches. The experiment was performed using an anthropomorphic robot that was programmed to move above the fabric. The camera attached to the robot detected defects in the fabric and triggered an alarm. A photoelectric sensor was installed on the conveyor belt and linked to the robot to notify it about an impending fabric. The CNN model architecture was enhanced to increase performance. Experimental findings show that the presented system can detect fabric defects with a 97.49% mean Average Precision (mAP).

Retriever: A view-based approach to reverse engineering software architecture models

Software systems, which have become an integral part of our daily lives, are constantly evolving and growing in complexity. The challenge of understanding and managing these systems has led to a shift towards automated extraction of models from software artifacts. However, extracting architectural models for web service or microservice systems is challenging due to the heterogeneity of formats, languages, and semantics. This challenge arises from the independent deployment and loose coupling of services, as well as the diverse technologies and platforms that comprise the heterogeneous artifacts. This paper addresses this issue with two key ideas: a knowledge representation model for diverse artifacts and a framework for integrating individual views into a unified architectural model. The former involves model-to-model transformations that consider technology-specific relationships and concepts. At the same time, the latter uses model transformation and composition techniques to create a unified model based on an existing metamodel. This work goes beyond the current state of the art by introducing an approach for reconstructing component-based software architectures and combining model-driven reverse engineering processes to capture information across multiple views. As empirical evidence of its applicability and accuracy, we report on its successful application in real projects of different sizes. Our evaluation shows scalability in generating architectural models and can reconstruct the components of a software system and their interactions with F1 values up to 0.94 for structural properties and up to 0.88 for behavioral properties.

Jaringan Saraf Tiruan Backpropagation untuk Prediksi Harga Bahan Pangan di Wilayah Kabupaten Deli Serdang

Foodstuffs are raw materials in the form of agricultural, vegetable and animal products that are used by the food processing industry to produce a food product. Food ingredients consist of plant foods and animal foods. Food is the most basic need for human resources in a country. Food prices sometimes experience erratic increases or decreases. The aim of this research is to determine the results of food price predictions in the Deli Serdang Regency area using the Backpropagation algorithm. The data used in this research is food price data from 2020 to 2023 which comes from the official National Food Ingredients website. This research uses the Backpropagation algorithm artificial neural network method which uses several architectural models and the results of this test will produce the best accuracy values. The test results show that the best architecture for research on implementing the backpropagation algorithm in predicting food prices in Deli Serdang Regency is 2-10-1 with an accuracy of 87.5% and the 2-3-8-1 architecture with an accuracy of 87.5%.