Damage identification and localization of pultruded FRP composites based on convolutional recurrent neural network and metaheuristic intelligent algorithms DOI

Xinquan Chang,

Xin Wang, Zhili He

и другие.

Polymer Composites, Год журнала: 2025, Номер unknown

Опубликована: Апрель 9, 2025

Abstract Fiber‐reinforced polymer (FRP) tendons are preferred in civil engineering for their lightweight properties, high strength, corrosion resistance, and electrical insulation. However, initial defects that arise during material preparation can adversely affect the mechanical performance service life of structures. Local identification technology is inadequate FRP products with variable thickness cross‐sections, especially tendons, resulting low detection efficiency. This article presents an innovative inverse problem‐solving framework aimed at simultaneously identifying location severity through frequency change rates. A convolutional recurrent neural network (CRNN) model was developed to establish mapping between rates associated damage information, including severity. The CRNN model's database generated from finite element models (FEM), which were validated against Euler beam vibration theory, demonstrating absolute error less than 1%. trained using this optimized data matrix reconstruction, refinement, dilated convolution, achieving a mean (Mae) 0.115% predicting rate. significantly surpassed CNN (0.318%), MLP (0.274%), LSTM (0.334%) models. served as surrogate problem, addressed Slime Mold Algorithm (SMA) model. prediction SMA under 0.5%, notably better FEM. Consequently, identifies defects' offering valuable insights applications various products. Highlights achieved MAE rates, 41.6% MLP. Optimized identified 97.8% accuracy. Hammering method effectively excited first 8 frequencies tendons. Experimental theoretical errors FEM analysis stayed below

Язык: Английский

Influence of treatment and fly ash fillers on the mechanical and tribological properties of banana fiber epoxy composites: experimental and ANN-RSM modeling DOI

S. Saravanakumar,

S. Sathiyamurthy,

G. Selvakumar

и другие.

Composite Interfaces, Год журнала: 2025, Номер unknown, С. 1 - 33

Опубликована: Янв. 6, 2025

This research examines the impact of chemical treatment and banana fly ash fillers on mechanical, tribological, water absorption characteristics fiber-reinforced epoxy composites. Alkaline enhanced fiber-matrix adhesion, markedly improving mechanical characteristics. The optimal performance occurred at 10% content, yielding a tensile strength 40.25 MPa, flexural 77.23 an 44.82 kJ/m2. Water studies indicated decline in moisture uptake, reducing from 40% untreated composites to 25% containing 15% ash, due bonding fewer voids. Tribological experiments demonstrated decrease Specific Wear Rate (SWR) Coefficient Friction (COF) with elevated concentration, signifying improved wear resistance. Predictive modeling Artificial Neural Networks (ANN) showed accuracy (mean error: 0.9584% for SWR, 0.50265% COF). RSM optimization identified input parameters minimizing SWR COF: sliding velocity 5.14491 m/s, distance 652.05 m, content 12.6236%, minimum COF values 15.63 × 10− 5 mm3/Nm 0.242241, respectively. SEM analysis confirmed that treated fibers minimized crack propagation while fracture toughness. results underscore promise ash-filled automotive, aerospace, structural applications necessitate moisture-resistant

Язык: Английский

Процитировано

3
Enhancing mechanical, degradation, and tribological properties of biocomposites via treatment and alumina content DOI

S. Saravanakumar,

S. Sathiyamurthy,

Ravikumar Natarajan

и другие.

Journal of Reinforced Plastics and Composites, Год журнала: 2025, Номер unknown

Опубликована: Фев. 17, 2025

The objective of this study is to investigate the effects alumina filler content and NaOH-treated Roselle fibers on mechanical, thermal, biodegradation, tribological properties while identifying optimal conditions for eco-friendly applications. Compression molding was employed fabricate composites, results revealed significant improvements in performance with chemical treatment content. Mechanical testing showed that 10% composite exhibited highest tensile, flexural, impact strengths due enhanced interfacial bonding uniform dispersion. Thermal analysis demonstrated improved stability, offering best thermal degradation resistance. Biodegradation studies indicated slower weight loss alumina-filled highlighting their environmental durability. Tribological evaluations achieved lowest specific wear rate (SWR) coefficient friction (COF), supported by SEM showing minimal debris surface damage. Optimization using a simulated annealing algorithm identified ideal (sliding velocity: 6.6 m/s, sliding distance: 500.33 m, content: 10.62%) minimized SWR (13.28 × 10⁻⁵ mm³/Nm) COF (0.278). These findings provide valuable insights into fiber composites sustainable applications automotive packaging industries.

Язык: Английский

Процитировано

2
Assessment of mechanical, thermal, and sliding wear performance of chemically treated alumina‐filled biocomposites using machine learning and response surface methodology DOI Creative Commons

V.S. Shaisundaram,

S. Saravanakumar,

Rajesh Mohan

и другие.

Polymer Composites, Год журнала: 2025, Номер unknown

Опубликована: Март 14, 2025

Abstract This study examines how NaOH treatment and alumina filler affect the mechanical properties, water absorption, thermal degradation, sliding wear of epoxy composites reinforced with pineapple leaf fiber. greatly improved composites' tensile, flexural, impact strengths by strengthening bond between fiber matrix. Furthermore, incorporation further elevated properties. The composite 10% showed peak values 41.4 MPa in tensile strength, 63.8 flexural 37.6 kJ/m 2 strength. Because hygroscopic parts were removed from treated composites, they absorbed much less water. 15% had lowest absorption at 18% after 192 h. Thermal degradation analysis that stability, having highest char residue (15.3%) 700°C. Sliding tests reinforcement significantly reduced specific rate (SWR) coefficient friction (COF). an SWR 0.2598 × 10 −5 mm 3 /Nm a COF 0.103 when 120 cm/s, 45 N load over 1500 m distance. A scanning electron microscopy found untreated experienced severe abrasive wear, while exhibited mild adhesive wear. shows treating PALF adding enhance their mechanical, thermal, tribological making them suitable for high‐performance industrial applications. Highlights Alumina (41.4 MPa) strength (63.8 MPa). NaOH‐treated moisture, enhancing durability. stability improved, 15.3% 700°C alumina. Optimized achieved (0.2598 /Nm). Artificial neural network response surface methodology accurately predicted optimized behavior.

Язык: Английский

Процитировано

1
Optimization and Comparative Analysis of Machining Performance of Al–Cu–SiC–GNP Composite: Influence of Reinforcement Variations Using Machine Learning, RSM, and ANOVA Validation DOI Open Access

Madduri Rajkumar Reddy,

Santhosh Kumar Gugulothu,

Talari Krishnaiah

и другие.

Advanced Engineering Materials, Год журнала: 2025, Номер unknown

Опубликована: Фев. 14, 2025

This study aims to optimize and analyze the machinability of Al–Cu–SiC–GNP composites using advanced techniques such as machine learning, (RSM), (ANOVA). The are fabricated an ex situ stir casting process with varying reinforcement percentages silicon carbide (SiC) graphene nanoplatelets (GNP) (2, 3, 5%), their is evaluated during water jet machining. key parameters analyzed material removal rate, surface roughness ( R a ), kerf width. Experimental findings reveal that significantly influence machinability. Optimal results achieved 5% SiC, 3% GNP, 300 MPa, 120 mm min −1 , balancing enhanced mechanical properties efficient ML models, including decision tree, random forest, support vector machine, artificial neural network (ANN), applied predict machining outcomes. Among these, ANN model exhibits highest predictive accuracy, capturing complex nonlinear interactions between input parameters. also validates through RSM ANOVA, confirming statistical significance on research provides robust framework for optimizing hybrid composite offers valuable insights into relationship content, parameters, performance outcomes, making it highly applicable aerospace automotive.

Язык: Английский

Процитировано

0
Advanced ensemble machine learning and response surface methodology for optimizing and predicting tribological performance of CMT-WAAM fabricated Al5356 alloy DOI

M.K. Nagarajan,

Manikandan Arumugam

International Journal on Interactive Design and Manufacturing (IJIDeM), Год журнала: 2025, Номер unknown

Опубликована: Март 20, 2025

Язык: Английский

Процитировано

0
Synergistic Effects of NaOH Treatment and Ceramic Fillers on the Mechanical and Tribological Behavior of Roselle Fiber-Reinforced Epoxy Composites DOI

S. Saravanakumar,

S. Sathiyamurthy,

Ravikumar Natarajan

и другие.

Fibers and Polymers, Год журнала: 2025, Номер unknown

Опубликована: Март 24, 2025

Язык: Английский

Процитировано

0
Damage identification and localization of pultruded FRP composites based on convolutional recurrent neural network and metaheuristic intelligent algorithms DOI

Xinquan Chang,

Xin Wang, Zhili He

и другие.

Polymer Composites, Год журнала: 2025, Номер unknown

Опубликована: Апрель 9, 2025

Abstract Fiber‐reinforced polymer (FRP) tendons are preferred in civil engineering for their lightweight properties, high strength, corrosion resistance, and electrical insulation. However, initial defects that arise during material preparation can adversely affect the mechanical performance service life of structures. Local identification technology is inadequate FRP products with variable thickness cross‐sections, especially tendons, resulting low detection efficiency. This article presents an innovative inverse problem‐solving framework aimed at simultaneously identifying location severity through frequency change rates. A convolutional recurrent neural network (CRNN) model was developed to establish mapping between rates associated damage information, including severity. The CRNN model's database generated from finite element models (FEM), which were validated against Euler beam vibration theory, demonstrating absolute error less than 1%. trained using this optimized data matrix reconstruction, refinement, dilated convolution, achieving a mean (Mae) 0.115% predicting rate. significantly surpassed CNN (0.318%), MLP (0.274%), LSTM (0.334%) models. served as surrogate problem, addressed Slime Mold Algorithm (SMA) model. prediction SMA under 0.5%, notably better FEM. Consequently, identifies defects' offering valuable insights applications various products. Highlights achieved MAE rates, 41.6% MLP. Optimized identified 97.8% accuracy. Hammering method effectively excited first 8 frequencies tendons. Experimental theoretical errors FEM analysis stayed below

Язык: Английский

Процитировано

0