Malnutrition is a medical disorder affected by deficiencies, excesses, or imbalances in nutritional components resulting from an unbalanced intake of vital nutrients compared to the body's needs. It includes both under nutrition, in which growth and development are hampered by insufficient consumption of calories, vital nutrients along with over nutrition, which is defined as an excessive calorie intake that result in obesity as well as related health problems. The study demonstrated the effectiveness of the suggested Dynamic Sailfish Optimized Advanced Support Vector Machine (DSFO-ASVM) for childhood malnutrition prediction. DSFO-ASVM addresses the specified problem and overcoming the obstacles related to the specific scenario in the examination. The proposed study utilized the publicly available Kaggle Childhood Malnutrition Health Dataset, which contains 15,000 records of children aged 1–59 months with 40 attributes, including 39 input features and one target variable representing four nutritional status categories: Normal, Underweight, Stunted, and Wasted. The preprocessed dataset using z-score normalization technique and employing the Principal Component Analysis (PCA) to extract the features. The proposed DSFO-ASVM was implemented in Python and evaluated using an 70:30 train–test split with 5-fold cross-validation, achieving 97.70
Nonlinear variants of several unconventional computing models have been developed with interesting results in real-world applications. We introduce a novel variant of the young computing paradigm of virus machines (VM): nonlinear parallel virus machines (NPVM). We provide a comprehensive overview of the architecture of NPVMs and discuss their advantages and drawbacks. NPVMs not only achieve a major leap in efficiency due to the new parallelism behavior, but now the hosts can also have a non-negative real amount of viruses. For this propose, a new type of weight in the channels between hosts is introduced where the amount of viruses consumed and replicated depends on nonlinear functions. The computational power of NPVM is established; more precisely, the computational completeness of NPVM is proved as number-generating devices. Finally, a small universal NPVM is constructed for function computing with 9 hosts and 23 instructions.
Multi-objective optimization involves the simultaneous optimization of multiple objectives, which are often conflicting. The multi-objective evolutionary algorithm based on decomposition (MOEA/D) has emerged as a prominent and successful technique in this domain, and in recent years has garnered significant research interest due to its notable performance and its advantages in terms of addressing multi-objective optimization problems. Furthermore, there has been a growing focus on the use of this successful approach to solve real-world challenges. Despite its commendable performance, however, a closer examination of MOEA/D reveals opportunities to enhance its search capabilities and convergence speed through the incorporation of an elitist resource allocation (ERA) mechanism and a novel crossover strategy. The algorithm proposed here, called MOEA/D-ERA, incorporates these enhancements and is rigorously tested on a comprehensive suite of 40 benchmark functions alongside established state-of-the-art algorithms. The empirical findings unequivocally show that our innovations enhance the structural integrity of the MOEA/D framework and have a significant positive impact on the overall performance of the algorithm. Having demonstrated its efficacy in regard to test functions, MOEA/D-ERA is applied to address a structural multi-objective optimization problem. The proposed algorithm effectively determines a solution space consisting of optimal design inputs for an I-beam, with the objectives of minimizing both the cost and the vertical deflection as outputs. In summary, this study introduces a novel decomposition-based algorithm that incorporates an elitist selection mechanism and an innovative crossover model. To the best of our knowledge, this is the first instance in which a previously solved structural design problem has been redefined and addressed as a multi-objective problem.
To solve the limitation of classical deep learning models for fault diagnosis towards unknown faults, the capsule network with strong interpretability and unique inter-layer calculations is applied to the bearing fault diagnosis. In this paper, an efficient hyperparameter-based capsule network (HP-CapsNet) is proposed, which significantly reduces the number of parameters and computations to enhance training efficiency in the model training process. Meanwhile, the threshold judgment method is proposed to improve the classifier structure of the HP-capsule network, which realizes the diagnosis of unknown faults. Results of rolling bearing fault diagnosis show that HP-CapsNet can effectively reduce the computation complexity while the fault diagnosis accuracy reaches 100
Steel–concrete composite box girder bridges have been used in large and medium-span girder bridges due to many advantages, such as large torsional stiffness, high bearing capacity and superior economic index. However, interface slip and web shear deformation of curved steel–concrete composite box girder bridges, coupled with significant flexural and torsion coupling effect of curved composite box girder bridges under eccentric load, lead to complex mechanical behavior of curved steel–concrete composite box girder bridges. The section no longer satisfies the assumption of a flat section. Aiming at the complex mechanical behavior of curved steel–concrete composite box girder bridges, especially the interface interaction mechanism and deformation characteristics, a multi-scale-space grillage modeling analysis method was proposed to explore the spatial mechanical effects of curved steel–concrete composite box girder bridges, which can realize the overall macroscopic mechanical characteristics of curved steel–concrete composite box girder bridges and local fine mechanical behavior analysis. The macro model is modeled by the space grillage method to consider the overall spatial mechanical properties. The multi-scale model was used to simulate and analyze the local main stress parts. Point-plane contact elements based on volume constraint are used to connect the macro-model of the space grillage method with the fine-model of the multi-scale method to ensure the coordinated deformation of the multi-scale and space grillage models at the interface, and to realize the direct transfer between the overall macro-model and the local fine-model of the composite beam. Taking the typical curved steel–concrete composite box girder bridge as the research object, the finite multi-scale-space grillage fine element model is established, and the model is verified. Based on the verified model, the static and dynamic behaviors of curved steel–concrete composite box girder bridges are investigated, and the interface interaction mechanism, deformation characteristics and influencing factors of curved steel–concrete composite box girder bridges are revealed. The results show that there is asymmetric slip at the interface of curved steel–concrete composite box girder bridge, and the asymmetric slip effect seriously affects the flexural and torsional coupling of the structure and the deformation characteristics in and out of the plane. Meanwhile, it also shows that the multi-scale-space grillage modeling analysis method proposed in this paper has the advantages of high calculation accuracy, low cost and high efficiency. It provides a new idea and strategy for the structural mechanical behavior simulation analysis of complex curved steel–concrete composite beam bridges.
