Wireless pressure monitoring of composite pressure vessels based on ultra-high frequency RFID

Abstract

To address the challenges of high cost, complex wiring, and electromagnetic interference in the online monitoring of composite pressure vessels, this paper proposes a wireless pressure monitoring method based on passive ultra-high frequency (UHF) RFID technology. A 2D axisymmetric finite element model of an aluminum-lined, glass-fiber-wrapped dual-layer cylinder was established using COMSOL Multiphysics to analyze the mechanical response under internal pressures ranging from 0 to 20 MPa. By constructing a force-electrical coupling model, the modulation mechanism was revealed, showing a “redshift” in the antenna’s resonant frequency and a linear attenuation of the Received Signal Strength Indicator (RSSI) induced by strain. Considering the curved features of the vessel, HFSS simulations were utilized to analyze the effects of different curvatures on tag performance, confirming a systematic evolution law where tag gain and read range decrease as the cylinder radius increases. Graded tensile experiments verified that the measured signal evolution trends are highly consistent with simulation predictions, with RSSI and surface strain exhibiting a significant linear negative correlation across the full range. This study successfully validates the “Pressure-Strain-Signal Strength” sensing logic, providing a wireless, passive, and low-cost solution for the full lifecycle health management of composite pressure vessels.

Keywords

composite pressure vessel pressure monitoring UHF RFID force-electrical coupling received signal strength indicator (RSSI)

Introduction

Composite pressure vessels, particularly Type IV and Type V all-composite containers, have been widely applied in fields such as hydrogen storage and transportation, aerospace, and natural gas vehicles due to their high specific strength, lightweight properties, and excellent corrosion resistance.¹ However, these vessels are subjected to cyclic internal pressure loads during long-term service, and their failure modes are complex, typically originating from the accumulation of micro-damage at the fiber-matrix interface and potentially leading to catastrophic burst failure.² Finite element analysis and experimental studies indicate that failure generally initiates in the “knuckle region” where the cylinder transitions to the dome, with fiber breakage occurring when the internal pressure reaches a critical value.³ Consequently, real-time, online monitoring of internal pressure and the resulting surface strain is critical for assessing structural integrity, predicting remaining life, and achieving predictive maintenance.⁴

To meet the growing demands for monitoring, academia has explored and applied various structural health monitoring (SHM) techniques based on physical properties. Traditional SHM techniques, such as acoustic emission (AE) and fiber Bragg grating (FBG) sensors, have been applied to damage detection and strain measurement in composite pressure vessels. As a passive monitoring method, acoustic emission technology can capture transient elastic waves generated by internal material damage and is frequently used to determine the first-ply failure pressure. By detecting the transient elastic wave energy released due to damage evolution during the loading process, real-time detection and localization of damage are achieved.⁵ In the field of optical fiber sensing, FBG technology has garnered significant attention due to its immunity to electromagnetic interference and quasi-distributed measurement capabilities. Researchers have utilized FBG sensing data to achieve real-time characterization of the vacuum level in aerospace vacuum pressure vessels,⁶ and have successfully implemented effective monitoring of vacuum leakage states in pressure vessels through quasi-distributed FBG sensing techniques.⁷ Other studies have achieved continuous structural health monitoring throughout the entire service life by winding sensing fibers spirally and circumferentially around polymer liners during the CFRP manufacturing process.⁸ Research by Alblalaihid et al. developed a linear capacitance-based sensor embedded within glass fiber and Kevlar fiber reinforced polymer composites for strain monitoring and early damage detection.⁹ Lu et al. developed a high-performance strain sensor based on laser-induced graphene/polydimethylsiloxane (A-LIG/PDMS) for monitoring and locating impact damage in composite structures.¹⁰ Barragán et al. reported on a new generation of embedded printed sensors for near-field monitoring of high-performance composites.¹¹ Bertram et al. proposed a “tag-like” sensor concept that integrates temperature-compensated full-bridge strain gauges, digital temperature sensors, and RFID transponders to wirelessly monitor resin infusion, the curing process, and mechanical loads during service.¹² Garcia-Etxabe et al. investigated the potential of embedding amorphous ferromagnetic microwires into composite materials to serve as wireless stress sensors.¹³ Since dimensional changes in a rectangular patch antenna in both horizontal and vertical directions lead to linear shifts in its resonant frequency, it serves as an excellent strain sensor.¹⁴ This frequency shift phenomenon has been widely applied in the monitoring of internal damage,¹⁵ surface cracks,¹⁶ and strain¹⁷ in composite materials.

However, despite certain progress made with existing technologies, traditional wired sensor systems still face significant bottlenecks in large-scale practical engineering applications. First, their deployment typically requires complex wiring, which not only increases installation costs and complexity but also makes large-scale, distributed integration difficult to achieve on filament-wound pressure vessels with rotationally symmetric structures.¹⁸ Second, the physical connection points between sensors and readout equipment often become weak links in reliability, being susceptible to damage in harsh industrial environments or under dynamic loads. Furthermore, many wired sensing solutions are prohibitively expensive and may negatively impact the mechanical properties of the composite structure itself, failing to meet the economic requirements for large-scale commercial applications.¹⁹ In addition, in specific application scenarios such as hydrogen energy, pressure vessels may face environments with strong electromagnetic interference, which imposes higher demands on the stability and anti-interference capabilities of traditional wired sensors that rely on electrical signal transmission. Consequently, there is an urgent demand in both industry and academia for an online pressure monitoring technology that is wireless, passive, low-cost, and easy to integrate. Such a technology must be seamlessly integrated onto the surface or within the interior of composite vessels to provide long-term, reliable monitoring of mechanical strain caused by internal pressure loads—without compromising structural performance—thereby providing critical data input for structural health management and safety assessment.²⁰

In the pursuit of more efficient monitoring solutions, passive Ultra-High Frequency (UHF) Radio Frequency Identification (RFID) technology stands out due to its unique energy transmission and communication mechanisms. This technology is an automatic identification method that achieves non-contact, bidirectional data communication and energy transfer via radio frequency signals. The basic system consists of a reader and a tag attached to the object being identified. The tag does not require a built-in power supply; its operating energy is derived entirely from the interrogation radio waves emitted by the reader.²¹ This zero-power consumption characteristic is one of the core features of passive RFID technology, allowing it to demonstrate immense potential in scenarios requiring long-term, maintenance-free monitoring.

UHF-band (typically 860–960 MHz) RFID technology offers faster transmission rates and smaller antenna sizes, and it is capable of reading a large number of tags in a short time, making it suitable for large-scale distributed monitoring.²² Furthermore, chipless RFID sensors further enhance these advantages in SHM. By removing the silicon chip, these sensors possess a simpler structure that enables ultra-low-cost, printable manufacturing and direct integration into composite material fabrication processes.²³ Chipless RFID sensors are typically based on frequency encoding, where the resonant frequency is sensitive to physical quantities such as strain and cracks.²⁴ In the field of SHM, particularly for complex structures like composite pressure vessels, passive UHF RFID technology has emerged as an attractive wireless sensing solution due to its unique advantages. Qiang Wang et al. discussed the importance of structural health monitoring and strain sensing in detail, systematically demonstrating the fundamental principles and component architecture of RFID technology. Furthermore, their research introduced various strain sensor design schemes, including passive, active, semi-passive, UHF, chipless, and multi-sensory RF-based systems, while elucidating their respective performance advantages.²⁵ Reza Arablouei et al. proposed a novel RF signal enhancement technique based on photoluminescence to improve the sensitivity of 3D-printed passive RFID sensor arrays. Depending on the propagation mode (near-field or far-field), this technique can increase the received signal level by 2 dB to 8 dB, effectively mitigating data loss issues without altering transmission power or the number of sensors.^26,27 Sarma et al. designed a novel “smart skin” sensor employing sensitive microwave structures composed of cascaded split-ring resonators coupled with coplanar waveguide transmission lines. This sensor is capable of identifying the initiation and propagation of cracks in architectural structures, enabling continuous monitoring of damage at any surface point and supporting the simultaneous detection of multiple structural disturbances.²⁸

In the field of composite materials monitoring, Daniel Visan et al. introduced a wireless system integrated with machine learning models specifically for monitoring CFRP structures in aerospace. The system collects data via carbon nanotube (CNT) piezoresistive sensors embedded within the CFRP layers and transmits it wirelessly to a central server for intelligent processing.²⁹ Kang et al. conducted an in-depth review of technologies such as CNTs, piezoelectric materials, RFID, wireless sensing, optical fibers, and computer vision in structural monitoring, clarifying the working principles, deployment status, and future research directions for each.³⁰ Additionally, Alassane et al. developed a wireless induction node specifically designed for monitoring reinforced concrete structures. As part of a cyber-physical system, this node utilizes an embedded design that requires no battery power, achieving energy autonomy and long-term monitoring through radiative electromagnetic wireless power transfer.³¹

In summary, passive UHF RFID technology has laid a solid technical foundation for the construction of zero-power, full-lifecycle autonomous wireless sensor networks. By combining the energy harvesting and communication capabilities of RFID with low-power sensing technologies, fully passive sensing nodes can be developed to achieve long-term, large-scale, real-time structural monitoring. This technical approach circumvents the challenges of battery replacement and maintenance inherent in battery-dependent systems, providing a viable technical solution for the full-lifecycle health management of composite pressure vessels.

Pressure monitoring mechanism and force-electrical mapping theory

Mechanical analysis of pressure vessel deformation under pressure

When filament-wound composite pressure vessels are subjected to high-pressure gas loads, their stress state exhibits typical axisymmetric characteristics. To explore the mechanical response laws of the vessel during the 0–20 MPa pressure cycling process and to establish a quantitative mapping model between the internal pressure $P$ and the surface strain $ε$ of the vessel body, this section begins with an analytical derivation based on pressure vessel mechanical theory. First, an analytical model is established based on the geometric features of the vessel. The cylindrical section of the vessel can be regarded as a multi-layer wound cylindrical shell composed of a high-strength glass fiber and epoxy resin matrix composite. When the vessel is filled with high-pressure gas, the vessel wall undergoes outward radial expansion and longitudinal extension under the action of normal pressure P. According to the equilibrium equations for thin-walled pressure vessels, under the simplified condition of neglecting the constraint effects of the domes at both ends, the hoop stress $σ_{θ}$ and axial stress $σ_{z}$ in the middle of the cylinder can be expressed by the simplified formulas of the first strength theory as shown in equation (1):

{\begin{cases} σ_{h} = \frac{P \cdot D}{2 t} \\ σ_{a} = \frac{P \cdot D}{4 t} \end{cases}

(1)

where:

P

is the real-time internal pressure load.

D

is the mean radius of the gas cylinder.

t

is the total wall thickness of the composite winding layer. As indicated by the formula, the hoop stress is consistently twice the axial stress. This implies that the deformation of the cylinder surface is dominated by circumferential expansion, which serves as the core basis for selecting the orientation of the sensing tag. Since glass fiber reinforced polymer (GFRP) exhibits significant orthotropy, its deformation behavior under internal pressure depends not only on the magnitude of the pressure but is also constrained by the fiber winding angle. According to the constitutive equations in composite mechanics shown in equation (2), the circumferential strain on the cylinder surface can be characterized by anisotropic elastic constants.

ε_{θ} = \frac{σ_{θ}}{E_{θ}} - ν_{θ_{z}} \frac{σ_{z}}{E_{z}}

(2)

By substituting the stress components and rearranging the terms, the theoretical mapping model describing the relationship between the cylinder’s internal pressure and surface strain can be obtained, as shown in equation (3).

ε_{θ} = P \cdot (\frac{D}{2 t E_{θ}} - \frac{ν_{θ_{z}} D}{4 t E_{z}})

(3)

where:

E_{θ}

and

E_{z}

are the equivalent elastic moduli of the cylinder body in the circumferential (hoop) and axial directions, respectively. This model clearly indicates that within the linear elastic range of the cylinder material, the surface strain maintains a strict linear proportional relationship with the internal pressure. Due to the lower elastic modulus of glass fiber (GFRP) compared to carbon fiber, the cylinder surface produces a more sensitive displacement response under the same internal pressure increment, thereby providing a wider signal modulation space for the embedded sensing tags.

RFID tag strain sensing mechanism

When an RFID tag attached to the surface of a GFRP laminate is subjected to a tensile load along the fiber direction, the tag antenna undergoes synchronous deformation with the substrate. Within the elastic range, this geometric deformation directly alters the electrical length of the antenna, thereby leading to a shift in the resonance frequency.

Assuming the dipole antenna is subjected to tension along its longitudinal direction (length direction) and the longitudinal strain of the substrate material is $ε_{L}$ . According to the definition of strain, the physical length L of the antenna will undergo elongation:

L (ε_{L}) = L_{0} (1 + ε_{L})

(4)

where

L_{0}

is the initial length in the stress-free state. Meanwhile, due to the Poisson Effect, the antenna width

w

will undergo transverse contraction:

W (ε_{L}) = W_{0} (1 - v ε_{L})

(5)

where

v

is the Poisson’s ratio of the tag substrate. For slender dipole antennas, the change in length dominates the resonance characteristics, and the effect of width can be neglected as a higher-order infinitesimal.

According to transmission line theory, the resonance frequency of a half-wave dipole antenna is inversely proportional to its electrical length. By substituting equation (4) into the resonance frequency formula, the instantaneous resonance frequency under the loaded state can be obtained:

f_{r e s} (ε_{L}) = \frac{c}{2 L (ε_{L}) \sqrt{ε_{e f f}}} = \frac{c}{2 L_{0} (1 + ε_{L}) \sqrt{ε_{e f f}}} \approx \frac{f_{r e s, 0}}{1 + ε_{L}}

(6)

As shown in equation (6), an increase in tensile strain $ε_{L}$ leads to an increase in the denominator, causing the resonance frequency to shift toward lower frequencies, a phenomenon known as “redshift”. By defining the frequency shift as $Δ f = f_{r e s} (ε_{L}) - f_{r e s, 0}$ , and assuming small strain, a linear relationship can be approximately derived:

\frac{Δ f}{f_{r e s, 0}} \approx - ε_{L}

(7)

This equation³² indicates that the rate of change in the antenna’s resonance frequency is numerically approximately equal to the applied mechanical strain, but in the opposite direction.

Similar to the logic used in curing monitoring, the reader emits signals at a fixed frequency, such as 915 MHz. On the surface of the composite material, the tag is initially in a detuned state. When tension occurs, the resonance frequency shifts further toward lower frequencies (redshift), causing the resonance peak to move further away from the reader’s carrier frequency. This process exacerbates impedance mismatch, causing the system’s operating point to slide down the falling edge of the resonance curve, or away from the peak. According to the Lorentzian line-shape characteristics of the power transmission coefficient, as shown in equation (8), the RSSI signal will exhibit monotonic attenuation as strain increases.

R S S I (ε_{L}) \propto 10 \log_{10} (\frac{1}{1 + Q^{2} {(δ_{0} + ε_{L})}^{2}})

(8)

where

δ_{0}

is the initial detuning factor. This model theoretically explains why, during the elastic stretching phase, the RSSI signal exhibits a linear or quasi-linear downward trend that is negatively correlated with strain.

Numerical simulation of force-electrical coupling based on COMSOL

Finite element modeling of internal pressure response in gas cylinders

To achieve efficient simulation of the mechanical behavior of the gas cylinder, this section utilizes the “Solid Mechanics” module in the COMSOL Multiphysics multi-physics simulation platform to establish a finite element model based on 2D Axisymmetric spatial dimensions. By exploiting the rotational symmetry of the cylinder structure, computational complexity can be significantly reduced and convergence speed improved while maintaining calculation accuracy.

Geometric characterization

A dual-layer shell geometric model, comprising an Aluminum Liner and a Composite Layer, was established for the research object. The specific dimensions of the model are set as follows: the internal radius $R_{i n}$ is 150 mm; the wall thickness of the aluminum liner $t_{A l}$ is 3 mm; the thickness of the external glass fiber composite layer $t_{c o m p}$ is 5 mm; and the cylinder height $H$ is taken as 50 mm. By constructing two adjacent rectangular regions, the composite structural interface of the cylinder body section was simulated. Figure 1 displays the geometric view of the cylinder section generated based on 2D axisymmetric rotation. Different color mappings are used to distinguish the aluminum liner (inner layer) from the glass fiber composite layer (outer layer). The distribution of these color gradients intuitively illustrates the spatial geometric arrangement of the dual-layer structure and provides a topological reference for the subsequent finite element mesh generation.

Figure 1.

Geometric view of the cylindrical section generated by 2D axisymmetric rotation.

Material properties and constitutive models

The model assigns distinct material properties to different layer domains to simulate realistic mechanical responses: An Isotropic Linear Elastic model was adopted. The Young’s modulus was set to $E = 70 GPa$ and the Poisson’s ratio to $ν = 0.33$ to simulate the sealing and support layer inside the vessel. An Orthotropic Linear Elastic model was employed. Considering the circumferential winding characteristics of fibers in actual production, a precise mapping of material principal axes was conducted in the 2D axisymmetric coordinate system $(r, ϕ, z)$ :

Fiber reinforcement direction (Direction 1): Mapped to the circumferential direction ( $ϕ$ ). Axial direction (Direction 2): Mapped to the $z$ . Thickness direction (Direction 3): Mapped to the radial direction ( $r$ ). The core mechanical parameters were set as: Hoop modulus $E_{ϕ} = 45 GPa$ , and Axial/Radial moduli $E_{z} = E_{r} = 12 GPa$ . Corresponding shear modulus and Poisson’s ratio matrices were strictly defined to reflect the orthotropic characteristics of the Glass Fiber Reinforced Polymer (GFRP).

Loading and boundary conditions

To simulate the pressurized conditions of the gas cylinder in actual operation, the model boundary conditions are set as follows:

Internal load

A normal boundary load is applied to the innermost surface of the aluminum liner. Using the Parametric Sweep function, the pressure $P_{i n}$ is set from 0 to 20 MPa to simulate a stepped pressurization process.

Displacement constraints

A Roller constraint is applied to the bottom edge of the geometric model. This setting restricts the axial displacement at the bottom ( $u_{z} = 0$ ), ensuring that the model does not undergo rigid body displacement while allowing the cylinder wall to expand freely in the radial direction ( $r$ ). This allows for the accurate acquisition of the expansion deformation of the gas cylinder under a plane strain state.

Meshing and solver settings

A Free Triangular Mesh is used to discretize the geometric domains, and second-order triangular elements are selected to enhance solution accuracy. Local mesh refinement is performed at the interface between the liner and the composite layer, as well as on the outer surface areas, to ensure the accuracy of the strain gradient collection. Regarding the solver settings, a Stationary Solver is selected, which works in conjunction with the parametric sweeper to extract the structural responses under different pressure steps.

Electromagnetic simulation model of the sensing tag

To quantitatively explore the modulation mechanism of the surface strain of glass fiber composite gas cylinders on the electromagnetic characteristics of RFID tags, this study established a simulation model based on the COMSOL Multiphysics platform, incorporating the RF Module and structural mechanical features.

Antenna geometric topology and refined modeling

The sensing tag prototype selected for this study is the Alien 9662 UHF RFID tag. As shown in Figure 2, a 1:1 geometric reconstruction was performed based on the actual physical dimensions to restore the real physical response in the simulation. The main structure of the antenna adopts a typical Meander Line Dipole configuration, which aims to increase the effective electrical length within a limited physical space through the bending structure, thereby reducing the resonance frequency to the ultra-high frequency (UHF) operating band. During the geometric modeling process, the antenna was placed at the center of the substrate, with its key structural parameters set as follows:

Figure 2.

Construction of tag antenna simulation model.

Radiation element design

To match the 915 MHz center operating frequency set for this study, the optimized total antenna length ( $L_{t o t a l}$ ) is set to 119 mm. Each radiation arm length ( $L_{a r m}$ ) is 44.5 mm, containing 5 cycles of a meander structure ( $N = 5$ ). The height of the meander section ( $H_{m e a n d e r}$ ) is 15 mm with a pitch ( $W_{p i t c h}$ ) of 3.5 mm. A widened solder pad of 15 mm × 8 mm is provided at the end to adjust the terminal capacitive effect for matching assistance.

Impedance matching network

A T-Match Loop structure is used at the center of the antenna for interconnection with the chip. The loop length ( $L_{l o o p}$ ) is designed to be 30 mm and the loop height ( $H_{l o o p}$ ) is 15 mm. This inductive loop structure effectively cancels the internal capacitive impedance of the chip, achieving maximum power transfer.

Wire specifications

The trace width ( $W_{t r a c e}$ ) is uniformly set to 0.8 mm. To simulate the skin effect and approximate actual etching processes, the conductive layer thickness ( $t_{c o p p e r}$ ) is set to 0.05 mm ( $50 μ m$ ).

Material properties and dielectric environment definition

The substrate material in the simulation model is used to simulate the cylindrical wall of the filament-wound glass fiber gas cylinder. Based on the actual operating conditions of composite cylinders, the substrate material is defined as FR-4 (equivalent to Glass Fiber Reinforced Polymer, GFRP). Its relative permittivity is set to $ε_{r} = 4.5$ , and the dielectric loss tangent is set to $\tan δ = 0.02$ . The geometric dimensions of the substrate are set to 180 mm × 80 mm with a thickness ( $h_{s u b}$ ) of 2 mm to fully cover the near-field radiation zone of the antenna. The aluminum liner inside the gas cylinder, acting as a continuous metallic surface, exerts an image charge effect on the near-field distribution of the antenna, thereby causing an impedance shift. In the electromagnetic simulation of this study, the aluminum liner is defined as a Perfect Electric Conductor (PEC) boundary. Given that the composite winding layer has a thickness of 5 mm, it effectively serves as an insulating dielectric buffer between the antenna and the metallic liner. This separation mitigates the shielding effect of the metal on the antenna gain, ensuring the signal transmission reliability of the tag during the pressure monitoring process. The material for the conductive layer of the antenna is defined as Oxygen-Free Copper, with the electrical conductivity set to $5.8 \times 10^{7} ° S / m$ . To simulate radiation characteristics in open space, a spherical Air Domain with a radius of 150 mm was established around the model. The outer surface of the air domain is designated with a Scattering Boundary Condition, which can completely absorb outward-radiating electromagnetic waves, effectively eliminating interference from boundary reflections on the calculation of S-parameters.

Port excitation and conjugate impedance matching

In the radio frequency (RF) simulation, the chip is not constructed as a physical geometric model but is instead represented by an equivalent Lumped Port. The port is positioned at the feed gap in the center of the T-match loop, with a gap width ( $L_{g a p}$ ) of 0.5 mm. To accurately simulate the nonlinear impedance characteristics of the tag chip (Impinj Monza R6), the reference impedance of the port is not set to the standard $50 Ω$ . Instead, it is defined as the conjugate value of the chip’s complex impedance. It is known that the input impedance of the Monza R6 chip at 915 MHz is $Z_{c h i p} = 12 - j 120 Ω$ . According to the Maximum Power Transfer Theorem, the input impedance of the simulation port is set to: $Z_{p o r t} = Z_{c h i p}^{*} = 12 + j 120 Ω$ , This setting ensures that in the initial undeformed state, the antenna and the port (chip) are in an optimal conjugate matching state. Consequently, the model can sensitively reflect the subsequent impedance mismatch phenomenon caused by mechanical deformation. The excitation settings of the lumped port at the chip location are shown in Figure 3.

Figure 3.

Schematic diagram of lumped port excitation setup for chip positions.

Simulation results and analysis

Finite element analysis of pressure vessel deformation under pressure

Distribution characteristics of surface strain field on the gas cylinder

Figure 4 illustrates the 3D strain distribution contours of the entire gas cylinder and its longitudinal section under an internal pressure of 20 MPa. As shown in the figure, the gas cylinder exhibits significant radial expansion characteristics after being pressurized. In the central region of the cylinder body, the distribution of the first principal strain (i.e., hoop strain) demonstrates extremely high uniformity; the color transition in the contour plot is smooth, with no evident stress concentration phenomena observed.

Figure 4.

Schematic diagram of lumped port excitation setup for chip positions.

This demonstrates that in regions far from the constraints of the domes, the mechanical response of the gas cylinder aligns with the ideal stress state of a thin-walled cylindrical shell. Further observation of the sectional strain gradient reveals a slight decreasing trend in strain values from the liner toward the outer surface of the composite material. This occurs because, during the synergistic deformation of the dual-layer structure, the external fiber layer serves as the primary load-bearing component, while the liner is mainly responsible for sealing.

Under the 20 MPa full-scale operating condition, the maximum circumferential strain on the outer surface is approximately $8286 μ ε$ . This is significantly lower than the ultimate tensile strain of glass fiber reinforced polymer (typically greater than 2%), proving that the gas cylinder possesses an ample safety margin within the monitored pressure range.

Quantitative evaluation of internal pressure–strain response

To verify the reliability of the linear mapping relationship, this study extracted the mean hoop strain values from the sensing tag attachment area on the outer surface during the pressurization process from 0 to 20 MPa. Meanwhile, theoretical values were calculated according to the analytical formulas derived previously, incorporating the gas cylinder’s geometric parameters (

D = 308 mm, t = 8 mm

) and the equivalent modulus of the material. A comparison between the extracted simulation values and the theoretical calculated values is presented in Table 1.

Table 1.

Data table for comparison of circumferential strain under different internal pressure loads.

Internal pressure load $P (MPa)$	Theoretical calculated value $(μ ε)$	Simulation extracted value $(μ ε)$
0 (初始)	0	0
5	2055.0	2071.4
10	4110.0	4142.8
15	6165.0	6214.3
20	8220.0	8285.7

Performing a linear regression on the simulation data, the resulting response equation is $ε_{θ} = 414.29 \cdot P - 0.02$ . The goodness of fit ( $R^{2}$ ) reaches 0.99. The slope of $414.29 μ ε / MPa$ characterizes the mechanical sensitivity of this gas cylinder structure.

It should be noted that the finite element model established in this study focuses primarily on the cylindrical mid-section of the gas cylinder, simplifying the analysis by neglecting the boundary constraint effects of the domes on the shell. In actual working conditions, the transition region between the dome and the cylinder exhibits significant strain gradients and complex stress states due to geometric discontinuities, which introduces a degree of non-uniformity into the pressure-strain mapping logic of that area.

Consequently, the monitoring method proposed in this paper is currently most applicable to the uniform strain zone in the middle of the cylinder. In future work, we intend to establish a complete 3D model of the cylinder including the dome structures to conduct an in-depth investigation of the constraint effects in the transition zone and their specific impact on radio frequency signal modulation, thereby further expanding the monitoring coverage of the sensing system.

Impact of strain on tag parameters

By observing the energy distribution on the antenna surface in Figure 5 and extracting the Surface Current and Near-field Electric Field Norm (Norm E), it is found that the current is mainly concentrated in the T-match loop and the initial sections of the meander arms. This demonstrates that the T-match structure effectively achieves power matching between the chip’s complex impedance and the antenna radiator.

Figure 5.

Surface current vector distribution plot.

As the internal pressure of the gas cylinder increases, the axial tensile strain generated on its surface gradually increases from 0 to $8285.7 μ ε$ . Figure 6 illustrates the evolution curves of the antenna return loss (S11) as a function of frequency under different strain levels.

Figure 6.

Comparison of return loss (S11) curves under different strains.

It is observed that as strain increases, the minimum point of the S11 curve (i.e., the resonance frequency) undergoes a distinct leftward translation, exhibiting a significant “redshift” characteristic. This is because the axial tension directly leads to an increase in the physical length of the antenna’s meander arms, which in turn increases the effective electrical length of the antenna. Experimental results indicate that the resonance frequency shifts continuously downward from the initial 915 MHz to approximately 913.5 MHz. Although the observed resonance frequency redshift within the 0 to 20 MPa pressure range is relatively small, this variation is sufficient to trigger a significant sensing response. According to Lorentzian lineshape characteristics, a minor frequency shift causes a severe impedance mismatch between the antenna and the chip, thereby converting mechanical deformation into a monotonic attenuation of the RSSI. Considering that the Impinj R420 reader used in the experiments possesses millisecond-level time resolution and high-precision signal strength detection capabilities, the signal fluctuations generated by a redshift of this magnitude are far above the environmental noise floor of the system, ensuring the robustness of the pressure monitoring logic.

To quantitatively evaluate the sensing sensitivity of the system, this study extracted the resonance frequency values under different strain states and performed a linear regression fit, as shown in Figure 7. The figure illustrates the linear fitting relationship between the resonant frequency and the surface strain. The fitting equation is $f_{r} = - 1.81 \times 10^{- 4} ε + 915$ , where the slope of $- 1.81 \times 10^{- 4} MHz / μ ε$ represents the strain sensitivity of the sensor (approximately $0.181 kHz / μ ε$ ). According to this linear model, a resonant frequency shift of 1 MHz corresponds to a strain change of approximately $5523.8 μ ε$ . This highly linear correlation demonstrates that by monitoring the frequency response of the tag, the mechanical deformation state of the composite cylinder can be accurately inverted.

Figure 7.

Linear fitting of resonance frequency versus strain.

Analysis of the RSSI mapping model

Based on the modified Friis backscatter transmission equation, the spectral distribution curves of the sensing tag’s reflected signal strength under different strain levels were obtained through simulation. As shown in Figure 8(a) as the strain gradually increases from 0, the spectral envelope of the signal strength exhibits significant asymmetric evolutionary characteristics. It can be seen from the spectral evolution trend that the position of the resonance peak undergoes a significant redshift as strain increases, which remains highly consistent with the previously described shift in S11 parameters. Simultaneously, the continuous degradation of the impedance matching state, caused by the antenna’s geometric deformation under strain, leads to a decrease in reflected power across various frequency points to varying degrees.

Figure 8.

Characterization of reflected signal intensity response to strain. (a) Spectrogram of reflected signal intensity. (b) Linear fitting of RSSI versus strain at a fixed frequency.

Considering that RFID readers typically operate at a fixed frequency, this study extracted the signal strength values at the center frequency (915 MHz) for quantitative analysis. By cascading the mechanical calibration data with the electromagnetic simulation results, the simulated mapping data was summarized as shown in Table 2.

Table 2.

Simulation mapping table of cylinder internal pressure, surface strain, and RF signal characteristics.

Internal pressure $P (MPa)$	Hoop strain $ε (μ ε)$	Fixed frequency $R S S I (915 MHz)$
0 (初始)	0	−47.000
5	2071.4	−47.098
10	4142.8	−47.171
15	6214.3	−47.291
20	8285.7	−47.390

As shown in the table, during the process of increasing the internal pressure of the gas cylinder from 0 MPa to 20 MPa, the RSSI value at the fixed frequency (915 MHz) attenuates monotonically from −47 dBm to −47.39 dBm. The Impinj R420 reader used in this study features a signal strength resolution of 0.1 dB. The observed variations are significantly higher than the hardware detection threshold, which is sufficient to support reliable pressure monitoring. To further quantify the sensing performance of this indicator, a linear regression analysis was performed on the fixed-frequency RSSI and strain data, and the results are illustrated in Figure 8(b).

Although the peak intensity of the reflected signal in Figure 8(a) shows only slight fluctuations with increasing strain, its impact on sensing performance is primarily reflected in the response at a fixed frequency point. Since the RFID reader operates under a single carrier at 915 MHz, the redshift of the resonance peak causes the operating state at this specific frequency to shift downward along the edge of the Lorentzian envelope. Figure 8(b) displays the linear regression analysis between RSSI and strain at a fixed frequency, yielding the fitting equation: $RSSI = - 48.7 ε - 46.9$ The slope of $- 48.7 dB / ε$ clearly defines the sensing sensitivity of the system, meaning that for every $1000 μ ε$ increase in strain, the signal strength decreases monotonically by approximately 0.05 dB. The fitting results show a very high linear correlation between the fixed-frequency RSSI and strain, with a coefficient of determination R² reaching 0.995. This indicates that RSSI exhibits robust response characteristics throughout the full monitoring range of the glass fiber cylinder.

Although the Lorentzian lineshape described by equation (8) possesses intrinsic nonlinearity, the RSSI attenuation observed in this study exhibits significant linear characteristics. This is primarily because, during the pressure cycling from 0 to 20 MPa, the maximum surface strain produced is approximately $8285.7 μ ε$ , corresponding to a resonant frequency redshift of 1.5 MHz. Relative to the carrier frequency of 915 MHz, this frequency shift is quite small. According to the Taylor series expansion, within the narrow bandwidth on either side of the resonance peak, the first-order derivative of the power transmission coefficient dominates. This results in the system’s operating point remaining consistently within a quasi-linear response region. When the strain further increases beyond $10000 μ ε$ or the frequency shift exceeds 2 MHz, the influence of second-order nonlinear terms will gradually emerge, causing the mapping logic to deviate from a linear response.

Analysis of the impact of curvature on tag response characteristics

Composite cylinders are typical curved structures, and in practical applications, tags are usually attached to a cylindrical surface with a specific curvature. Compared to a flat installation state, surface attachment further alters the geometric configuration of the tag antenna, the current distribution paths, and the surrounding electromagnetic boundary conditions, thereby affecting its impedance matching state. Therefore, based on the experimental results of flat specimens, it is necessary to further analyze the impact of curvature factors on the tag’s response characteristics.

To analyze the mechanism of curvature’s effect on the tag’s electromagnetic properties, and with reference to the equivalent analysis method for cylindrical cavities, the tag dielectric structure attached to a cylindrical metal surface can be approximated as a cylindrical cavity model. In this case, the conductor surface and the grounded cylindrical surface are treated as electric wall boundaries, while the dielectric layer forms an equivalent resonant cavity. Under the curved state, the electric field distribution within the cavity satisfies the wave equation in cylindrical coordinates, as shown in equation (9):

\frac{1}{ρ} \frac{\partial E_{ρ}}{\partial ρ} - \frac{1}{ρ^{2}} \frac{\partial^{2} E_{ρ}}{\partial ϕ^{2}} - \frac{\partial^{2} E_{ρ}}{\partial z^{2}} + \frac{\partial^{2} E_{z}}{\partial z \partial ρ} - k^{2} E_{ρ} = 0

(9)

where:

E_{ρ}

and

E_{z}

are the radial and axial components of the electric field in the cylindrical coordinate system, respectively;

ρ

ϕ

, and

z

are the radial, azimuthal, and axial coordinates of the cylindrical system, and k is the wavenumber. The transverse wavenumber in the curved state can be expressed as equation (10):

k_{ρ} = \frac{m π}{2 θ (a + h)}

(10)

Furthermore, the expression for the resonant frequency of the tag antenna can be derived as equation (11):

f_{r} = \frac{1}{2 \sqrt{ε μ}} \sqrt{{(\frac{m π}{2 θ (a + h)})}^{2} + {(\frac{n π}{L})}^{2}}

(11)

where: a is the cylinder radius, h is the thickness of the dielectric layer,

θ

is the bending angle, L is the characteristic dimension of the tag’s straight edge, and

ε

and

μ

are the equivalent permittivity and permeability of the medium, respectively.

As indicated by the above equation, the resonant frequency is directly related to the cylinder radius and the bending angle. When a tag transitions from a flat state to a curved attachment state, both its equivalent electrical length and the cavity mode conditions undergo changes, leading to a shift in the resonance point. Simultaneously, variations in curvature alter the electric field storage state and power dissipation within the cavity, which in turn affects the radiation efficiency and radiation gain of the tag antenna. For passive UHF RFID tags, these changes are ultimately reflected in key electromagnetic performance indicators such as the power transmission coefficient, gain, and read range. Consequently, it is evident that curvature is a significant factor influencing the electromagnetic response characteristics of the tag.

To verify the aforementioned mechanisms at the simulation level, this study established a 3D electromagnetic model in HFSS featuring a tag antenna attached to the surface of cylindrical metallic objects with varying radii. While maintaining constant structural parameters, material properties, and excitation conditions for the tag, only the cylinder radius was varied—specifically at 15 mm, 20 mm, 25 mm, 30 mm, and 35 mm—to analyze the tag’s transmission coefficient, antenna gain, and read range.

Figure 9(a) presents the 2D radiation pattern of the tag when attached to the surface of a cylindrical object. As shown in the figure, the tag maintains a distinct primary radiation direction under cylindrical surface conditions; however, its radiation morphology differs from the ideal flat state. This indicates that changes in curvature exert an influence on the far-field radiation distribution of the tag antenna. Figure 9(b) illustrates the relationship between the tag antenna’s read range and frequency under different cylindrical radii. The results show that near 915 MHz, there are significant differences in the read range across the various radii. Furthermore, as the cylinder radius increases, the peak read range gradually decreases. This demonstrates that curvature not only shifts the resonant state of the tag but also further impacts the ultimate read/write performance.

Figure 9.

RFID tag performance under different curvature conditions. (a) The radiation pattern of a tag attached to a cylindrical surface. (b) Under different radius conditions, the read range curves of the tag antenna.

Table 3 summarizes the simulation results of the tag antenna attached to the surfaces of cylindrical metallic objects with different radii at 915 MHz. The results in the table indicate that as the cylinder radius increases, the power transmission capability of the tag antenna remains relatively stable overall; however, both the radiation gain and read range exhibit a downward trend. This suggests that the curvature of the cylindrical surface has a direct impact on the electromagnetic response characteristics of the tag. As the tag transitions from a surface with a smaller radius to one with a larger radius, its reading performance shows a degree of attenuation.

Table 3.

Simulation results of the tag antenna mounted on cylinders of different radii.

Parameter	15 mm	20 mm	25 mm	30 mm	35 mm
Transmission coefficient	0.54	0.57	0.54	0.56	0.54
Gain (dBi)	−11.68	−13.34	−14.84	−15.84	−16.85
Read range	2.51	2.13	1.75	1.58	1.38

Based on the combined theoretical analysis and HFSS simulation results, it can be concluded that curvature exerts a systematic influence on the electromagnetic response characteristics of the tag by altering its initial detuning state, electric field distribution, and radiation capability. Consequently, while experiments on flat specimens can reveal the fundamental response patterns of the tag to strain variations, the additional effects brought about by curvature must be further considered when extending these findings to real-world curved composite cylinder scenarios.

Experimental validation and data discussion

Experimental materials and specimen preparation

The experimental specimens were fabricated using glass fiber fabric reinforcement combined with epoxy resin to form the composite material representing the tested samples. The glass fiber fabric used has a density of 130 $g / m^{2}$ and is characterized by high-strength properties. Specimen preparation strictly followed the ISO 527-5:2009 standard (Plastics—Determination of tensile properties—Part 5: Test conditions for unidirectional fibre-reinforced plastic composites), ensuring that both geometric dimensions and mechanical load-bearing characteristics conform to international academic norms. It should be noted that while actual gas cylinders include an aluminum liner structure, this experiment utilized a single-material GFRP (Glass Fiber Reinforced Polymer) plate for fundamental characterization. The primary objective of this configuration was to isolate complex metallic near-field interference, thereby accurately determining the intrinsic strain response characteristics of the RFID sensor on a composite substrate. Since the mechanical behavior of a gas cylinder under pressure involves the transfer of strain from the liner to the outer composite layers, the GFRP plate tensile test effectively simulates the actual stress state of the cylinder’s outer surface.

To ensure consistency in the curing quality of the composite materials, the specimen preparation process was conducted in a controlled laboratory environment. During preparation, the ambient temperature was strictly maintained at 25 ± 2°C, and the relative humidity was kept at 50 ± 5%. To investigate the influence of substrate structural dimensions on sensing characteristics, two types of equivalent specimens with different thicknesses were prepared by controlling the number of material stacking layers, as shown in Figure 10:

1. Thin Specimen: Fabricated using a 4-layer symmetrical stacking sequence, aimed at simulating the sensing response of thin-walled sections of gas cylinders or low-pressure vessels.

2. Thick Specimen: Fabricated using an 8-layer symmetrical stacking sequence, used to evaluate the impact of increased dielectric thickness on radio frequency (RF) energy attenuation and strain transfer sensitivity.

Figure 10.

Characterization of reflected signal intensity response to strain.

The lamination process was conducted on a glass surface without additional vacuuming or traction operations on the top. The samples were allowed to air-dry naturally for 24 h and were then peeled from the glass surface after curing, followed by mechanical processing to achieve a uniform shape.

The sensing unit consists of an RFID tag and metal foil resistance strain gauges, forming a collaborative monitoring system that enables cross-validation of experimental data. A high-strength, low-loss epoxy adhesive was utilized to attach the tag to the geometric center of the specimen. To ensure the fidelity of strain transfer, the longitudinal axis of the tag’s meander dipole must be strictly aligned with the tensile axis of the specimen. A resistance strain gauge was attached to the back of the specimen, symmetrical to the RFID tag’s position; this gauge serves as the “gold standard” for strain measurement, providing real-time monitoring of the actual strain values during the tensile process.

The bonding of the resistance strain gauges followed a standardized procedure: First, the predetermined locations on the specimens were lightly sanded with 400-grit sandpaper to remove surface laitance, followed by repeated wiping with anhydrous ethanol until no visible impurities remained. Subsequently, an appropriate amount of high-strength cyanoacrylate adhesive was applied, and the strain gauge was precisely aligned with the central axis of the specimen. During the bonding process, constant finger pressure was applied for approximately 60 s to expel air bubbles and ensure an extremely thin and uniform adhesive layer. Finally, the specimens were left undisturbed for over 1 h to guarantee complete strain transfer integrity. Figure 11 illustrates the antenna sensor integrated onto the specimen surface along with the five metal foil resistance strain gauges used for benchmark calibration below it.

Figure 11.

Physical GFRP flat specimen and sensor integration layout.

Experimental setup and synchronized data acquisition

In terms of hardware link construction, the load application module utilized an Instron 5985 microcomputer-controlled electronic universal testing machine. This equipment possesses a load capacity of up to 200 kN and micron-level displacement control precision, providing a robust quasi-static tensile load for the GFRP specimens. The specimens were vertically secured using the testing machine’s high-performance hydraulic grips. During the clamping process, the gripping pressure was strictly set to 0.3 MPa, a parameter that effectively avoids local crushing damage to the composite substrate.

For strain benchmark monitoring, an array of metal foil resistance strain gauges was integrated onto the specimen surface. Data were captured in real-time via a DH3816 N static strain testing system, serving as the benchmark data for evaluating the sensing accuracy of the RFID tags. The RF sensing module employed an ultra-high frequency (UHF) RFID reader based on the Impinj R420 chip, connected via low-loss RF cables to a 9 dBi gain circularly polarized antenna. To mitigate strong multipath interference caused by the metal components of the testing machine, the reader antenna was fixed at a vertical distance of 30 cm from the center of the GFRP specimen, ensuring the tag remained within the antenna’s far-field radiation zone. Simultaneously, the longitudinal axis of the RFID tag’s meander dipole was kept strictly parallel to the primary tensile axis of the testing machine. Prior to the experiment, a system self-check and baseline zeroing were performed to record the initial signal strength in the unloaded state. The physical layout of the experiment is shown in Figure 12.

Figure 12.

Experimental procedure for the tensile testing of GFRP flat specimens.

The experiment adopted a displacement-controlled quasi-static stepped loading mode with a loading rate set at 1 mm/min to simulate the stress conditions of a slowly pressurized gas cylinder. Loading was terminated once the tensile strain reached the upper limit preset in the simulation. The system utilized a synchronization strategy of “unified software triggering and post-hoc timestamp alignment,” simultaneously activating the testing machine controller, the strain gauge, and the reader monitoring terminal at the start of loading. The reader operated in Dense Reader Mode, returning data with millisecond-level timestamps via an Ethernet interface in real-time. To address the random fluctuations of RF signals in complex laboratory environments, the system continuously collected no fewer than 500 signal samples at each strain point and calculated their statistical mean. Background noise was effectively filtered out by eliminating anomalous outliers.

Although this study currently focuses on static pressure mapping, future dynamic monitoring can refer to methods for structural modal parameter identification and damage detection under ambient excitation.³³ By introducing more advanced state-recognition algorithms to perform a deep deconstruction of the fluctuation characteristics in the reflected signals, real-time health diagnosis of composite gas cylinders under complex dynamic loads can be achieved.

Analysis of experimental results

Based on the timestamp alignment scheme designed above, this study extracted the load data output by the Instron 5985 testing machine, the benchmark strain data recorded by the DH3816 N strain gauge, and the RSSI characteristic values acquired by the reader. To eliminate random errors caused by specimen fabrication processes and stochastic fluctuations in the spatial electromagnetic environment, the RSSI values listed in Table 4 represent the statistical mean values obtained from repeated experiments under each respective operating condition.

Table 4.

Summary mapping table of strain and signal strength for GFRP laminates.

Sample ID	Reference strain $ε (μ ε)$	Mean RSSI of 4-layer specimen	Mean RSSI of 8-layer specimen
GFRP-130-XL-0	0	−47.125	−47.312
GFRP-130-XL-5	2071.4	−47.228	−47.428
GFRP-130-XL-10	4142.8	−47.359	−47.511
GFRP-130-XL-15	6214.3	−47.410	−47.610
GFRP-130-XL-20	8285.7	−47.552	−47.769

By comparing the simulation data presented above with the measured data in Table 3, it can be observed that both maintain high consistency in terms of magnitude and evolutionary trends. In the initial zero-strain state, the RSSI value of the 8-layer (8L) specimen is slightly lower than that of the 4-layer (4L) specimen, which validates the theoretical prediction mentioned earlier: an increase in dielectric thickness leads to greater dielectric loss of radio frequency (RF) energy. The least squares method was employed to perform a linear regression analysis on the experimental 915 MHz fixed-frequency RSSI data, the results of which are displayed in the fitting plots in Figure 11. The analysis reveals that for both the thin specimen (4L) and the thick specimen (8L), the evolution of reflected signal strength with mechanical strain exhibits a significant negative linear correlation (Figure 13).

Figure 13.

Fitting plot of RF signal intensity evolution versus reference strain for GFRP laminates. (a) Fitting curve of 4-layer layup specimens. (b) Fitting curve of 8-layer layup specimens.

As shown in the figures, compared to the 4-layer (4L) specimen, the absolute value of the signal attenuation slope for the 8-layer (8L) specimen increases significantly. This indicates that in the surface-sensing mode, a thicker substrate exerts a stronger amplification effect on the impedance modulation caused by strain, thereby enhancing the monitoring sensitivity of the system. Although the study shows that increasing substrate thickness can significantly improve sensing sensitivity, a balance must be struck between the shear lag in mechanical transfer and the dielectric loss of electromagnetic waves. Experimental data suggest that an integrated thickness in the range of 1.0–1.5 mm (corresponding to 4L–8L ply stacks) represents the “golden window” for achieving high sensitivity and high-fidelity monitoring. If the thickness is increased further, the system may face the dual challenges of signal distortion and structural weight gain.

Additionally, it should be noted that the read distance significantly affects the system signal-to-noise ratio through the Friis Transmission Equation. In this experiment, the distance was fixed at 30 cm to ensure the initial signal strength (−47 dBm) was well above the reader’s detection threshold. Limited by the hardware signal resolution of 0.1 dB and combined with the fitted sensitivity of −48.705 dB/ $ε$ , the theoretical minimum detectable strain of this system is estimated to be approximately 2053 $μ ε$ . When the surface strain is lower than this threshold, the signal attenuation may be masked by the quantization error of the reader, which constitutes the lower monitoring limit under the current configuration.

Considering that the established simulation model is based on the typical wall thickness of a gas cylinder, while this section utilizes ultra-thin scaled specimens, a systematic bias exists between the two in terms of initial RSSI absolute values and absolute sensitivity (slope) due to dielectric scale effects. To eliminate the influence of dimensions and thickness gains, this study employs a normalization analysis method. The initial RSSI values for both simulation and experiment are calibrated as benchmark reference points, focusing on the incremental mapping trends of the signal as it changes with strain. Superimposing the normalized measured data onto the simulation curves reveals that the linear evolution trend of the measured RSSI relative to the benchmark strain exhibits extremely high consistency with the simulation prediction model. Specifically, the relative deviation between the sensitivity slope of the 4L specimen and the theoretical prediction is controlled within 6.47%. Meanwhile, the 8L specimen demonstrates a larger signal attenuation slope (with a deviation of approximately 12.64%), further confirming the enhancement effect of fringing field modulation caused by the increased dielectric substrate thickness. The primary reasons for these minor deviations can be summarized into the following two points:

Contact thermal resistance and adhesive layer effects

Despite using high-strength adhesives, the non-uniformity of the adhesive layer thickness during the manual attachment process may introduce slight initial impedance fluctuations, which are difficult to fully replicate in a purely idealized simulation model.

Multipath interference effects

Although layout precautions were taken, the metal hydraulic grips of the testing machine inevitably cause slight distortions in the far-field radiation pattern of the reader antenna, leading to a minor offset between the absolute RSSI values and the simulation.

Although this study currently relies on single-point resistance strain gauges as a benchmark for performance evaluation, future work could introduce Digital Image Correlation (DIC) technology to perform full-field strain characterization across the entire antenna attachment area. This would facilitate the quantitative verification of strain transfer uniformity between the antenna and the composite substrate, further confirming the integrity of the antenna-substrate coupling logic in the spatial dimension. Compared to traditional SHM technologies such as FBG or AE, the passive UHF RFID sensing logic employed in this method maintains linearity while significantly reducing system integration costs and wiring complexity. While FBG offers advantages in static measurement precision, the wireless and passive nature of this method—achieved through mechanical-electrical coupling—is better suited for the lifecycle management of large-scale distributed gas cylinder arrays, such as those found in hydrogen refueling stations.

Conclusion

This paper proposes and validates a wireless sensing method utilizing passive UHF RFID technology for monitoring the internal pressure of composite gas cylinders. The primary conclusions are as follows:

1. Established a complete mechanical-electrical mapping theoretical model. Analytical derivations and finite element simulations demonstrate that within the linear elastic range of 0–20 MPa, the surface strain of the cylinder body and the internal pressure satisfy a rigorous linear proportional relationship.

2. Revealed the physical mechanism of signal modulation. Mechanical strain induces a “redshift” in the resonant frequency by altering the electrical length of the antenna. At a fixed carrier frequency, this is converted into a monotonically attenuating RSSI signal.

3. Elucidated the impact of curvature on sensing characteristics. Through cylindrical cavity model analysis and HFSS simulations, it was found that curvature variations exert a systematic influence on tag gain and reading performance by altering the initial detuning state and electric field distribution. Specifically, the read range decreases as the cylinder radius increases.

4. Experimentally verified the reliability of the monitoring system. Experimental results from glass fiber reinforced polymer (GFRP) laminates show that the RSSI response possesses extremely high linearity and excellent sensitivity. Due to its wireless and passive advantages, this method is well-suited for large-scale distributed monitoring, such as in hydrogen refueling stations. Future work will focus on introducing DIC technology to verify full-field strain transfer and establishing full-scale 3D models including the domes to optimize monitoring accuracy in transition zones.

Footnotes

ORCID iD

Jiazhong Xu

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Intelligent Wireless Detection Technology and In-service Health Monitoring Research for Plastic Liner Fiber-Wound Gas Cylinders; CQSJKJDW2024017.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.

References

Farhood

Singal

. Prediction of first ply failure of composite pressure vessels under internal pressure: a review. Anbar J Eng Sci 2022; 13(1): 76–84. https://doi.org/10.37649/aengs.2022.175883

Bouhala

Karatrantos

Reinhardt

, et al. Advancement in the modeling and design of composite pressure vessels for hydrogen storage: a comprehensive review. J Compos Sci 2024; 8(9): 339. https://doi.org/10.3390/jcs8090339

Zeng

, et al. Study on winding forming process of glass fiber composite pressure vessel. Materials 2025; 18(11): 2485. https://doi.org/10.3390/ma18112485

Bouhala

Polesel

Karatrantos

, et al. Review of state-of-the-art of structural health monitoring in hydrogen composite pressure vessels. Compos, Part C: Open Access 2025; 18: 18. https://doi.org/10.1016/j.jcomc.2025.100635

Jiang

Liu

, et al. Damage characterization of carbon fiber composite pressure vessels based on modal acoustic emission. Materials 2022; 15(14): 4783. https://doi.org/10.3390/ma15144783

Gong

Yan

, et al. Real-time vacuum degree characterization of aerospace vacuum pressure vessels based on FBG monitoring information. Meas Sci Technol 2026; 37: 015102. https://doi.org/10.1088/1361-6501/ae26b0

Gong

, et al. Vacuum loss state monitoring of aerospace vacuum pressure vessels based on quasi-distributed FBG sensing technology. Struct Durab Health Monit 2025; 19(3): 473–498. https://doi.org/10.32604/sdhm.2024.057916

Karapanagiotis

Breithaupt

Duffner

, et al. Real-time monitoring of hydrogen composite pressure vessels using surface-applied distributed fiber optic sensors. JPhys Photonics 2025; 7(2): 025016. https://doi.org/10.1088/2515-7647/adb9ac

Alblalaihid

Aldoihi

Alharbi

. Structural health monitoring of fiber reinforced composites using integrated a linear capacitance based sensor. Polymers 2024; 16(11): 1560. https://doi.org/10.3390/polym16111560

10.

Feng

Wang

, et al. A high-performance, sensitive, low-cost LIG/PDMS strain sensor for impact damage monitoring and localization in composite structures, nanotechnology. IoP Publ Ltd 2024; 35(35): 355702. https://doi.org/10.1088/1361-6528/ad5298

11.

Barragan

Kell

Liu

, et al. Next-generation embedded printed sensors for near-field monitoring of high-performance composites. Adv Eng Mater 2025; 27(4): 2401332. https://doi.org/10.1002/adem.202401332

12.

Bertram

Brink

Lang

. Wireless, material-integrated sensors for strain and temperature measurement in glass fibre reinforced composites. Sensors 2023; 23(14): 6375. https://doi.org/10.3390/s23146375

13.

Garcia-Etxabe

Mendinueta

Camacho-Iglesias

, et al. Influence of sample position on strain monitoring in composite materials using magnetic microwires. Sensors 2025; 25(16): 4892. https://doi.org/10.3390/s25164892

14.

Wan

, et al. A passive and wireless sensor based on RFID antenna for detecting mechanical deformation. IEEE Open J 2020; 1: 426–434. https://doi.org/10.1109/OJAP.2020.3015782

15.

Pecho

Hruz

Novak

, et al. Internal damage detection of composite structures using passive RFID tag antenna deformation method: basic research. Sensors 2021; 21(24): 8236. https://doi.org/10.3390/s21248236

16.

Wang

Dong

Wang

, et al. Passive wireless dual-tag UHF RFID sensor system for surface crack monitoring. Sensors 2021; 21(3): 882. https://doi.org/10.3390/s21030882

17.

Wang

Leon

, et al. Wireless crack sensing using an RFID-based folded patch antenna. In: Biondini

Frangopol

(eds). Bridge maintenance, safety, management, resilience and sustainability. CRC Press-Taylor & Francis Group, 2012, pp. 824–830.

18.

Azeem

Alam

, et al. Application of filament winding technology in composite pressure vessels and challenges: a review. J Energy Storage 2022; 49: 49. https://doi.org/10.1016/j.est.2021.103468

19.

Klute

Metrey

Gary

, et al. In-Situ structural health monitoring of composite-overwrapped pressure vessels. SAMPE J 2016; 52(2): 7–17.

20.

Meemary

Vasiukov

Deleglise-Lagardere

, et al. Sensors integration for structural health monitoring in composite pressure vessels: a review. Compos Struct 2025; 351: 118546. https://doi.org/10.1016/j.compstruct.2024.118546

21.

Zhang

Liu

Shen

, et al. A review of radio frequency identification sensing systems for structural health monitoring. Materials 2022; 15(21): 7851. https://doi.org/10.3390/ma15217851

22.

Ran

S-C

Wang

Q-A

Wang

J-F

, et al. A concise state-of-the-art review of crack monitoring enabled by RFID technology. Appl Sci-Basel 2024; 14(8): 3213. https://doi.org/10.3390/app14083213

23.

Wang

Lin

, et al. Condition monitoring of composite overwrap pressure vessels using MXene sensor. Int J Pres Ves Pip 2021; 191: 104349. https://doi.org/10.1016/j.ijpvp.2021.104349

24.

Patre

. Passive chipless RFID sensors: concept to applications-a review. IEEE J Radio Freq Identif 2022; 6: 64–76. https://doi.org/10.1109/JRFID.2021.3114104

25.

Liu

Wang

Jiao

, et al. Review of wireless RFID strain sensing technology in structural health monitoring. Sensors 2023; 23(15): 1–23. https://doi.org/10.3390/s23156925

26.

Pekgor

Arablouei

Nikzad

, et al. Displacement estimation via 3D-Printed RFID sensors for structural health monitoring: leveraging machine learning and photoluminescence to overcome data gaps. Sensors 2024; 24(4): 1–18. https://doi.org/10.3390/s24041233

27.

Pekgor

Arablouei

Nikzad

, et al. Displacement estimation using 3D-Printed RFID arrays for structural health monitoring. Sensors 2022; 22(22): 1–19. https://doi.org/10.3390/s22228811

28.

Dey

Bhattacharyya

Sarma

, et al. A novel “smart skin” sensor for chipless RFID-based structural health monitoring applications. IEEE Internet Things J 2021; 8(5): 3955–3971. https://doi.org/10.1109/jiot.2020.3026729

29.

Pop

Tudose

Visan

, et al. A machine learning-driven wireless system for structural health monitoring. INCAS Bulletin 2024; 16(3): 77–93. https://doi.org/10.13111/2066-8201.2024.16.3.8

30.

Kang

Zhu

Cai

, et al. A concise review of state-of-the-art sensing technologies for bridge structural health monitoring. Sensors 2025; 25(17): 1–18. https://doi.org/10.3390/s25175460

31.

Loubet

Sidibe

Herail

, et al. Autonomous industrial IoT for civil engineering structural health monitoring. IEEE Internet Things J 2024; 11(5): 8921–8944. https://doi.org/10.1109/jiot.2023.3321958

32.

Marindra

AMJ

Pratama

Suroso

. Non-invasive frozen meat monitoring system using UHF RFID tag antenna-based sensing and RSSI. Int J Adv Sci Eng Inf Technol 2023; 13(1): 1–7. https://doi.org/10.18517/ijaseit.13.1.16919

33.

Zhang

Lai

Wang

. Structural modal parameter recognition and related damage identification methods under environmental excitations: a review. Struct Durab Health Monit 2025; 19(1): 25–54. https://doi.org/10.32604/sdhm.2024.053662