The integration of Artificial Intelligence (AI) and the Internet of Medical Things (IoMT) presents significant computational and engineering challenges, especially in the deployment of real-time diagnostic systems. Therefore, this study proposed a numerically optimized and computationally efficient framework that combines deep learning (DL) architectures with IoMT-enabled deployment for the automated detection and classification of brain tumors using Magnetic Resonance Imaging (MRI). The methodology revolves around rigorous numerical preprocessing techniques, including normalized resizing, advanced data augmentation, and computationally efficient feature extraction via both custom and pre-trained Convolutional Neural Networks (CNNs). The key contributions of this study are tailored toward the evaluation of algorithmic performance beyond diagnostic accuracy, incorporating metrics such as model convergence, inference latency, memory footprint, and numerical stability under varied input conditions. The proposed AI-IoMT framework, known as I-BRAINDETECT, implemented as a web-based IoMT platform, demonstrates how algorithmic design and computational modeling can address the limitations of real-time medical image analysis. Performance evaluation and comparative analysis have shown that EfficientNetB0 and DenseNet121 achieved optimal performance in binary classification with 98.5 ± 0.202% accuracy, while ResNet50 excelled in multiclass classification with 95.4 ± 1.01% accuracy, both within a computationally constrained IoT environment. Validation of the trained models on an external dataset (Figshare) has shown that DenseNet121 achieved the best result with 93.99% accuracy. This work underscores the necessity of numerical robustness and algorithmic efficiency in bridging AI and IoT for scalable biomedical engineering solutions.
Currently, technologies such as convolutional neural network (CNNs) and deep Q network (DQNs) are undergoing intensive research and rapid development, driving vigorous advancement in the field of artificial intelligence. Nevertheless, there is still room for improvement in addressing practical industrial problems and enhancing learning efficiency and accuracy. To address the core challenges of traditional evolutionary algorithms (EAs) in complex optimization problems, such as insufficient scalability, limited environmental adaptability, and low computational efficiency, this paper proposes an evolutionary algorithm optimization framework (DLRL-EAF) that fuses deep learning and reinforcement learning. To verify the effectiveness of the proposed method, six standard test functions (Sphere, Rastrigin, Griewank, etc.) and three practical engineering optimization problems (mechanical parts design, logistics path planning, photovoltaic array layout) are selected for comparison experiments. The performance of DLRL-EAF is evaluated using the standard genetic algorithm (SGA), the particle swarm optimization algorithm (PSO), and the adaptive genetic algorithm (AGA). Experimental results show that DLRLEAF improves the accuracy of optimal solutions by an average of 23.6%, accelerates iterative convergence by 31.2%, demonstrates greater stability in high-dimensional, complex problems, and improves scalability by more than 40%. At the same time, the proposed method significantly reduces the time and resource costs of problem-solving in practical engineering applications and demonstrates its practical value in industrial settings.OPEN ACCESS Received: 27/12/2025 Accepted: 26/03/2026
To address the multi-objective collaborative optimization of quality, energy consumption, and yield under dynamic conditions in the Portland cement combined grinding process, this paper proposes a novel algorithm, CGDS-LTL, based on cooperative game theory and temporal perception. First, a hybrid temporal model combining Linformer, TCN, and LSTM was developed to dynamically track process conditions in the Portland cement combined grinding process. Second, an optimization objective function was established, and a cooperative game theory framework was introduced to address the challenge of achieving multi-objective optimization, which could not be effectively solved with a single paretooptimal solution. Meanwhile, volatility metrics were used to quantify the adjustment range of operational variables, allowing for the dynamic optimization of decision constraints. This approach mitigated the deviation of the pareto front from current decision settings caused by high population randomness, ultimately identifying the optimal solution for the multiobjective collaborative optimization problem. Finally, experiments using real production data from a Portland cement plant demonstrated that, compared with NSGA-II and C-TAEA, the proposed method improved the hypervolume indicator by 95% and 33.3%, respectively, indicating a more uniform solution distribution and better convergence. This demonstrated the interpretability and effectiveness of the proposed framework for dynamic multi-objective optimization in Portland cement combined grinding.OPEN ACCESS Received: 14/04/2026 Accepted: 21/05/2026
The motivation behind the creation of new statistical models is primarily driven by the need to accurately describe complex data and related phenomena. This article introduces the arctan power half logistic distribution, a new and more flexible extension of the power half logistic distribution. The proposed model’s hazard rate function is highly versatile, capable of displaying decreasing, J-shaped, or reversed J-shaped patterns. We derive its key statistical properties and investigate its application to progressively Type-II censored data. Parameter estimation is conducted using both maximum likelihood and Bayesian frameworks, the latter incorporating informative and non-informative priors across multiple loss functions. Given the analytical intractability of the posterior distributions, we employ Markov Chain Monte Carlo techniques for numerical approximation. Monte Carlo simulations demonstrate that Bayesian point and interval estimates generally outperform frequentist approaches, maintaining coverage probabilities near 95%. Finally, the model’s superiority is validated using a real-world engineering dataset, where it consistently outperforms several established competing distributions.OPEN ACCESS Received: 21/02/2026 Accepted: 20/04/2026
The formulation of constrained families of probability distributions is crucial for analyzing data inherently limited to the unit interval, such as proportions, rates, and probabilities. These models have extensive applications across various domains, such as epidemiology, dependability, finance, and environmental research, where adaptable and resilient statistical instruments are essential for accurately representing intricate real-world phenomena. Motivated by this, our work presents the bounded Uma distribution (BUmD), an innovative probability model obtained by transforming the Uma distribution to the unit interval (0, 1). The suggested distribution maintains the flexibility of the Uma family while broadening its applicability to data represented as proportions, probabilities, and rates. We examine the essential statistical characteristics of the BUmD, encompassing its moments, quantile function, extropy, and reliability metrics, with a focus on its capacity to represent various hazard rate behaviors. Sixteen traditional and contemporary estimation techniques are employed to estimate the model parameters, and their efficacy is assessed through comprehensive Monte Carlo simulations using criteria such as bias, mean squared error, and goodness-of-fit measures. The simulation findings indicate that the maximum likelihood and maximum product of spacings estimators are the most efficient. The practical relevance of the BUmD is evidenced by its application to epidemiological and public health datasets, where it demonstrates enhanced flexibility and fitting performance relative to conventional models. The suggested distribution provides a robust and flexible tool for modeling bounded lifetime, reliability, and proportional data in the applied sciences.OPEN ACCESS Received: 07/04/2026 Accepted: 20/05/2026 Published: 21/07/2026
Due to increased system uncertainty, nonlinear dynamics, and marketdriven power exchanges, modern deregulated multi-area power systems with high penetration of renewable energy sources (RES) like wind and solar, and electric vehicle (EV) charging stations present serious challenges to automatic generation control (AGC). Degraded frequency regulation and tie-line power control under deregulated environments result from the limited robustness, slow dynamic response, poor handling of stochastic RES/EV variations, and susceptibility to local optimal solutions of both conventional PI/PID controllers and recently reported optimizationbased AGC schemes. However, with a view to addressing these issues, this study focuses on reducing area control errors (ACE) during different operational shifts, such as frequency fluctuation (f) and tie line variations (Ptie). The main objective of this study is to determine the optimal gain settings for the Fractional Order Proportional-Integral-Derivative controller (FOPIDC) using the mother optimization algorithm (MOA). The proposed control strategy looks at how generators 3-AMS behave in a deregulated environment and emphasises the significance of FOPIDC optimization in preserving system stability with the goal of reducing integral time and absolute error (ITAE). Furthermore, the effectiveness of the proposed method is verified by comparing it with the Walrus Optimization Algorithm (WOA). As case studies, the effectiveness of the suggested strategy is also evaluated under Poolco, bilateral agreements, stability, and sensitivity analysis. In terms of generator outputs, tie-line power variations, and frequencies across different locations, comparative data unequivocally demonstrate that the suggested MOA-adjusted FOPIDC performs better than alternative approaches. In case 1, by implementing the MOA optimized controller, the settling time is 8.5 s, and the value of the objective ITAE for the transient responses is 0.0005292. However, these values are less than the values obtained by WOA. Similarly, in case 2, the settling time is 11.5 s, and ITAE is 0.000395 less
Multi-object tracking (MOT) remains challenging in conditions involving occlusion, small objects, rapid motion, and crowding, wherein the accuracy of detection and the quality of association degrade simultaneously. We propose AdaptiveTrack, an online MOT framework featuring a closed-loop, confidence-aware association and recovery design: CSI-IoU adapts spatial overlap based on confidence and scale, EAMO refines similarity through density, velocity, and scale cues, and DCR updates detection confidence utilizing association context before NMS and assignment. A lightweight continuity module additionally preserves identities during missed detections. On MOT17/MOT20, AdaptiveTrack achieves HOTA 67.33 and 66.73, MOTA 82.55 and 78.30, and IDF1 83.20 and 82.57, operating at 23.5 FPS.
Real-time monitoring of oil-gas wellbore temperature profiles provides important support for drilling strategy adjustment and model parameter calibration. MEMS-based micro-measurers can collect fullwellbore temperature data by flowing with drilling fluid circulation, but converting their time-series records to well depth relies on idealized steady-motion assumptions that neglect wall collisions, sticking, and local flow disturbances, introducing systematic positioning errors. In this study, a deep learning-based time-depth conversion method is proposed. A gated recurrent unit (GRU)-based temporal neural network is developed to extract motion-state features from six-axis dynamic signals, and a bounded velocity correction mechanism is introduced to compensate deviations from the idealized terminal velocity. The results show that: (1) The proposed model effectively learns motion-state deviations from dynamic response data and provides stable velocity correction under complex downhole conditions. (2) The corrected time-velocity curve exhibits transient fluctuations relative to the idealized terminal velocity, accurately capturing non-ideal behaviors while preserving directional stability. (3) During model training, the loss function decreases progressively and stabilizes, indicating good convergence and training stability. (4) Comparative analysis shows that the proposed GRU-Full method achieves a mean absolute anchor-point error of 1.23 m (0.24% of the well depth), outperforming the MLP and LSTM alternatives. Ablation experiments confirm that both the physical constraints and regularization terms contribute substantially to the model accuracy. This study enhances the spatial mapping accuracy of temperature data acquired by micro-measurers under complex downhole dynamic conditions. The established physically constrained deep learning framework provides a new technical pathway for refined wellbore thermal-field characterization and intelligent drilling decision-making.
Energy consumption is an emerging concern in many fields, including information technology, particularly in data warehousing environments where Extract, Transform, Load (ETL) processes account for a significant portion of operational costs and resource utilization. Despite advances in hardware-level optimization, limited attention has been given to software-level energy optimization within ETL workflows. In reality, software is as important as hardware, and it is equally responsible for a decrease or increase in energy consumption. We argue that for modern applications in which energy efficiency is a priority, ETL processes should be optimally designed. This paper addresses this gap by proposing a Green ETL (GETL) approach designed to reduce energy consumption while maintaining high performance. The proposed method integrates transformation-level reuse through a shared transformation cache and adaptive parallel execution using Apache Spark, enabling efficient resource utilization and elimination of redundant computations. The proposed GETL removes unnecessary calculations and reduces both execution time and energy consumption, without requiring any modifications to the underlying data processing engines. To evaluate the effectiveness of the proposed GETL, experiments were conducted using the Transaction Processing Council Data Integration (TPC-DI) benchmark across multiple scale factors. The results demonstrate that the proposed approach achieves an average energy reduction of approximately 30%, with higher savings observed under large-scale workloads. In addition, GETL improves execution efficiency and reduces resource utilization compared to a traditional Spark-based ETL implementation.
Student-supervisor relationships (SSR) play a central role in postgraduate training, academic development, and research well-being. In the context of the rapid expansion of Chinese graduate education, understanding mentorship quality has become increasingly important for both educational governance and student development. However, prior SSR studies have often relied on small-scale surveys, context-specific qualitative evidence, or linear analytical approaches, which limits their ability to capture heterogeneous, non-linear, and system-level patterns across institutions and disciplines. To address this gap, we propose Interpretable Mentorship Analytics (IMA), a scalable analytical framework built on large-scale anonymous student evaluations. First, we construct a multi-platform dataset of anonymous supervisor evaluations and transform unstructured review text into structured mentorship indicators through a preprocessing and LLM-assisted feature engineering pipeline. Second, we employ gradient-boosting models, including XGBoost, LightGBM, and Gradient Boosting, to model the relationship between mentorship-related features and overall evaluation outcomes. Third, we apply explainable machine learning methods, particularly SHAP and LIME, to identify global feature importance, local decision patterns, and non-linear interactions among mentorship dimensions. The results show that IMA can effectively uncover the key drivers of mentorship satisfaction, especially the central role of teacher-student relationship quality, while also revealing substantial heterogeneity across regions, institution types, and disciplines. By combining large-scale anonymous evaluations with interpretable predictive modeling, this study provides a transparent and data-driven framework for evaluations with interpretable predictive modeling, this study provides a transparent and data-driven framework for understanding SSR and offers empirical evidence for improving postgraduate supervision and educational policy.
Due to the dynamics of industrial production, it is a great challenge to learn robust feature representations from industrial process data for building an accurate soft sensor model. The traditional variational autoencoder (VAE) can learn robust features that can adapt to the dynamic process better, but cannot be directly applied in soft sensor modeling. Therefore, a novel regression-based dynamic VAE (REGDVAE) is introduced in this paper. Firstly, the encoder of the DVAE, which is constructed by the graph attention and the convolutional neural networks, is utilized to obtain the robust spatio-temporal features. The decoder of the DVAE is responsible for reconstructing the input data via transposed convolution. Secondly, the Transformer is employed to capture the dynamic associations between the robust spatio-temporal features and corresponding outputs. Moreover, with the purpose of guaranteeing the confidence of the proposed method, the Gaussian loss function is used as the optimization target of the REG-DVAE to enhance the confidence of each predicted value. The proposed REG-DVAE is implemented in a real-world melt index modeling of the polypropylene production process. The results show that compared with other baseline methods, the REG-DVAE not only can achieve the best performance, but also can provide a confidence interval, which greatly enhances the credibility of the prediction results.
sup>Efficient service placement is a critical challenge in large-scale Internet of Things (IoT) environments, where fog computing must balance deployment cost and resource utilization under heterogeneous and dynamic conditions. To address this challenge, this paper proposes a hybrid metaheuristic approach that combines Rat Swarm Optimization (RSO) and Sunflower Optimization (SFO), leveraging the strong global exploration capability of RSO and the efficient local exploitation behavior of SFO. The proposed RSO–SFO framework integrates both strategies within a unified fitness function designed to minimize deployment cost while ensuring efficient allocation of fog resources. Extensive simulation results demonstrate that the proposed hybrid algorithm consistently outperforms state-of-the-art optimization techniques, including Grey Wolf Optimization (GWO), Particle Swarm Optimization (PSO), and the standalone RSO and SFO methods. Specifically, the RSO–SFO approach achieves a fitness improvement of 45.38%, reduces deployment costs by 43.71%, and maintains a high average resource utilization of 78.83%. These results confirm the effectiveness and robustness of the proposed hybrid strategy for optimal service placement in fog-based IoT environments.
This study explores the existence, uniqueness (EU), and Hyers-Ulam stability (HUS) for a class of weighted fractional Itô-Doob stochastic integral equations (WFIDSIEs). To establish these crucial properties, we utilize the Banach fixed-point theorem (BFPT) as a foundational tool, alongside a variety of essential mathematical inequalities. These inequalities offer valuable insights into the underlying structure and behavior of WFIDSIEs, particularly in the context of fractional and stochastic dynamics. By leveraging these techniques, we demonstrate the conditions under which solutions to these equations exist and are unique, as well as how small perturbations affect the stability of these solutions in the Hyers-Ulam sense. Our results contribute to a deeper understanding of the stability and solvability of WFIDSIEs, which are important in modeling complex systems exhibiting both stochasticity and memory effects. Additionally, these findings open avenues for further research into the robustness of solutions in fractional stochastic models across various applied disciplines, such as finance, engineering, and biology.OPEN ACCESS Received: 12/09/2025 Accepted: 11/11/2025 Published: 21/07/2026
A highly adaptable censoring framework, known as the unified TypeI progressive hybrid censoring scheme, has been recently introduced as an enhancement to the unified hybrid censoring approach. However, this censoring method has a significant limitation: it assumes that the removal pattern is fixed and established prior to conducting the experiment, which may lack practical applicability. This study presents, for the first time, the unified Type-I progressive hybrid censoring scheme incorporating binomial removal, which enables the random withdrawal of surviving units following each failure, making it more suitable for survival studies. Assuming that the underlying distribution of the test units follows the Nadarajah–Haghighi distribution, we examine both point and interval estimation problems for the model parameters, binomial parameter, and two key survival measures. Maximum likelihood estimation serves as the classical methodology for obtaining point estimates and approximate confidence intervals. For Bayesian estimation, we employ the squared error loss function in conjunction with Markov Chain Monte Carlo sampling techniques. A comprehensive simulation analysis is conducted to assess the performance of the various point and interval estimators. In addition, to demonstrate the practical relevance of the proposed methodologies in survival analysis, a real-world data application involving the death times of a group of 45 gastric cancer patients is investigated.OPEN ACCESS Received: 02/11/2025 Accepted: 11/12/2025 Published: 21/07/2026
The complexity of data necessitates the development of novel distributions. The introduction of new models allows us to enhance this data and remain contemporary. In this paper, we proposed and explored a new flexible distribution, referred to as the extended inverse Weibull model. This study presents a novel probability model, the extended inverse Weibull (E-IW) distribution. The recently introduced distribution combines the E-X approach with the inverse Weibull distribution. The novel distribution facilitates the evaluation of real-world data due to its analytical feasibility and applicability. The suggested distribution may accommodate several types of datasets. Numerous statistical characteristics of the proposed model were acquired. The features include quantile functions, ordinary moments, order statistics, and moment-generating functions. functions. The estimated parameters of the new model are determined using several estimation techniques, including maximum likelihood, least squares, weighted least squares, maximum product spacing, and the Bayesian method under the square error loss function. The simulation analysis performed on the suggested model validated the dependability and consistency of its parameters. Furthermore, actuarial measures were computed, along with simulation study for these actuarial metrics was executed. Five real data sets were taken from several sectors to illustrate the importance and usefulness of the proposed model. Our empirical findings underscore the significance of the recommended model as a flexible and reliable tool for statistical modeling, with implications for improving data-driven analyses and integrating parametric modeling into modern applications.OPEN ACCESS Received: 10/11/2025 Accepted: 27/01/2026 Published: 21/07/2026
The deployment of communication base stations establishes an efficient information transmission network; however, implementing deployment in mountainous areas with complex terrain remains amajor challenge. To address the issues of large topographical variations, dispersed villages, and low coverage efficiency, this study focuses on application issues, develops a mountainous deployment environment model that incorporates terrain elevation increments and village exclusion zones. On this basis, a Differential Evolution algorithm with Multiple Mutation Strategies (MSM-DE) is proposed to improve the balance between global exploration and local exploitation. The algorithm introduces a probabilistic multi-mutation mechanism that dynamically selects among several mutation strategies according to population diversity, and an adaptive parameter memory archive that guides the search toward promising regions.These modifications enhance both convergence speed and robustness in complex terrain optimization. Three objectives—coverage rate, village coverage satisfaction, and signal security—are combined into a weighted multi-objective function, and experiments are performed under two deployment scenarios (fixed and random village distributions).The results demonstrate that MSM-DE achieves significantly faster convergence and higher coverage performance than benchmark DE variants, validating that the proposed mutation synergy and adaptive parameter control effectively strengthen the algorithm’s optimization capability and stability in mountainous base station deployment.
DC motors are frequently utilized in industrial and automation applications where accurate speed control is crucial. Although conventional Proportional-Integral-and Derivative (PID) controllers are widely utilized, their constant gain values make them less effective in managing dynamic loads and disturbances. It’s difficult to get optimal transient and steady-state performance with traditional PID tuning methods. To overcome these limitations, more adaptable and dependable control systems are needed. This study introduces a novel control strategy by optimizing a Fractional-Order PID (FOPID) controller using the Ant Lion Optimization (ALO) method. Mathematical modeling is used to determine the DC motor’s transfer function. An ALO, a metaheuristic algorithm, is then implemented to improve five FOPID parameters using Integral Timeweighted Absolute Error (ITAE). The simulation is done in MATLAB-Simulink software. According to the findings, the enhanced ALO-FOPID controller decreased settling time (0.0728 s) and rising time (0.0455 s) when compared to the PID controller, which is taken as a reference. It is noted that the proposed ALO-tuned FOPID demonstrated enhanced response over the conventional methods, demonstrating the usefulness of bioinspired algorithms for precision control applications. The comparison of the proposed methodology is also done with other studies during different operating conditions. The results show that intelligent optimizationbased control in industrial systems is feasible, which aids in the creation of reliable and flexible automation solutions.
It is necessary to construct constrained families of probability distributions when dealing with data that is unit-limited by definition, such as rates, proportions, and probabilities. These models are widely used in fields such as ecology, epidemiology, reliability, and economics. To make sense of the complex real-world events, robust and adaptable statistical methods are required across all these fields. Unit power Shanker distribution (UPSD) is a novel bounded probability distribution that we introduce for this purpose. Key statistical properties, including the quantile function, reliability measures, moments, cumulative distribution function, and probability density function, are derived and studied in detail. Furthermore, the analytical links between the Tsallis entropy and other entropy measures, such as extropy, Shannon, Rényi, and Arimoto, are clarified. Maximum likelihood, Anderson-Darling, Cramér-von Mises, maximum product of spacings, least squares, and variants thereof are among the parameter estimation approaches that we examine. We assess the efficacy of several goodness-of-fit estimation methodologies for bias and mean squared error using a thorough Monte Carlo simulation study. The findings demonstrate that, in every case, the maximum likelihood estimation method outperforms the maximum product of spacings method. In various technical and scientific settings, the UPSD offers a robust and adaptable alternative for analyzing limited data.OPEN ACCESS Received: 07/04/2026 Accepted: 09/05/2026 Published: 21/07/2026
This study investigates the utility of the newly modified Kies–Rayleigh distribution for analyzing progressively first-failure censored samples. We develop both classical and Bayesian estimators for the distribution’s parameters and derived reliability measures, including the reliability function and hazard rate. Interval estimation is addressed via approximate confidence intervals and Bayesian credible intervals. Classical estimates are obtained numerically by solving the likelihood equations, whereas Bayesian inference is conducted through Markov chain Monte Carlo sampling from the posterior distribution. To assess and compare the performance of the classical and Bayesian procedures, we carry out a comprehensive simulation study under multiple experimental scenarios. We also consider the design problem of choosing an optimal progressive sampling plan and evaluate competing plans using four standard optimality criteria. We analyze a dataset comprising time-to-failure observations for turbocharged engines from a single model of diesel engine. The results highlight the flexibility of the modified Kies–Rayleigh model and provide guidance on estimation and design choices for progressively censored reliability data.OPEN ACCESS Received: 21/09/2025 Accepted: 07/11/2025 Published: 21/07/2026
This study introduces a novel coupled computational framework that seamlessly fuses the Isogeometric Boundary Element Method (IGABEM) with the Polynomial Chaos Expansion (PCE) to address transient heat conduction problems under parametric uncertainty. The proposed approach transcends traditional numerical paradigms by constructing a unified formulation that intertwines geometric precision, computational efficiency, and probabilistic rigor. At the heart of the framework lie three tightly integrated representations. First, Non-Uniform Rational B-Splines (NURBS) basis functions serve a dual purpose—they simultaneously embody geometric modeling and field discretization. This duality allows for the direct transformation of ComputerAided Design (CAD)-level geometric fidelity into the numerical domain, preserving intricate boundary characteristics while discretizing the governing boundary integral equations with remarkable exactness. Second, through the introduction of Bézier extraction, the computational overhead associated with NURBS basis evaluations is dramatically reduced. This operation reformulates complex NURBS structures into more manageable Bézier entities, unlocking substantial efficiency gains without compromising the inherent smoothness or precision of the geometric representation. Third, the framework incorporates Polynomial Chaos Expansion to systematically quantify and propagate uncertainties in material parameters. By coupling this stochastic representation with an advanced radial integration scheme, the method elegantly circumvents the need for conventional domain meshing while maintaining high accuracy in evaluating domain integrals. The result is a streamlined yet robust uncertainty quantification process embedded directly within the boundary element context. A series of numerical experiments corroborates the effectiveness of the proposed hybrid strategy. The findings reveal that the method not only preserves the celebrated advantages of IGABEM—such as geometric exactness and reduced-dimensional computation—but also extends its capability into the stochastic domain, enabling a comprehensive evaluation of how parameter uncertainties influence transient thermal responses. In essence, this integrated IGABEM–PCE framework bridges the gap between deterministic modeling and probabilistic analysis, yielding a unified, highfidelity numerical tool. Its capacity to deliver both geometric accuracy and uncertainty quantification positions it as an indispensable technique for complex engineering scenarios, particularly in thermal management, energy conversion, and high-precision design systems, where reliability under uncertainty is paramount.
