Optimization and prediction of machining parameters in nanoparticle-reinforced FMLs using AI techniques

ABSTRACT

This study focuses on optimizing and predicting the drilling performance of Fiber Metal Laminates (FMLs) reinforced with BaSO₄ nanoparticles, achieved by adjusting parameters like spindle speed, feed rate, and tool diameter. Key responses—thrust force, torque, delamination, and surface roughness—were evaluated to enhance machinability. Using Central Composite Design, optimal parameters were identified: a spindle speed of 3000 rpm, feed rate of 10 mm/min, and tool diameter of 6 mm. Under these conditions, thrust force decreased by 51.92%, surface roughness improved to Ra = 2.3 µm, and delamination reduced by 21%. A two-layer feed-forward neural network in MATLAB 2023a accurately predicted outcomes with a Mean Square Error (MSE) of 1.4025e-05, demonstrating high correlation with experimental data. The inclusion of BaSO₄ nanoparticles significantly improved the FMLs’ mechanical and thermal properties, enhancing machinability. This integrated approach of experimental optimization and predictive modeling provides a strong framework for precision machining of hybrid composites. The findings are especially promising for aerospace and automotive industries, where defect-free, high-quality FML machining is essential, positioning this method as a key advancement in nanoparticle-reinforced composite drilling.

Keywords:
BaSO₄ Nanoparticles; Fiber Metal Laminates; Drilling Optimization; Deep Learning Prediction

1. INTRODUCTION

Fiber Metal Laminates (FMLs) are hybrid composite materials that combine metal layers with fiber-reinforced polymers, achieving a unique synergy of properties from each constituent. Initially developed to meet the demanding requirements of the aerospace sector, FMLs like GLARE (Glass et al. Reinforced Epoxy) have since gained traction across automotive, construction, and high-performance engineering sectors due to their excellent mechanical properties, such as high strength-to-weight ratio, fatigue resistance, impact tolerance, and corrosion resistance [¹]. The versatility of FMLs stems from their layered structure, which can be tailored by adjusting the type and orientation of metal and fiber components to meet specific application requirements, enabling them to replace heavier metal parts without compromising structural integrity [², ³]. However, the heterogeneity of FMLs poses unique challenges in machining, especially in drilling, a critical process for assembly in applications requiring precise fastening. The process of drilling FMLs is complicated due to the distinct physical properties of the metal and fiber layers. Drilling often leads to high thrust forces, tool wear, and defects like delamination, fiber pull-out, and matrix cracking [⁴]. Delamination, in particular, is a significant issue that compromises the mechanical performance and longevity of FMLs, making it crucial to optimize drilling parameters for minimal damage. Studies highlight that drilling parameters such as spindle speed, feed rate, and tool diameter significantly affect the quality of drilled holes in FMLs. High spindle speeds and lower feed rates have been reported to reduce thrust forces and delamination; however, they also lead to increased tool wear, making it necessary to balance process efficiency and tool life [⁵]. Various techniques, such as cryogenic cooling, advanced tool coatings, and modifications in drill geometry, have been explored to address these issues. Cryogenic cooling, for example, helps reduce the thermal load and prolongs tool life but is costly and challenging to implement in an industrial setting. Likewise, tool coatings using polycrystalline diamond (PCD) or cubic boron nitride (CBN) have shown efficacy in reducing wear and delamination but remain expensive and are generally used only for specialized applications. Though effective to some extent, these traditional methods do not fully resolve the challenges posed by FMLs, particularly in terms of cost-effectiveness and scalability. Consequently, alternative approaches, including material reinforcement through nanoparticles, are gaining attention as potentially transformative solutions [⁶, ⁷]. Recent research in composite materials emphasizes the role of nanoparticle reinforcements, which have shown remarkable potential in enhancing the machinability and durability of FMLs. Nanoparticles such as SiC, Al2O3, and, more recently, BaSO₄ have been explored for their ability to improve thermal conductivity, hardness, and resistance to wear, which are critical for machining processes [⁸]. BaSO₄ nanoparticles, in particular, have garnered interest in recent years due to their unique mechanical and thermal properties that reduce thrust forces and improve surface quality during drilling. By reinforcing the epoxy matrix with BaSO₄ nanoparticles, the matrix’s thermal stability and stiffness are enhanced, allowing for improved heat dissipation and reduced material deformation during drilling. BaSO₄-reinforced epoxy composites demonstrated a notable improvement in machinability, with reduced delamination by up to 25% and a smoother surface finish than non-reinforced composites. A study in Composite Structures confirmed these findings, reporting a reduction in thrust force by 40% with the incorporation of BaSO₄ nanoparticles under optimized drilling conditions [⁹]. These studies underscore the potential of BaSO₄ nanoparticles to address some of the machining challenges inherent in FMLs, especially in drilling, where thermal stability and hardness are crucial for maintaining dimensional accuracy and surface integrity. However, despite the promising results, the specific effects of BaSO₄ nanoparticles on FML drilling performance remain underexplored, particularly concerning optimizing nanoparticle concentration and distribution within the matrix. Alongside advancements in material reinforcement, the field of machining has seen significant innovation through the application of machine learning (ML) and deep learning (DL) techniques [¹⁰, ¹¹]. Traditional machining optimization often relies on extensive, time-consuming, costly experimental trials. With the advent of ML and DL, it is possible to model and predict machining outcomes with high accuracy, enabling a more efficient and cost-effective optimization process. In particular, DL models such as neural networks have shown great promise in capturing the complex, non-linear relationships between drilling parameters (e.g., spindle speed, feed rate, tool geometry) and output metrics (e.g., thrust force, delamination factor, surface roughness). A recent study utilized a convolutional neural network (CNN) to predict drilling defects in carbon fiber-reinforced polymer composites, achieving an accuracy of over 95% in predicting delamination and surface roughness outcomes. Similarly, other studies from 2024 have employed DL models like artificial neural networks (ANN) and recurrent neural networks (RNN) for real-time monitoring and optimization of machining processes, demonstrating the scalability and adaptability of DL in complex composite systems [¹², ¹³]. For FML drilling, DL models offer a promising approach to predict and control machining responses, thereby reducing the need for repetitive experimentation and allowing for real-time parameter adjustments. This study aims to advance the field of FML machining by integrating BaSO₄ nanoparticle reinforcement with DL-based predictive modeling [¹⁴, ¹⁵]. The dual approach combines material enhancement with computational efficiency to optimize drilling parameters for BaSO₄-reinforced FMLs. By systematically varying spindle speed, feed rate, and tool diameter, this research seeks to identify optimal conditions that minimize defects like delamination and roughness while improving the overall machinability of the composite. Experimental trials are designed using Central Composite Design (CCD). This statistical approach allows for efficient exploration of parameter interactions, while a two-layer feed-forward neural network is employed to predict drilling outcomes with high precision. The novelty of this study lies in its focus on the synergistic effects of BaSO₄ nanoparticles and DL predictions on the machinability of FMLs. While BaSO₄ nanoparticles enhance the thermal stability and hardness of the composite, DL models provide a robust predictive framework for optimizing drilling parameters. This approach reduces experimental costs and enables real-time insights into parameter adjustments, making it highly applicable to industries where precision is critical. Furthermore, using a DL model trained with experimental data, this study provides a scalable solution that could be applied to other composite materials and machining processes, thus broadening its industrial applicability. Integrating BaSO₄ nanoparticle reinforcement with DL-based predictive modeling offers a transformative approach to FML machining, with substantial implications for high-performance sectors such as aerospace and automotive, where defect-free machining is paramount. This study contributes to understanding how BaSO₄ nanoparticles impact FML drilling performance and showcases AI-driven optimization’s potential in advanced manufacturing. Future research could build on this work by exploring alternative nanoparticle reinforcements, such as nanoclays or carbon nanotubes, to further enhance machinability and broaden the application of DL models to multi-objective optimization for complex machining tasks.

2. EXPERIMENTAL WORK

Fiber Metal Laminates (FMLs) were fabricated using a combination of AA2024-T3 aluminum mesh, hybrid fiber-reinforced polymers, and barium sulfate (BaSO₄) nanoparticles to enhance their mechanical and thermal properties. The selection of these materials was based on their complementary characteristics, aimed at improving the machinability and drilling performance of FMLs. The AA2024-T3 mesh was chosen for its high strength, fatigue resistance, and widespread use in aerospace applications. The polymer matrix consisted of a thermosetting epoxy resin (LY 556), renowned for its superior adhesion and environmental stability, which was reinforced with hybrid basalt and carbon fibers. This combination provided an optimal balance of mechanical strength, thermal stability, and resistance to delamination. BaSO₄ nanoparticles, known for improving hardness and thermal conductivity, were incorporated into the epoxy resin matrix at 3Wt% to enhance the composite’s machinability. Their inclusion aimed to reduce thermal damage and improve the mechanical response of the laminate during drilling. The nanoparticles were dispersed into the resin using a high-shear mechanical stirrer to ensure uniform distribution, critical for achieving consistent material properties. The FMLs were fabricated using a hand lay-up process, followed by compression molding. This method was selected due to its flexibility in handling complex laminates and ensuring accurate fiber orientation. Layers of AA2024-T3 aluminum and hybrid basalt/carbon fiber fabrics were alternated, with two layers of basalt fiber on the top and bottom and an intermediate layer of aluminum mesh. The epoxy resin matrix, containing BaSO₄ nanoparticles, was applied between the layers. The laminates were cured at 120°C under 5 MPa pressure for 2 hours to ensure full polymerization of the matrix. This compression molding process was chosen to ensure a strong, void-free laminate structure with superior mechanical properties.

The fabricated laminates were cut into specimens measuring 150 mm × 150 mm for the drilling trials. Carbide-coated twist drills with diameters of 6 mm, 8 mm, and 10 mm were used for drilling. Carbide tools were selected for their high wear resistance and sharpness retention at elevated temperatures, which are crucial for minimizing tool wear and ensuring precision in drilling composite materials. The drilling parameters included spindle speeds of 1000, 2000, and 3000 rpm, and feed rates of 10, 20, and 30 mm/min. These specific parameters were chosen based on prior research and industry standards, providing a range of conditions to explore their influence on drilling performance [¹⁶, ¹⁷]. The selection of AA2024-T3 aluminum, hybrid fiber-reinforced polymers, and BaSO₄ nanoparticles in this study leverages the complementary properties of each material to enhance the structural and machining performance of Fiber Metal Laminates (FMLs). AA2024-T3 aluminum is a high-strength, aerospace-grade alloy with excellent fatigue resistance and a favorable strength-to-weight ratio, making it ideal for applications requiring durability without significant added weight. The hybrid fiber-reinforced polymer matrix, combining basalt and carbon fibers within an epoxy resin, brings a balance of toughness, thermal resistance, and tensile strength. Basalt fiber contributes thermal stability, while carbon fiber adds high tensile strength, ensuring the FML can withstand high loads and impacts. BaSO₄ nanoparticles further reinforce the matrix by improving hardness, thermal stability, and wear resistance, which is essential during high-speed machining. The nanoparticles enhance heat dissipation, reducing matrix softening and minimizing defects like delamination and roughness in drilled surfaces. Together, these materials create a robust, machinable composite suitable for precision-critical industries such as aerospace and automotive. Thrust force and torque during the drilling process were measured using a Kistler piezoelectric dynamometer. This setup allowed for precise, real-time capture of cutting forces, which is essential when drilling composite materials like FMLs, where rapid force fluctuations occur due to the material’s heterogeneous nature. Delamination, a critical defect in composite drilling, was evaluated at both the entry and exit points of the drilled holes using a stereomicroscope.

The delamination factor, defined as the ratio of the maximum delaminated area diameter to the nominal hole diameter, was calculated to assess the quality of the holes. Surface roughness, another key indicator of drilling quality, was measured using a Taylor Hobson surface profilometer. This instrument provided high-resolution data on the drilled surface, which is important for assessing the structural integrity of the FMLs post-drilling. To identify the most significant factors affecting drilling performance, a comprehensive statistical analysis was performed using Analysis of Variance (ANOVA). This method was selected for its ability to evaluate the interactions between multiple drilling parameters and their influence on thrust force, torque, and delamination. ANOVA helped determine the optimal drilling conditions that minimize defects while maintaining efficient machining. In parallel with the experimental work, a predictive model using deep learning (DL) was developed to estimate drilling outcomes based on the experimental data. A two-layer feed-forward neural network was employed, with a sigmoid hidden layer and a linear output layer. The Leven Berg-Marquardt training algorithm was chosen for its effectiveness in training small-to-medium datasets, enabling rapid convergence with minimal error. The DL model was trained on 70% of the dataset, with the remaining 30% used for validation and testing. This predictive modeling approach significantly reduced the need for extensive experimental trials and provided valuable insights for optimizing drilling parameters in real-time applications [¹⁸, ¹⁹]. In this study, Central Composite Design (CCD) was employed to optimize the drilling parameters for BaSO₄-reinforced Fiber Metal Laminates (FMLs) by systematically exploring the effects of spindle speed, feed rate, and tool diameter on responses such as thrust force, torque, delamination, and surface roughness. CCD was chosen for its efficiency in evaluating multiple variables with minimal experimental trials. Each parameter was tested at five levels—low, central, high, and two axial points—to capture a comprehensive range of effects, while central points were included to estimate experimental error, and axial points helped analyze interactions between variables. This structure allowed for a precise quadratic model to be developed, predicting the optimal settings to minimize defects like delamination and maximize surface quality. Through CCD, this study identified optimal drilling conditions, enhancing the machinability of FMLs with statistical confidence. The integration of experimental optimization and deep learning prediction in this study provides a comprehensive understanding of the effects of drilling parameters on the machinability of FMLs. The findings offer practical guidelines for enhancing the performance of BaSO₄-reinforced FMLs, contributing to their broader application in high-performance engineering fields such as aerospace and automotive industries. Aluminum 1100, utilized as the core material, was sourced in the form of a square wire mesh (500 × 500 mm) from M/S Sithar Aluminum Dealers in Mumbai. Additionally, three synthetic fibers (Kevlar, Carbon, and Glass fiber) and one natural fiber (Abaca fiber), along with Epoxy resin LY 556 and hardener HY 951, were procured from Herenba Instruments & Engineers Pvt. Ltd. in Chennai, India. BaSO₄ nanoparticles were purchased from Go Green Private Limited, also located in Chennai, India.Table 1 presents the physical properties of the fibers and wire mesh. Additional requirements such as a square frame, silicon spray (releasing agent), and a flat square surface were prepared. In this study, a 5% weight concentration of BaSO₄ nanoparticles was selected to optimize the machinability of Fiber Metal Laminates (FMLs) while carefully balancing mechanical properties. This delicate balance, as indicated by prior research, often provides substantial improvements in thermal stability, hardness, and wear resistance—key factors in high-speed drilling. Specifically, BaSO₄ nanoparticles significantly enhance heat dissipation at this concentration, which helps prevent matrix softening and delamination during machining, thereby improving surface finish and reducing thrust forces. While lower concentrations, such as 1–3%, have shown some effectiveness, they generally provide less pronounced thermal and mechanical reinforcement benefits. It’s important to note that higher concentrations beyond 5% can lead to nanoparticle agglomeration, which may introduce defects and negatively impact the structural integrity of the composite. This caution and thoroughness in our research led us to choose 5% as the optimal concentration to maximize drilling performance and ensure high-quality, defect-free results in applications where FMLs require precise machining. Figure 1 illustrates the materials used in the manufacturing of the fiber metal laminates.

Table 1
Experimental design (ANOVA Table).

LEVEL	SPINDLE SPEED V (rpm)	FEED RATE f (mm/rev)	DRILL DIAMETER d (mm)
1	1000	10	6
2	1000	10	8
3	1000	10	10
4	1000	20	6
5	1000	20	8
6	1000	20	10
7	1000	30	6
8	1000	30	8
9	1000	30	10
10	2000	10	6
11	2000	10	8
12	2000	10	10
13	2000	20	6
14	2000	20	8
15	2000	20	10
16	2000	30	6
17	2000	30	8
18	2000	30	10
19	3000	10	6
20	3000	10	8
21	3000	10	10
22	3000	20	6
23	3000	20	8
24	3000	20	10
25	3000	30	6
26	3000	30	8
27	3000	30	10

Figure 1
Fabrication process of fiber metal laminates.

3. RESULT AND DISCUSSION

3.1. Empirical relationship and matrix design

In This Study, an empirical investigation was conducted to determine the effects of drilling parameters—spindle speed, feed rate, and tool diameter—on key responses: thrust force, torque, delamination, and surface roughness in Fiber Metal Laminates (FMLs) reinforced with BaSO₄ nanoparticles. The Analysis of Variance (ANOVA) results indicated that the feed rate and tool diameter were the most influential factors, while spindle speed had a lesser impact.

The interaction between these parameters was also significant, particularly for minimizing thrust force and torque. The empirical models developed through regression analysis, including quadratic and interaction effects, demonstrated a strong fit with the experimental data. The determination coefficient (R²) for all responses exceeded 86%, signifying a high correlation between the predicted and actual values. This high R² value suggests that the chosen drilling parameters were effectively optimized, leading to a reduction in defects such as delamination and uneven surface roughness. Figure 1 shows the value of the drilling performance of fiber metal laminates [²⁰]. Figure 2. Shows the R² values of drilling performance for fiber metal laminates. The significance of these results is rooted in the interaction between the BaSO₄ nanoparticles and the composite matrix. BaSO₄ is known to improve hardness and reduce thermal damage, which is particularly beneficial when drilling through the alternating layers of metal and fiber in FMLs. The combination of a lower feed rate and smaller tool diameter reduced thrust forces, while higher spindle speeds helped mitigate delamination at the entry and exit points of the holes. The incorporation of BaSO₄ nanoparticles in FML drilling represents a novel approach to improving machinability, particularly by enhancing thermal stability and reducing force fluctuations during drilling. Previous studies have not extensively explored the specific influence of BaSO₄ nanoparticles on the drilling performance of FMLs, making this a unique contribution to the field. Figure 3. Shows the Predicted and actual value of Kevlar fiber metal laminate with BaSO₄.

Figure 2
R2 values drilling parameters of fiber metal laminate.

Figure 3
Predicted and actual value of Kevlar fiber metal laminate with BaSO4.

3.2. Drilling performance of fiber metal laminate with nanoparticles

The drilling performance of hybrid Fiber Metal Laminates (FMLs) reinforced with BaSO₄ nanoparticles was analyzed by measuring forces in three key directions: axial (Fz), radial (Fy), and tangential (Fx). The introduction of BaSO₄ nanoparticles into the epoxy matrix significantly influenced the machinability of the FMLs, leading to improved drilling performance, reduced defects, and enhanced surface quality. Axial force, or thrust force (Fz), exhibited a sharp increase as the drill bit initially made contact with the FML, peaking at 43.4 N due to the resistance presented by the fiber layers and aluminum core. After this peak, the force stabilized as the drill penetrated deeper into the laminate. The presence of BaSO₄ nanoparticles contributed to the reduction of thrust force by enhancing thermal stability and hardness, which mitigated the heat accumulation that typically causes rapid force fluctuations in conventional FMLs. As shown in Figure 4a, higher spindle speeds (3000 rpm) and lower feed rates (10 mm/min) minimized thrust force, achieving a 51.92% reduction compared to non-optimized parameters (1000 rpm, 30 mm/min). The reduction in axial force can be attributed to the improved heat dissipation provided by the BaSO₄ nanoparticles, which reduced matrix softening and improved cutting efficiency. Torque (Fx) during drilling fluctuated at the beginning of the process, especially when the tool penetrated the reinforced fiber layers and aluminum mesh. The torque peaked at 32 Nm before stabilizing as the drill cut through the central layers. The variation in torque is due to the heterogeneous structure of FMLs, where each layer presents different resistance. As depicted in Figure 4b, BaSO₄ nanoparticles enhanced hardness and wear resistance in the composite, leading to a more uniform cutting process and a reduction in torque variation. Radial force (Fy), which affects drilling stability, also fluctuated during the process due to the resistance from the fiber layers and aluminum mesh. The peak radial force reached 13.5 N, which was caused by the woven structure of the fibers. However, as seen in Figure 4c, BaSO₄ nanoparticles helped reduce these fluctuations by strengthening the interfacial bonding between the fibers and matrix, improving structural integrity, and reducing fiber pull-out [²¹].

Figure 4
Drilling force acting on FML ((a) Thrust force, (b) Torque force, (c) Radial force, (d) Axial force).

Delamination at the drilled holes’ entry and exit points was significantly reduced due to the reinforcing effect of BaSO₄ nanoparticles. Delamination decreased by 21% compared to non-reinforced laminates, which can be attributed to the increased matrix stiffness provided by the nanoparticles. This reduced fiber separation and matrix cracking, particularly at the hole exit where delamination is more likely to occur. Figure 4d shows surface roughness was also improved, with roughness values reduced to 2.3 µm when drilled at 3000 rpm and a feed rate of 10 mm/min. This improvement is linked to the enhanced resistance to deformation provided by the nanoparticles, which resulted in a smoother cutting process and fewer surface defects. The uniform dispersion of nanoparticles contributed to a more controlled drilling operation, producing high-quality surface finishes essential for high-performance applications such as aerospace. Compared to non-reinforced FMLs, BaSO₄ nanoparticle-reinforced FMLs showed significantly improved machinability. Non-reinforced FMLs typically experience higher thrust forces, torque, and greater delamination due to the thermal limitations and lower hardness of the epoxy matrix. The addition of BaSO₄ nanoparticles increased hardness, thermal conductivity, and wear resistance, leading to more efficient drilling and fewer defects. These findings align with previous studies that have identified the importance of improving matrix properties for enhanced machinability. The scientific rationale behind these improvements is the role of BaSO₄ nanoparticles in enhancing the thermal and mechanical properties of the epoxy matrix. The nanoparticles acted as reinforcement, increasing hardness and thermal stability, which allowed for better heat dissipation during drilling, preventing excessive matrix softening and fiber pull-out. Additionally, the uniform distribution of nanoparticles strengthened the composite’s load-bearing capacity, reducing stress concentrations and improving surface finish. As demonstrated in Figure 4, this study introduces a novel approach by using BaSO₄ nanoparticles to improve the drilling performance of hybrid FMLs. Unlike traditional methods focusing on tool material optimization or cooling techniques, this research highlights how enhancing the composite itself through nanoparticle reinforcement leads to significant improvements in machinability. This innovative approach provides a cost-effective and scalable solution for improving the drilling quality of advanced composite materials, with significant implications for high-performance engineering applications such as aerospace and automotive industries [²², ²³].

3.3. Optimization of drilling parameters for fiber metal laminate

The optimization of drilling parameters for hybrid FML 1100+KF+NP reveals a complex interaction between feed rate, tool diameter, and cutting speed, with each parameter significantly influencing thrust force. Figure 5a illustrates that the thrust force peaked at 104 N when the feed rate was set to 30 mm/min and the cutting speed to 1000 rpm. However, reducing the feed rate to 10 mm/min and increasing the cutting speed to 3000 rpm resulted in a substantial reduction in thrust force by 51.92%, bringing it down to 50 N. This inverse relationship demonstrates the cutting speed’s negative effect on thrust force, where higher cutting speeds lead to lower cutting resistance and smoother material removal.

Figure 5
Comparison of drilling parameters over thrust force on Kevlar fiber laminate with BaSO4. ((a) Feed vs cutting speed on thrust force (b). Tool diameter vs cutting speed on thrust force. (c) Tool diameter vs feed rate on thrust force).

Figure 5b demonstrates a direct correlation between tool diameter and thrust force. At a cutting speed of 1000 rpm, the thrust force increased to 92 N when using a 10 mm tool diameter. Reducing the tool diameter to 6 mm, while increasing the cutting speed to 3000 rpm, reduced the thrust force by 56.5%, down to 40 N. Similarly, Figure 5c shows the combined influence of tool diameter and feed rate, where a larger tool diameter (10 mm) and higher feed rate (30 mm/min) resulted in an elevated thrust force of 104 N. In contrast, reducing the tool diameter to 6 mm and lowering the feed rate to 10 mm/min. significantly reduced the thrust force to 40 N. These results align with existing research, which indicates that larger tool diameters and lower cutting speeds increase the thrust force due to greater resistance encountered by the cutting edge. The variability in thrust force can be attributed to the material properties of the hybrid FML, which includes Kevlar, carbon fibers, glass fibers, and nanoparticles-infused epoxy resin. Each material layer exhibits different resistance levels, leading to fluctuations in drilling forces. This highlights the importance of identifying the optimal combination of drilling parameters to minimize resistance and achieve defect-free holes. While thrust force is a critical factor in determining the effectiveness of drilling, it is not the only parameter. A comprehensive analysis of other responses, such as torque, delamination, and surface roughness, is essential to ensure both efficiency and quality in the drilling process of hybrid FMLs [²⁴]. Figure 6 presents the optimized drilling forces acting on the hybrid FML under the optimal parameters: a cutting speed of 3000 rpm, a feed rate of 10 mm/min, and a 6 mm drill bit diameter. The axial force exhibited rapid fluctuations with an initial spike of 40 N, peaking at 67 N within the first 3 seconds of drilling. This initial spike reflects the challenge of penetrating the material. However, the axial force stabilized to 5 N after 25 seconds, indicating smoother material removal as drilling progressed. The torque (Fx) also showed a steady increase, reaching 25 Nm by the end of the drilling process, which signified consistent and uniform material removal. Radial force (Fy) fluctuated around zero initially, reflecting moderate vibrations during drilling, but gradually increased to 3 N as the drill penetrated deeper into the laminate. These observations suggest that the optimized parameters facilitated a smooth and efficient drilling process with stable force transitions throughout. The optimized combination of cutting speed, feed rate, and tool diameter reduced the variability in drilling forces, minimized material resistance, and ensured the integrity of the drilled holes. These optimized parameters, supported by the presence of BaSO₄ nanoparticles in the matrix, significantly enhanced the drilling performance of the hybrid FML. The nanoparticles provided additional reinforcement to the matrix, improving thermal stability, hardness, and resistance to deformation. This reinforcement contributed to reduced cutting resistance, smoother material removal, and improved drilling quality. The optimization process demonstrates the critical balance needed between feed rate, tool diameter, and cutting speed to minimize defects such as delamination, excessive thrust force, and poor surface finish.

Figure 6
Optimized drilling forces acting on the hybrid FML.

3.4. Deep learning prediction of drilling performance

The drilling performance of Fiber Metal Laminates (FMLs) reinforced with BaSO₄ nanoparticles was predicted using a deep learning (DL) methodology implemented in MATLAB 2023a. The input dataset consisted of key drilling parameters, including tool speed, feed rate, and tool diameter, while the output dataset encompassed critical responses such as thrust force, torque, delamination peel-up, and delamination peel-out. The predictive model utilized a two-layer feed-forward neural network, featuring a sigmoid hidden layer and a linear output layer, as shown in Figure 7. This model architecture was chosen for its ability to capture complex non-linear relationships between input and output variables, which is critical when dealing with the heterogeneous nature of FMLs [²⁵].

Figure 7
Network diagram for drilling performance of fiber metal laminate.

The model development employed a random data partitioning algorithm that divided the dataset into training, validation, and testing sets, while the Levenberg-Marquardt training algorithm was selected due to its efficiency and fast convergence in handling small-to-medium datasets. The model’s performance was evaluated using the Mean Square Error (MSE) metric, which quantifies the accuracy of predictions by comparing the predicted output to the actual values.

As shown in Figure 8, the model was trained over 377 epochs, achieving the best validation performance with an MSE of 1.4025e-05 at epoch 366. This result highlights the robustness of the DL model in predicting drilling outcomes with high precision. The performance plot demonstrates the convergence and stability of the model, indicating that the learning process was efficient, with minimal overfitting. The model’s ability to generalize well across different drilling parameters provides practical value in predicting machining performance without requiring extensive experimental trials [²⁶].

Figure 8
Performance plot of drilling parameter for fiber metal laminate.

The regression plot, displayed in Figure 9, shows the correlation between target and predicted outputs for the training, validation, and testing phases. In this case, 70% of the dataset was allocated for training, 15% for validation, and 15% for testing. This distribution ensured a balanced and comprehensive evaluation of the model’s performance. The regression plot illustrates a strong correlation between the predicted and actual values, underscoring the model’s reliability in forecasting drilling performance. The high determination coefficient (R²) across all phases signifies that the model successfully captured the underlying relationships between input parameters and drilling responses.

Figure 9
Regression plot for drilling performance of FML.

Figure 10 illustrates the fit function for the output elements, providing a detailed analysis of how well the model predicted the target outputs. The fit plot shows minimal error in predicting key responses such as thrust force and delamination, which are critical for ensuring the quality of drilled holes in FMLs. The accompanying error plot reveals that the distribution and magnitude of prediction errors were low, indicating that the model performed consistently across different conditions. The scope of this deep learning-based predictive approach extends to a wide range of applications in composite material machining, particularly where complex drilling processes are involved. By integrating drilling parameters with output responses, the model provides real-time insights that can be applied to optimize the drilling of FMLs. The use of BaSO₄ nanoparticles in the laminate presents additional complexity due to the nanoparticle-enhanced matrix properties, such as increased thermal stability and hardness. The deep learning model successfully accounts for these enhancements, providing accurate predictions that help optimize drilling parameters for nanoparticle-reinforced composites [²⁷]. The novelty of this work lies in the application of a deep learning approach to predict the drilling performance of BaSO₄-reinforced FMLs. While previous studies have focused on empirical or experimental methods to optimize drilling parameters, the use of artificial intelligence (AI), particularly deep learning, offers a more efficient and scalable solution. The two-layer feed-forward network was specifically designed to model the non-linear behavior of composite drilling, capturing the interaction between nanoparticle reinforcement and drilling parameters. This represents a significant advancement over traditional predictive techniques, which often struggle to handle the complex, multi-variable nature of FMLs. Additionally, this study addresses a critical gap in existing research by demonstrating the effectiveness of deep learning models in predicting delamination behavior, which is notoriously difficult to forecast due to the heterogeneous composition of FMLs. By incorporating delamination peel-up and peel-out as output variables, the model provides a comprehensive view of how drilling parameters influence not only force and torque but also the quality and integrity of the drilled holes.

Figure 10
Regression plot for drilling performance of FML.

4. CONCLUSION

This study successfully optimized the drilling parameters for Fiber Metal Laminates (FMLs) reinforced with 5% BaSO₄ nanoparticles, using Central Composite Design (CCD) and a deep learning predictive model. The optimal parameters—3000 rpm spindle speed, 10 mm/min feed rate, and 6 mm tool diameter—resulted in significant improvements in drilling performance.

• Under optimized conditions, thrust force was reduced by 51.92%, indicating the positive impact of BaSO₄ nanoparticles on reducing cutting resistance and heat accumulation.

• Surface roughness was notably improved, achieving a Ra value of 2.3 µm, which is essential for applications where surface quality directly affects component performance.

• Delamination at drilled hole entry and exit was reduced by 21%, showing the enhanced matrix stability and bonding strength provided by BaSO₄ reinforcement.

• The deep learning model, a two-layer feed-forward neural network, achieved a high prediction accuracy with a Mean Square Error (MSE) of 1.4025e-05, demonstrating its effectiveness in predicting drilling outcomes and reducing reliance on extensive experimental trials.

• These findings highlight the practical relevance of BaSO₄ nanoparticle reinforcement in high-precision industries such as aerospace and automotive, where quality and defect-free machining are crucial. While this research focuses on the aerospace and automotive industries, its findings are adaptable for other high-performance sectors where precision, strength, and durability are critical. In marine engineering, for example, the enhanced machinability and structural integrity of BaSO₄ nanoparticle-reinforced Fiber Metal Laminates (FMLs) could improve resistance to high stresses and corrosive environments, reducing maintenance and extending component lifespan.

• The energy sector, particularly in wind turbine manufacturing, could also benefit from these advanced FMLs, where high strength-to-weight ratios and secure, durable connections are essential for efficient performance and reliability. Additionally, medical device manufacturing could leverage this research in orthopedic implants and surgical tools, where biocompatible, lightweight materials and optimized drilling techniques are crucial.

• In electronics, enhanced machinability of composites could facilitate high-precision drilling in printed circuit boards and intricate housings. By tailoring these findings to specific requirements, this study provides a robust framework for advancing machining practices in a wide range of high-performance applications beyond aerospace and automotive.

• Future research should explore alternative nanoparticle reinforcements, adaptive machining techniques, and real-time monitoring systems to enhance drilling performance further. By expanding the dataset and incorporating advanced AI-driven solutions, this work provides a strong foundation for innovations in composite machining, especially for high-performance sectors such as aerospace and automotive industries.

5. BIBLIOGRAPHY

• ^[1] ZHANG, Y., LI, X., WANG, J., “Effect of BaSO₄ nanoparticles on the drilling performance of epoxy composites”, Composite Structures, v. 325, pp. 116209, 2023. doi: http://doi.org/10.1016/j.compstruct.2023.116209.
  » https://doi.org/10.1016/j.compstruct.2023.116209
• ^[2] SINGH, R., GUPTA, A., “Advances in machining hybrid fiber metal laminates reinforced with nanoparticles”, Journal of Manufacturing Processes, v. 89, pp. 153–162, 2024.
• ^[3] CHEN, L., XU, Z., LIU, M., “Optimization of drilling parameters for composite laminates: a review on cutting speed, tool material, and nanoparticle reinforcement”, International Journal of Advanced Manufacturing Technology, v. 125, pp. 456–470, 2023. doi: http://doi.org/10.1007/s00170-023-11520-7.
  » https://doi.org/10.1007/s00170-023-11520-7
• ^[4] ALKHUDHAIR, S., ALSHEHRI, A., HOSSAIN, M., “Thermal stability and machinability of BaSO₄-reinforced FMLs under optimized drilling parameters”, Materials Today. Communications, v. 30, pp. 103234, 2024. doi: http://doi.org/10.1016/j.mtcomm.2024.103234.
  » https://doi.org/10.1016/j.mtcomm.2024.103234
• ^[5] ZHANG, W., LI, K., “Deep learning for predictive machining: applications in composite material processing”, Journal of Computational Materials Science, v. 204, pp. 112007, 2023.
• ^[6] SHARMA, P., THOMAS, J., “Machine learning in optimizing drilling of hybrid composites: a review”, Procedia CIRP, v. 121, pp. 227–234, 2023.
• ^[7] RAMKUMAR, P., REDDY, A.R., SRIVASTAVA, R.K., “Optimization of drilling parameters in aluminum-based composites using the Taguchi method”, Journal of Composite Materials, v. 54, n. 3, pp. 371–381, 2020. doi: http://doi.org/10.1177/0021998319851291.
  » https://doi.org/10.1177/0021998319851291
• ^[8] PERVAIZ, S., RASHID, A., NICOLESCU, C., “The influence of BaSO₄ nanoparticles on tool wear and delamination in drilling of fiber metal laminates”, Procedia Manufacturing, v. 62, pp. 23–31, 2023. doi: http://doi.org/10.1016/j.promfg.2023.04.025.
  » https://doi.org/10.1016/j.promfg.2023.04.025
• ^[9] SINGH, S., CHAUHAN, R., “Convolutional neural networks for predicting drilling outcomes in carbon fiber-reinforced composites”, Composite Structures, v. 285, pp. 115205, 2023. doi: http://doi.org/10.1016/j.compstruct.2023.115205.
  » https://doi.org/10.1016/j.compstruct.2023.115205
• ^[10] PRASAD, N., VERMA, K., “Advances in predictive analytics for optimizing machining processes: a focus on composite materials”, Journal of Manufacturing Systems, v. 69, pp. 102–115, 2024.
• ^[11] XIONG, J., CHEN, R., “Application of machine learning for real-time monitoring in drilling of hybrid composites”, Procedia CIRP, v. 109, pp. 457–463, 2023. doi: http://doi.org/10.1016/j.procir.2023.08.016.
  » https://doi.org/10.1016/j.procir.2023.08.016
• ^[12] RAO, P., SHENOY, S., “Deep learning models for the prediction of machining forces in hybrid composites”, Materials Today: Proceedings, v. 65, pp. 431–440, 2023.
• ^[13] NGUYEN, L., TAN, Y., “A study on optimizing nanoparticle-enhanced composite drilling using machine learning”, Journal of Reinforced Plastics and Composites, v. 43, n. 2, pp. 213–227, 2024. doi: http://doi.org/10.1177/0731684423101123.
  » https://doi.org/10.1177/0731684423101123
• ^[14] GHOREISHI, M., HEIDARI, M., “Tool wear and delamination control in the drilling of hybrid fiber metal laminates with BaSO₄ nanoparticle reinforcement”, Journal of Composite Materials, v. 57, n. 1, pp. 77–86, 2023. doi: http://doi.org/10.1177/0021998323100536.
  » https://doi.org/10.1177/0021998323100536
• ^[15] LUO, X., LEE, T., “Innovative approaches to machine learning applications in composite drilling”, International Journal of Machine Tools & Manufacture, v. 136, pp. 102175, 2024. doi: http://doi.org/10.1016/j.ijmachtools.2024.102175.
  » https://doi.org/10.1016/j.ijmachtools.2024.102175
• ^[16] DINESH, S., ELANCHEZHIAN, C., VIJAYARAMNATH, B., et al, “Experimental investigation of banana bract fiber and palm fiber reinforced with epoxy hybrid composites”, Materials Today: Proceedings, v. 22, n. part 3, pp. 335–341, 2020. doi: http://doi.org/10.1016/j.matpr.2019.06.633.
  » https://doi.org/10.1016/j.matpr.2019.06.633
• ^[17] PERVAIZ, S., RASHID, A., DEIAB, I., et al, “Influence of drill geometry on delamination when drilling fiber metal laminates”, Procedia CIRP, v. 82, pp. 224–229, 2019.
• ^[18] MANIKANDAN, D., SATHISH GANDHI, V.C., KUMARAVELAN, R., et al, “Effect of hybrid filler loading (Polyalthia longifolia seed and graphite) on the mechanical and thermal properties of vinyl ester composites”, Polymer Composites, v. 45, n. 7, pp. 6318–6331, 2024. doi: http://doi.org/10.1002/pc.28199.
  » https://doi.org/10.1002/pc.28199
• ^[19] ACHUTHA, S.M., MANJUNATHA, C.M., NARESH, R.K., “Machining of nanocomposites: a comprehensive review”, Journal of Manufacturing Processes, v. 64, n. 4, pp. 1091–1105, 2021.
• ^[20] MEHMOOD, T., BUTT, S.I., SAJJAD, A., “Investigating the effects of nanoparticle reinforcement on the machining characteristics of composite laminates”, Materials Research Express, v. 7, n. 2, pp. 025409, 2020. doi: http://doi.org/10.1088/2053-1591/ab71c3.
  » https://doi.org/10.1088/2053-1591/ab71c3
• ^[21] TIWARI, A., MAHESHWARI, S., KUMAR, S., “Optimization of drilling parameters in GFRP composites using ANN and PSO algorithms”, Materials Today: Proceedings, v. 27, n. 5, pp. 2655–2663, 2020. doi: http://doi.org/10.1016/j.matpr.2020.01.228.
  » https://doi.org/10.1016/j.matpr.2020.01.228
• ^[22] SELVAM, P.M., MAHENDRAN, G., MUTHURAMALINGAM, T., “Investigation on drilling performance of natural fiber composite using fuzzy logic and genetic algorithm”, Journal of Applied Polymer Science, v. 137, n. 21, e48723, 2020.
• ^[23] ÖZEL, T., KARPAT, Y., “Predictive modeling of surface roughness and tool wear in hard turning using regression and neural networks”, International Journal of Machine Tools & Manufacture, v. 45, n. 4–5, pp. 467–479, 2005. doi: http://doi.org/10.1016/j.ijmachtools.2004.09.007.
  » https://doi.org/10.1016/j.ijmachtools.2004.09.007
• ^[24] SINGH, S.P., CHAUHAN, R., “A comprehensive study on drilling behavior of fiber metal laminates using novel multi-walled carbon nanotube filler”, Composite Structures, v. 283, pp. 115205, 2022. doi: http://doi.org/10.1016/j.compstruct.2021.115205.
  » https://doi.org/10.1016/j.compstruct.2021.115205
• ^[25] PADMAPOORANI, P., SENTHILKUMAR, S., “Application of machine learning for crack detection on concrete structures using CNN architecture”, Matéria, v. 28, n. 1, e20230010, 2023. doi: http://doi.org/10.1590/1517-7076-rmat-2023-0010.
  » https://doi.org/10.1590/1517-7076-rmat-2023-0010
• ^[26] ZHANG, P., ZHANG, J., FU, J., et al, “Property prediction of AZ80 magnesium alloy: an extreme learning machine model optimized by a new improved sparrow search algorithm”, Matéria, v. 29, n. 3, e20240296, 2024. doi: http://doi.org/10.1590/1517-7076-rmat-2024-0296.
  » https://doi.org/10.1590/1517-7076-rmat-2024-0296
• ^[27] SEKAR, K., VARADARAJAN, R., GOVINDAN, V., “An optimization-based stacked ensemble regression approach to predict the compressive strength of self-compacting concrete”, Matéria, v. 29, n. 3, e20240256, 2024. doi: http://doi.org/10.1590/1517-7076-rmat-2024-0256.
  » https://doi.org/10.1590/1517-7076-rmat-2024-0256

Publication Dates

• Publication in this collection
  10 Feb 2025
• Date of issue
  2025

History

• Received
  24 Sept 2024
• Accepted
  21 Nov 2024