The development of effective treatments and the improvement of patient survival rates depend heavily on the timely and precise diagnosis of brain tumors. Nevertheless, the current deep-learning methods have trouble generalizing, are prone to overfitting with small datasets, and are imprecise in identifying the subtle boundaries of tumors, all of which could limit their application in clinical settings. To tackle these concerns, this paper has introduced an innovative BrainMorphNet (BMN) structure which can automatically detect and categorize brain tumors on MRI images completely, hence overcoming these problems. The first step of the approach is the application of data-specific preprocessing to clear the images of artifacts and improve their quality. Afterwards, the multilevel feature extraction generates the global and local structures of the tumor regions simultaneously. Lastly, the framework applies adaptive attention mechanisms to spatial and contextual features and thus improve them. There is a very strong multiclass separation between gliomas, meningiomas, pituitary tumors, and nontumor cases, and the whole setup is such that cancers can be diagnosed at a particular place. Our rigorous testing on the Brain Tumor MRI and Figshare datasets gave a model accuracy of 0.993, precision of 0.985, and F1 score of 0.988. The system is a very effective and time-saving tool for brain tumor analysis automation, and it is a significant step towards more intelligent medical imaging and diagnostic systems. These impressive numbers show that BMN has good generalization and clinical reliability.
This study designs, simulates, and optimizes an automated robotic vehicle for a smart cell fabrication plant. AISI 316 stainless steel was chosen for its strength and durability. Vehicle performance was evaluated across three layouts—dynamic path, grid-based, and free navigation—using Technomatix Plant Simulation. Structural and dynamic analyses were performed in ANSYS. Response surface methodology (RSM) optimized key performance indicators including collision rate, system responsiveness, handling error rate, and downtime from navigation errors. The RSM models showed high accuracy (R2 = 0.9732–0.9817). Optimization achieved a collision rate of 0.7936 collisions/h, 77.10
Soft sensing technology has recently attracted significant attention in industrial processes, particularly chemical processes, as it can ensure process safety and lay the foundation for optimizing operation by predicting product quality in real time. In the acetic acid synthesis process via acetaldehyde oxidation, real-time estimation of the composition of crude acetic acid, the product, is crucial for improving process safety and product quality. In this paper, we propose a distributed adaptive neuro fuzzy inference system (ANFIS) based on principal component analysis (PCA) and fuzzy C means (FCM) clustering, and apply it to the soft sensing of the acetic acid synthesis process. Simulation and experimental results demonstrate the effectiveness of the proposed soft sensing method using the distributed ANFIS.
The continuous need for increased data speeds and reduced latency in contemporary wireless communication systems, especially 5G and the upcoming 6G, requires the advancement of antennas with improved performance. Microstrip patch antennas (MPAs) are excellent choices because of their slim design, adaptability, and affordability. Nonetheless, traditional MPAs face fundamental limitations such as restricted bandwidth, minimal gain, and limited power handling capability. This paper introduces an innovative design of a high-efficiency MPA functioning in the 28 GHz 5G mm-wave spectrum. The suggested antenna features a Sierpiński carpet fractal structure for size reduction and multiband functionality, combined with a hexagonal split-ring resonator (H-SRR) metamaterial layer to markedly improve gain and bandwidth. A thorough evaluation utilizing a finite-difference time-domain (FDTD) approach, executed in Python, is presented to confirm the design. The simulated outcomes indicate a –10 dB impedance bandwidth of 4.2 GHz (15
Aiming at the problems of large trajectory offset, low trajectory tracking control accuracy and long control time in the traditional trajectory tracking control algorithm of robot bionic manipulator, a trajectory tracking control algorithm of robot bionic manipulator based on RBF neural network is proposed. Through the position description and orientation description of the robot bionic manipulator in space, the spatial coordinate transformation of the manipulator is carried out. Using the Lagrange method, the dynamic model is constructed through the kinetic energy and potential energy of the robot bionic manipulator. On this basis, the trajectory of the manipulator is tracked and controlled through the designed RBF neural network controller. The simulation results show that the trajectory of the robot bionic manipulator controlled by the proposed algorithm is consistent with the theoretical control trajectory, and the trajectory tracking control accuracy is high and the control time is short.
The variability of loads in near-zero-energy buildings with photovoltaic direct-drive inverter air-conditioning systems poses challenges for traditional load control methods, which often rely on simple models and fixed rules. This makes it difficult to optimize shutdown, duty cycles, and temperature control, affecting energy utilization and operational stability. To tackle these issues, we propose a cluster load control method utilizing improved modeling and algorithms tailored for these systems. This involves constructing detailed thermodynamic and electrical models to analyze load characteristics and establish effective control strategies. We also introduce a load control model based on an enhanced DEPSO algorithm, focusing on minimizing operational costs while meeting conditions. Experimental results indicate that this approach leads to stable, efficient operation and significantly reduces energy consumption and carbon emissions, demonstrating its effectiveness in optimizing air-conditioning cluster load control in near-zero-energy buildings.
A precise positioning method for the motion track of a multiaxis manipulator using PLC technology was developed to enhance workability and efficiency. The system integrates manual, zero finding, automatic, and semiautomatic operational modes, with an action sequence including pitch down, stop rotation, open claw, reach out, close claw, retract, reverse, pitch up, and stop. The control system, based on SIEMENS PLCS7-300, includes a PLC, touch screen, step drive, sensors, and a pneumatic servo positioning device. I/O port distribution was designed, and control code was created to coordinate the step motor and pneumatic devices, ensuring seamless operation. Experimental results indicate a success rate over 98.5
The fluid flow inside the natural gas cyclone gas-liquid separator is complex, and the difficulty in determining the liquid level switching gain leads to separator shaking and unstable liquid level control. To this end, a liquid level self correction control method based on ANFTSMC algorithm is studied: accurate liquid level change values are obtained using an algebraic sliding hybrid model based on flow field assumptions; Design a liquid level controller based on terminal sliding mode, achieve finite time convergence through nonsingular fast terminal sliding mode surface, and use adaptive law to estimate disturbance upper limit online to avoid variable singularity; Design adaptive sliding mode control law, adjust switch gain online to compensate for disturbances, reduce jitter, and stabilize liquid level; Introducing ANFTSMC algorithm to achieve liquid level self correction control. The experiment shows that this method adjusts the liquid level quickly, has a short rise time, and almost zero steady-state error, with good results.
The fertilizer discharge flow of fertilizer seeder is easily influenced by the change of the rotation speed of fertilizer discharge shaft, which leads to fertilizer waste and environmental pollution. Faced with this situation, it is of great practical significance to accurately control the fertilizer flow of fertilizer seeder. Therefore, a sectional control method of fertilizer flow of fertilizer seeder based on differential evolution algorithm is proposed. The method is divided into two parts. In the first part, the capacitive sensor is used to detect the fertilizer flow and the wavelet threshold method is used to denoise. In the latter part, PID subsection control model is studied on the basis of traditional PID control, and three parameters of the model are improved by differential evolution algorithm to realize subsection control of fertilizer flow of fertilizer seeder. The results show that the accuracy of the control method is above 93
As autonomous driving advances towards real-world implementation, accurately predicting the behavior of traffic participants remains a critical challenge for ensuring both safety and efficiency in driving. This study emphasizes the use of convolutional neural networks (CNN), employing long short-term memory (LSTM) and multilayer perceptron (MLP) as comparative models to investigate deep learning applications in behavior prediction. Utilizing Kaggle’s “prediction_for_validation_data.csv” dataset, the research develops a predictive system following data cleaning, feature extraction, and oversampling procedures. Experimental results indicate that CNN effectively extracts visual features through its convolution-pooling architecture, achieving an accuracy of 94.12
Social media, in the era of digital information, is a useful tool for communication. Text categorization, image mining, and evaluation of sentiment have accumulated a lot of attention, partially because of the amount of unstructured data on social networking. Many of these standards are derived from online sharing networks, which provide a vibrant social environment and gather images with cyberbullying posts. These clusters provide rich data that is logically helpful for text retrieval and image classification. It is shown that two important deep learning (DL) and artificial intelligence (AI) methodologies are beneficial for evaluating toxicity from social media interactions (SMI). This paper describes the many procedures involved in toxic word analysis, such as preprocessing, evaluating, and displaying social media content. Using DL approaches, such as bidirectional encoder representations from transformers (BERT) to assess toxic terms in social interactions, the text is first categorized. A BERT-based legislative text classification technique is suggested to identify the policy field defined in the text more precisely. The approach first vectorizes the sentence-level attributes of the legislative text using the BERT-trained language model. The resulting feature vector is then fed into the classifier to do the classification. BERT categorizes text into predetermined categories by first extracting characteristics from the text using filters. The second approach properly identifies photos from social networks using AI techniques. A weighted stack on the social network image association network was created using a neural network. The harmful photographs were then filtered using AI by combining three distinct types of social network photos: RGB (red, green, and blue), grayscale, and depth. Three-dimensional conventional neural networks (3D CNNs) were built to gather and filter social media online bullying posts in real-time during a crisis event to categorize cyberbullying images.
Vehicle network data is being created in massive quantities due to the Internet of Vehicles (IoV) rapid expansion. Network communication security is challenged by the volume of data. The significant amount of data created within the vehicle network presents time-consuming detection issues, even while intrusion detection technologies can help protect the system from unwanted attacks. In this manuscript, advancing connected vehicle security using advanced deep learning and optimized feature selection for intrusion detection (ACVS-OFSID-NCGNN) is proposed. The CIC-IDS-2017 dataset is where the data is first gathered. The gathered information is then sent to preprocessing. In prior to processing, to employ unsharp mask guided filtering (UMGF), identify the missing values, clean the data and standardization are carried out. Next, the previously processed data are provided to superb fairy-wren optimization algorithm (SFOA) for feature selection. SFOA selected 10 optimum features features from CIC-IDS-2017 information. The node-level capsule graph neural network (NCGNN) is then fed the chosen features to detecting the intrusion and classify as benign, brute force, DoS, portscan, web attack, bot, and infiltration. generally speaking, NCGNN doesn’t articulate how to modify optimization techniques to identify the best parameters to guarantee Internet of Vehicles. Therefore, the purpose of the house swallow optimizer (HSO) is to optimize the node-level capsule graph neural network, which correctly classifies intrusion detection. The proposed EID-IoV-NCGNN Python is used to implement this strategy. Using performance criteria such as accuracy, recall, FPR, precision, F1-score, and detection time, the effectiveness of the suggested approach was evaluated. The proposed NCGNN-HSO approach.
This research integrates deep physics-informed neural networks (PINNs) with Krupková’s geometrical theory to propose a novel framework for modeling and directing nonholonomic robotic systems. We derive reduced equations of motion for a differential-drive mobile robot under dynamic constraints, such as time-varying payloads and terrain-induced friction, using geometrical mechanics. Deep PINNs are employed to solve these nonlinear equations and develop an adaptive control framework for robust trajectory tracking and motion planning. The originality lies in the seamless fusion of classical geometrical mechanics with modern deep learning, while the novelty is demonstrated through the application of deep PINNs to address nonholonomic constraints and nonconservative effects. Under nominal, high payload, and high friction situations, extensive numerical comparisons with Runge–Kutta (RK4) confirm the method’s higher accuracy (36–60
Intelligent transportation systems (ITSs) are revolutionizing transportation by integrating advanced technologies to optimize efficiency, safety, and sustainability. Traffic prediction, a core component of ITS is quite important in improving traffic management, enhancing public safety, and promoting sustainable urban mobility. However, deep learning models’ efficacy in traffic prediction is heavily reliant on the quality of input data. Data noise, biases, and missing values can significantly hinder model accuracy by introducing spurious correlations that do not reflect true causal relationships. Additionally, imbalanced training data can skew the model’s learning process, leading to suboptimal performance and reduced generalization. Traditional feature selection methods often find it difficult to depict intricate, nonlinear interactions in high-dimensional datasets, further complicating the task. To overcome these obstacles, this research proposes a novel graph-GRU temporal fusion network (GGTFN) that combines graph neural networks (GNNs) and gated recurrent units (GRUs) to capture both spatial and temporal dependencies in traffic data. The model also incorporates regression-based imputation during preprocessing to handle missing data and uses SMOTE-NC to address class imbalance. Experimental findings indicate that the GGTFN performs better than other previous techniques, achieving the lowest RMSE values for short- to medium-term traffic predictions (15–30 min). Although performance decreases for longer prediction horizons, the proposed model demonstrates robustness and effective generalization across diverse traffic conditions. This research sets the stage for more precise and effective traffic management, helping to the creation of more intelligent and environmentally friendly transportation networks.
The paper proposes an approach for fast tuning the nonlinear attitude control system applied for a quadcopter. First, a mathematical model of quadcopter’s motion is estimated by a simple identification process using a low-cost test bench. After that, the controllers’ parameters are computed via optimization-based synthesis using simulation-in-loop framework. The proposed approach therefore does not require the development of a complex mathematical model built on the theories of aerodynamics, flight dynamics and DC motors, as well as geometry and mass/inertial properties of a quadcopter. There is also no need to apply any algorithm for nonlinear control system analysis. Due to its simplicity, the proposed approach can be used for quick synthesis of new control systems or adjusting the existing ones. The adequacy of the proposed approach is confirmed by bench studies of identification accuracy and control performance. Besides this, the paper also firstly describes the mathematical principle of the square-root controller implemented in well-known flight controller software like Ardupilot.
An insulator is one of the crucial parts of the high-voltage overhead cables. Leakage currents may circulate on the insulator’s surface as a result of external factors and contaminants adhered to the surface. Large leakage currents (LC) have the potential to produce flashover, heat losses and surface damage to the insulator. In order to prevent early flashover, this research offers a method that measures the severity of the insulator surface using harmonic measurements of leakage currents. In this work, on 11 kV polymer insulators, the leakage currents were assessed at various equivalent soluble deposit density (ESDD) levels. Discrete Wavelet transform topology is incorporated to extract features from the LC signal. Random forest (RF) with fuzzy inference system (FIS) using a diverse range of LC signals collected over an extended period is incorporated. From results, it is evident that the suggested pollution severity classifier, is highly effective with an accuracy about 95
The development of science and technology has promoted the unmanned aerial vehicle (UAV) industry. Due to its small size, lightweight, low cost, and other characteristics, UAV can integrate with multiple industries, and promote social development, which broadens the use of UAV itself. UAVs have been widely used in aerial photography, agriculture, and disaster rescue. This paper analyzed the application of UAV in geological disaster rescue. Using UAV remote sensing to photograph the roads to the geological disaster area, the road conditions of different roads could be analyzed, providing the best rescue route in the disaster area. The current point-feature-based methods fail to accurately identify and analyze the target in the UAV image. This paper proposed a convolution neural network (CNN) based model to analyze the UAV image by automatically identifying the image targets. We investigated the accuracy of vehicle recognition using traditional UAV image recognition and our CNN-based model. The experimental results showed that the proposed method improved the average recognition accuracy by 9.35 and 9.08
Underactuated mechanical systems (UMSs) present significant challenges in control design due to limited actuation, nonlinear coupling, and susceptibility to unmatched and time-varying disturbances. Traditional passivity-based methods often assume full state availability and matched disturbances, limiting their applicability in uncertain, sensor-constrained environments. This study proposes a learning-augmented interconnection and damping assignment passivity-based control (iIDA–PBC) framework that enhances robustness and adaptability for UMSs. The approach integrates real-time disturbance estimation using Gaussian Math. Comput. Appl. (GPR) and Math. Comput. Appl. (LSTM) networks, alongside nonlinear observer designs for reconstructing unmeasured actuator states. The overall control architecture preserves the port-Hamiltonian structure while enabling adaptive compensation for unknown external perturbations. Theoretical analysis ensures Math. Comput. Appl. (ISS) under bounded estimation errors. Simulation results on a benchmark underactuated system demonstrate improved disturbance rejection, tracking accuracy, and robustness compared to conventional adaptive and passivity-based controllers. The proposed method is suitable for complex, partially observable systems such as aerial vehicles, autonomous robots, and marine platforms.
This study introduces an innovative control strategy for a DC motor speed control system, referred to as a black box approach. The methodology involves system identification through two models: a fractional-order model, increasingly prevalent in identification processes, and a conventional integer-order model. The identification phase employs advanced heuristic optimization techniques, including particle swarm optimization (PSO), artificial bee colony (ABC), ant colony optimization (ACO), and genetic algorithms (GA). The subsequent control phase utilizes a fractional-order proportional-integral-derivative (FOPID) controller, widely recognized as a leading fractional-order control solution. The FOPID controller is characterized by five parameters: proportional gain ( K_p ), integral gain ( K_I ), derivative gain ( K_d ), integral order ( λ ), and derivative order ( μ ). These parameters are optimized using the aforementioned algorithms to enhance the performance of the fractional-order model, particularly in terms of energy efficiency during the stabilization phase, compared to the traditional integer-order model. Performance is evaluated using the integral of time and absolute error (ITAE) criterion, alongside standard metrics such as overshoot and settling time. A robustness test, involving parameter variations, is conducted to validate the effectiveness of the proposed approach. Experimental validation is performed using Matlab/Simulink interfaced with an Arduino Uno board to determine the parameters of a real-world DC motor. The results demonstrate that fractional calculus significantly enhances the efficacy of DC motor speed control.
The paper presents an implementation of an algorithm for finding graph centers. A characteristic feature of the problem statement is the lack of general information about the structure of the graph, the number of its vertices, and the number of edges incident to a particular vertex. A vertex, in this case, is a separate “entity” about which it is known how many edges are incident to it and what their weight is. The only requirements imposed on the graph are connectivity, nonnegativity of edge lengths, and their nondirectivity. The essence of the algorithm lies in the local exchange of “messages” between the graph nodes, which form the weight of each of them: a value identical to the distance from it to the most distant node of the graph.
