Intelligent Fault Diagnosis Across Varying Working Conditions Using Triplex Transfer LSTM for Enhanced Generalization DOI Creative Commons

Misbah Iqbal,

C.K.M. Lee, K. L. Keung

et al.

Mathematics, Journal Year: 2024, Volume and Issue: 12(23), P. 3698 - 3698

Published: Nov. 26, 2024

Fault diagnosis plays a pivotal role in ensuring the reliability and efficiency of industrial machinery. While various machine/deep learning algorithms have been employed extensively for diagnosing faults bearings gears, scarcity data limited availability labels become major bottleneck developing data-driven approaches, restricting accuracy deep networks. To overcome limitations insufficient labeled domain shift problems, an intelligent, approach based on Triplex Transfer Long Short-Term Memory (TTLSTM) network is presented, which leverages transfer fine-tuning strategies. Our proposed methodology uses empirical mode decomposition (EMD) to extract pertinent features from raw vibrational signals utilizes Pearson correlation coefficients (PCC) feature selection. L2 regularization utilized mitigate overfitting problem improve model’s adaptability diverse working conditions, especially scenarios with data. Compared traditional such as TCA, BDA, JDA, demonstrate accuracies range 40–50%, our model excels identifying machinery minimal by achieving 99.09% accuracy. Moreover, it performs significantly better than classical methods like SVM, RF, CNN-based networks found literature, demonstrating improved performance fault under varying conditions proving its applicability real-world applications.

Language: Английский

A deep learning approach: physics-informed neural networks for solving a nonlinear telegraph equation with different boundary conditions DOI Creative Commons
Alemayehu Tamirie Deresse, Alemu Senbeta Bekela

BMC Research Notes, Journal Year: 2025, Volume and Issue: 18(1)

Published: Feb. 19, 2025

The nonlinear Telegraph equation appears in a variety of engineering and science problems. This paper presents deep learning algorithm termed physics-informed neural networks to resolve hyperbolic telegraph with Dirichlet, Neumann, Periodic boundary conditions. To include physical information about the issue, multi-objective loss function consisting residual governing partial differential initial conditions is defined. Using multiple densely connected networks, feedforward proposed scheme has been trained minimize total results from function. Three computational examples are provided demonstrate efficacy applications our suggested method. Python software package, we conducted several tests for various model optimizations, activation functions, network architectures, hidden layers choose best hyper-parameters representing problem's optimal solution. Furthermore, using graphs tables, approach contrasted analytical solution literature based on relative error analyses statistical performance measure analyses. According results, method effective resolving difficult non-linear issues

Language: Английский

Citations

0
The Impact of RObotic Assisted Rehabilitation on Trunk Control in Patients with Severe Acquired Brain Injury (ROAR-sABI) DOI Creative Commons
Letizia Castelli, Claudia Loreti,

Anna Maria Malizia

et al.

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(5), P. 2539 - 2539

Published: Feb. 26, 2025

Daily activities require balance and control posture. A severe Acquired Brain Injury (sABI) disrupts movement organization, execution, affecting trunk balance. Trunk therapy for difficult patients requires known novel methods. This study analyzes how hunova® robotic platform affects sABI patients’ sitting control. Twenty-six were randomized into the experimental group (HuG) that employed in addition to traditional (CoG) received only conventional rehabilitation. Clinical assessments performed trunk, balance, cognitive motor performance, disability, autonomy, quality of life, fatigue. Both static dynamic assessed with hunova®. HuG CoG significant intragroup analysis. Intergroup comparisons showed substantial differences control, affected side function, Only improved statistically instrumental assessment Between-group analysis a difference emerged COP path movement. The found effectiveness adaptability rehabilitation, showing improvement life fatigue patients. Registration: NCT05280587.

Language: Английский

Citations

0
Development of a Bayesian Network-Based Parallel Mechanism for Lower Limb Gait Rehabilitation DOI Creative Commons
H. L., Yanping Bao, Chao Jia

et al.

Biomimetics, Journal Year: 2025, Volume and Issue: 10(4), P. 230 - 230

Published: April 8, 2025

This study aims to address the clinical needs of hemiplegic and stroke patients with lower limb motor impairments, including gait abnormalities, muscle weakness, loss coordination during rehabilitation. To achieve this, it proposes an innovative design method for a rehabilitation training system based on Bayesian networks parallel mechanisms. A network model is constructed expert knowledge structural mechanics analysis, considering key factors such as scenarios, motion trajectory deviations, goals. By utilizing characteristics mechanisms, we designed device that supports multidimensional correction. three-dimensional digital developed, multi-posture ergonomic simulations are conducted. The focuses quantitatively assessing kinematic hip, knee, ankle joints while wearing device, establishing comprehensive evaluation includes range (ROM), dynamic load, optimization matching trajectories. Kinematic analysis verifies reasonable, aiding in improving patients’ gait, enhancing strength, restoring flexibility. achieves personalized goal through probability updates. mechanisms significantly expands joint motion, hip sagittal plane mobility reducing thereby validating notable effect

Language: Английский

Citations

0
Inverse kinematics solution for a six-degree-of-freedom upper limb rehabilitation robot using deep learning models DOI Creative Commons
Muhammad Faizan Shah, Naveed Ahmad Khan, Prashant K. Jamwal

et al.

Neural Computing and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: April 23, 2025

Language: Английский

Citations

0
How Do Humans Recognize the Motion Arousal of Non-Humanoid Robots? DOI Creative Commons

Qisi Xie,

Zihao Chen, Ding-Bang Luh

et al.

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(4), P. 1887 - 1887

Published: Feb. 12, 2025

As non-humanoid robots develop and become more involved in human life, emotional communication between humans will common. Non-verbal communication, especially through body movements, plays a significant role human–robot interaction. To enable to express richer range of emotions, it is crucial understand how recognize the movements robots. This study focuses on underlying mechanisms by which perceive motion arousal levels It proposes general hypothesis: Human recognition robot’s based perception overall motion, independent mechanical appearance. Based physical constraints, are divided into two categories: those guided inverse kinematics (IK) constraints forward (FK) constraints. Through literature analysis, suggested that amplitude has potential be common influencing factor. Two psychological measurement experiments combined with PAD scale were conducted analyze subjects’ expression effects different types at various amplitudes. The results show can used for expressing across Additionally, FK end position also certain impact. validates hypothesis paper. patterns roughly same robots: degree corresponds closely arousal. research helps expand boundaries knowledge, uncover user cognitive patterns, enhance efficiency

Language: Английский

Citations

0
Transformer-Based Approach for Predicting Transactive Energy in Neurorehabilitation DOI Creative Commons
Naveed Ahmad Khan, Tanishka Goyal, Fahad Hussain

et al.

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2024, Volume and Issue: 33, P. 46 - 57

Published: Dec. 11, 2024

Advancements in robotic neurorehabilitation have made it imperative to enhance the safety and personalization of physical human-robot interactions (pHRI). Estimation management energy transfer between humans robots is essential for enhancing during rehabilitation. Traditional control methods, which rely on coordinate-based monitoring robot velocity external forces, often fail unstructured environments due their susceptibility sensor noise limited adaptability individual patient needs. This paper introduces concept transactive energy, a coordinate-invariant entity that captures dynamics human robot-assisted rehabilitation can be used personalized control. However, estimation such complex process therefore, we developed transformer-based model predict potential energy. The proposed implemented an ankle compliant parallel provides required three rotational degrees freedom (DOF). learns from data obtained experiments carried out using with five stroke patients two types controllers: impedance controller operated zero mode trajectory tracking controller. study baseline, future research energy-based mechanisms pHRI applications, by utilizing advanced deep learning models.

Language: Английский

Citations

2
Intelligent Fault Diagnosis Across Varying Working Conditions Using Triplex Transfer LSTM for Enhanced Generalization DOI Creative Commons

Misbah Iqbal,

C.K.M. Lee, K. L. Keung

et al.

Mathematics, Journal Year: 2024, Volume and Issue: 12(23), P. 3698 - 3698

Published: Nov. 26, 2024

Fault diagnosis plays a pivotal role in ensuring the reliability and efficiency of industrial machinery. While various machine/deep learning algorithms have been employed extensively for diagnosing faults bearings gears, scarcity data limited availability labels become major bottleneck developing data-driven approaches, restricting accuracy deep networks. To overcome limitations insufficient labeled domain shift problems, an intelligent, approach based on Triplex Transfer Long Short-Term Memory (TTLSTM) network is presented, which leverages transfer fine-tuning strategies. Our proposed methodology uses empirical mode decomposition (EMD) to extract pertinent features from raw vibrational signals utilizes Pearson correlation coefficients (PCC) feature selection. L2 regularization utilized mitigate overfitting problem improve model’s adaptability diverse working conditions, especially scenarios with data. Compared traditional such as TCA, BDA, JDA, demonstrate accuracies range 40–50%, our model excels identifying machinery minimal by achieving 99.09% accuracy. Moreover, it performs significantly better than classical methods like SVM, RF, CNN-based networks found literature, demonstrating improved performance fault under varying conditions proving its applicability real-world applications.

Language: Английский

Citations

0