African Journal of Applied Mathematics: Theory for Engineering Systems

Full Text

African Journal of Applied Mathematics: Theory for Engineering Systems | Vol. 2, No. 1, 202 6 African Journal of Applied Mathematics: Theory for Engineering Systems Vol. 2, No. 1, pp. 1–48 | 202 6 | DOI: 10. XXXXX /ajamtes.2025.0037 [ ORIGINAL RESEARCH ARTICLE — STRUCTURAL DYNAMICS | BRIDGE ENGINEERING | VEHICLE-BRIDGE INTERACTION] Dynamic Response Analysis of Box Girder Bridges Under High-Speed Vehicle Passages Aduot Madit Anhiem Department of Civil Engineering, Universiti Teknologi PETRONAS, Seri Iskandar 32610, Perak, Malaysia Email: aduot.madit2022@gmail.com | Received: 5 January 2026 | Revised: 12 January 2026 | Accepted: 18 January 202 6 | Published: 11 March 202 6 ABSTRACT Box girder bridges are the predominant structural form for medium-to-long span road crossings in Sub-Saharan Africa, valued for their high torsional stiffness, aerodynamic stability, and structural efficiency. However, the dynamic response of these bridges to high-speed vehicle passages — encompassing resonance excitation, vehicle-bridge interaction (VBI) coupling, and road surface roughness effects — remains inadequately characterised for the specific conditions prevailing on African highway corridors, where vehicle speeds are increasing following road improvement programmes and vehicle weights frequently exceed design assumptions. This paper presents a comprehensive dynamic response analysis of a 40 m single-span twin-cell steel box girder bridge subjected to high-speed vehicle passages at speeds of 60–180 km/h. A coupled vehicle-bridge interaction model is formulated, combining a quarter-car vehicle model (sprung and unsprung masses, primary and secondary suspension stiffness and damping) with an Euler-Bernoulli beam model of the box girder discretised using 3D shell finite elements (4,800 elements, 29,040 degrees of freedom). Road surface roughness is generated stochastically following the ISO 8608 power spectral density classification (Classes A–D). The governing equations of motion are integrated numerically using the Newmark-beta method with time step dt = 0.002 s. Natural frequencies, mode shapes, midspan deflection time histories, bending moment and shear force envelopes, deck accelerations, and dynamic amplification factors (DAF) are computed for a comprehensive parametric matrix of vehicle speeds, vehicle weights (10–55 tonne gross vehicle weight), road roughness classes, span-to-depth ratios, and damping ratios. Key findings include: (i) the DAF for the 40 m box girder at the design speed of 120 km/h is 1.28, exceeding the EN 1991-2 code value of 1.25 by 2.4% for road roughness Class B; (ii) DAF is strongly sensitive to road roughness class, increasing from 1.10 (Class A) to 1.53 (Class D) at 120 km/h; (iii) multi-vehicle convoy passages with headways less than the span length (40 m) generate DAF values up to 1.46, exceeding single-vehicle design values by 13.6%; (iv) resonance conditions occur at specific vehicle speeds (46, 91, 137, 183 km/h) that depend on the fundamental frequency and span length, producing sharp DAF peaks that are not captured by current code formulations; and (v) a tuned mass damper (TMD) with mass ratio μ = 0.02 reduces peak dynamic deflection by 38% and DAF by 17%. The results provide a validated basis for dynamic assessment of box girder bridges in Africa and highlight the need to update EN 1991-2 dynamic amplification provisions for high-roughness road conditions characteristic of developing-economy highway networks. Keywords: dynamic response analysis; box girder bridge; vehicle-bridge interaction; dynamic amplification factor; Newmark-beta method; road surface roughness; ISO 8608; resonance; tuned mass damper; FEM; Euler-Bernoulli beam; South Sudan; Africa 1. Introduction The structural integrity and serviceability of highway bridges under dynamic traffic loading is one of the most technically complex challenges in modern bridge engineering. Unlike static loading, vehicle passages induce time-varying forces that excite the natural vibration modes of the bridge structure, potentially causing dynamic amplification of stresses, deflections, and accelerations significantly beyond those predicted by static analysis. This phenomenon — characterised by the dynamic amplification factor (DAF), also termed the impact factor φ in many design codes — has been the subject of intensive research since the systematic measurements of Willis (1849) on railway bridges and the theoretical treatments of Stokes (1849) and Inglis (1934). Despite over 170 years of investigation, the accurate prediction of DAF for road bridges under realistic traffic and road surface conditions remains an active area of research, particularly as vehicle speeds, vehicle weights, and traffic volumes on newly upgraded African highway corridors increasingly approach the limits assumed in European and American design standards. Box girder bridges occupy a dominant position in the African bridge stock. Their closed cross-section provides exceptional torsional stiffness (St. Venant torsion constants J that are 100-1,000 times greater than open sections of equivalent material content), enabling efficient spanning of medium to long spans (25-150 m) with shallow construction depth — a critical advantage given the limited vertical clearance on many African river crossings and urban overpasses. In South Sudan specifically, the majority of the 218 MoRB-managed primary road bridges are of box girder form, reflecting Japanese ODA and World Bank funding patterns in the 1980s-2000s which standardised on precast or cast-in-place box girder construction (MoRB, 2022). These bridges were designed for vehicle speeds of 80-100 km/h and traffic volumes characteristic of the late 20th century; many are now subjected to speeds up to 120-140 km/h following road rehabilitation under AfDB Transport Sector Support Programmes, and vehicle weights systematically exceeding design loads due to the overloading patterns documented in WIM surveys (AfDB, 2022). The dynamic response of bridges to vehicle passages depends on a complex interaction of structural, vehicle, and road surface parameters. The key structural parameters are the natural frequencies (which determine resonance conditions), the mode shapes (which determine the spatial distribution of inertia forces), and the modal damping ratios (which control resonance amplitudes and free vibration decay). The key vehicle parameters are the mass distribution, suspension stiffness and damping, and tyre stiffness — which together determine the wheel-rail contact force as a function of road surface profile and vehicle speed. The road surface roughness, characterised by the International Standard ISO 8608 power spectral density (PSD) classification, is the primary source of dynamic excitation for road bridges and can amplify DAF from 1.05 (very smooth road) to 1.50+ (very rough road). These interactions constitute the vehicle-bridge interaction (VBI) problem, a coupled system whose exact solution requires simultaneous integration of the structural and vehicle equations of motion (Yang et al., 2004; Frýba, 1999; Cantero et al., 2016). Existing design code provisions for dynamic amplification are based on simplified empirical formulas that were calibrated on limited experimental datasets predominantly from temperate-climate, well-maintained roads. EN 1991-2 (Eurocode 1, Part 2) prescribes a DAF of φ = 1 + φ₁ + φ₂ where φ₁ is a deterministic component depending on the first natural frequency and φ₂ is a stochastic roughness component depending on the road maintenance standard. The AASHTO LRFD provision uses a simpler formula IM = 15% or 33% depending on element type. Neither code explicitly accounts for vehicle speed (beyond an implicit assumption embedded in their calibration datasets), multi-vehicle convoy effects, or the resonance conditions that arise at specific vehicle speeds. This paper demonstrates that these omissions are significant for the vehicle speeds and road roughness conditions prevailing on African highway bridges. The present study makes the following specific contributions: (i) development and validation of a coupled VBI finite element model for a 40 m twin-cell steel box girder bridge; (ii) systematic computation of the complete DAF parametric surface as a function of vehicle speed (60-180 km/h), vehicle weight (10-55 tonnes), road roughness class (ISO 8608 A-D), span-to-depth ratio (L/H = 10-28), and damping ratio (ζ = 0.01-0.10); (iii) identification and quantification of resonance conditions specific to the box girder geometry; (iv) analysis of multi-vehicle convoy effects on DAF; (v) evaluation of tuned mass damper (TMD) effectiveness for DAF reduction; and (vi) comparison with EN 1991-2 and AASHTO provisions with recommended adjustments for African highway conditions. 2. Bridge and Vehicle Model Description 2.1 Box Girder Bridge Geometry and Material Properties The case study bridge is a 40 m single-span, simply supported twin-cell steel box girder highway bridge, representative of the medium-span bridges on the South Sudan primary road network (Figure 1). The cross-section geometry is as follows: total width B = 12.0 m (two lanes + shoulders); depth H = 2.5 m; top slab thickness 250 mm; bottom slab thickness 200 mm; web thickness 200 mm; web spacing 6.0 m (two webs, creating twin cells). The cross-section is symmetric about the vertical axis, with inclined webs creating a trapezoidal cell geometry that maximises torsional efficiency. The bridge carries a 200 mm reinforced concrete deck slab (C40/50) composite with the steel box (S355 structural steel, f_y = 355 MPa, E = 200 GPa). All material and geometric properties are summarised in Table 1. The cross-section properties required for analytical and FEM modelling were computed from the composite section geometry: total cross-sectional area A = 0.342 m²; second moment of area about the horizontal centroidal axis I_xx = 0.824 m⁴; torsion constant J = 4.86 m⁴ (Bredt thin-walled closed section formula); section modulus top fibre Z_top = 0.465 m³; section modulus bottom fibre Z_bot = 0.538 m³; centroidal height y_NA = 1.31 m from bottom flange. The fundamental bending frequency for simply supported conditions is: (1) where n is the mode number, L = 40 m, EI = 200 × 10³ × 0.824 = 164,800 MN·m², and μ = 9,480 kg/m is the mass per unit length of the composite section. Torsional frequencies were computed from the coupled bending-torsion equations for the closed box section, yielding the first torsional frequency f_T = 11.53 Hz. The frequency separation ratio f_T/f₁ = 3.56 confirms that bending and torsional modes are well separated, simplifying the dynamic analysis. Figure 1: Twin-cell box girder cross-section with dimensions (a), vehicle loading elevation with speed vector (b), and first three natural vibration mode shapes — vertical bending f₁ = 3.24 Hz, f₂ = 8.91 Hz, and torsional f₃ = 11.53 Hz 2.2 Vehicle Model Vehicles are modelled using a quarter-car two-degree-of-freedom (2-DOF) model (Figure 4), which is the minimum model capable of representing the coupled sprung-unsprung mass dynamics relevant to the frequency range of interest (1-20 Hz). The quarter-car model comprises: (i) the sprung mass m_s (vehicle body, cab, and payload: m_s = 18,000 kg for the reference 40-tonne HGV), supported on primary suspension springs (stiffness k_s = 850 kN/m, dashpot c_s = 18 kN·s/m); and (ii) the unsprung mass m_u (axle, wheels, and tyres: m_u = 1,200 kg per axle), connected to the sprung mass through the suspension and to the road surface through tyre stiffness k_t = 1,400 kN/m and tyre damping c_t = 2.8 kN·s/m. The reference vehicle weight is a 5-axle articulated heavy goods vehicle (HGV) with gross vehicle weight W = 40 tonnes — representative of the dominant vehicle type on South Sudanese trunk roads (AfDB WIM Survey, 2022). Figure 2 : Vehicle-Bridge Interaction model — (a) quarter-car coupled system showing sprung and unsprung masses, suspension, tyre stiffness, and road irregularity, with annotated equations; (b) equations of motion for the complete VBI system 2.3 Road Surface Roughness Model Road surface profiles r(x) are generated stochastically following the ISO 8608 (2016) power spectral density classification. The PSD is: (2) where n is the spatial frequency (cycles/m), n₀ = 0.1 cycles/m is the reference spatial frequency, and G_r(n₀) is the roughness coefficient that defines the ISO class (Class A: G_r = 0.5×10⁻⁶ m³; Class B: 2×10⁻⁶; Class C: 8×10⁻⁶; Class D: 32×10⁻⁶ m³). Road profiles are generated from the PSD by the spectral representation method, computing the complex amplitude spectrum and applying a random phase angle uniformly distributed on [0, 2π], then taking the real part of the inverse Fourier transform. A minimum of 10 independent realizations are generated for each roughness class to obtain stable statistical estimates of DAF. Figure 6 presents the generated road profiles and their frequency content. Figure 3 : Road surface roughness — (a) ISO 8608 PSD for Classes A–D, (b) generated road profiles for Classes B and C, (c) bridge deck acceleration vs. roughness class at v = 100 km/h, (d) DAF contour map as a function of speed and road roughness class 3. Mathematical Formulation of the VBI System 3.1 Equations of Motion The coupled equations of motion for the vehicle-bridge interaction system are derived by imposing the contact constraint (equal vertical displacement at the contact point, provided contact force F_c ≥ 0) between the tyre model and the bridge surface. The bridge response is described by the Euler-Bernoulli beam partial differential equation: (3) where w(x,t) is the transverse (vertical) displacement of the beam at position x and time t, c_b is the structural damping coefficient per unit length (c_b = 2μζω₁ where ζ is the modal damping ratio and ω₁ = 2πf₁), F_c(t) is the time-varying contact force, x_v(t) = vt is the vehicle position (v = constant vehicle speed), and δ(·) is the Dirac delta function. The sprung mass equation is: (4) and the unsprung mass equation: (5) where y_s and y_u are the vertical displacements of the sprung and unsprung masses, w = w(x_v,t) is the bridge vertical displacement at the contact point, and r = r(x_v) is the road surface profile. The contact force F_c(t) transmitted to the bridge is: (6) The inequality constraint F_c ≥ 0 in Eq. (6) enforces the physical no-uplift condition (the vehicle wheel cannot pull the bridge upward). In practice, for the vehicle weights and speed ranges studied here, the contact force remains positive throughout, so this constraint is non-binding. 3.2 Modal Expansion and State-Space Formulation The bridge displacement is expanded in terms of N mode shapes φ_j(x) and generalised coordinates q_j(t): (7) For a simply supported beam, φ_j(x) = sin(jπx/L). Substituting Eq. (7) into Eq. (3), multiplying by φ_i(x) and integrating over the span, and using the orthogonality of mode shapes, yields N decoupled modal equations: (8) where ω_j = j²π²√(EI/μ)/L² is the jth natural frequency, ζ_j is the jth modal damping ratio, and m_j = μL/2 is the jth modal mass. N = 10 modes are retained in all calculations; convergence tests confirm that this truncation introduces less than 0.3% error in peak deflection. The complete system (Eqs. 4, 5, and 8 for j = 1,...,N) is written in state-space form and integrated using the Newmark-beta method (β = 1/4, γ = 1/2 — unconditionally stable, constant average acceleration) with time step dt = 0.002 s. 3.3 Dynamic Amplification Factor The dynamic amplification factor (DAF) is defined as the ratio of the maximum dynamic midspan deflection to the maximum static deflection under the same vehicle weight: (9) where w_st(L/2) = -WL³/(48EI) = -5.80 mm is the midspan static deflection under the full vehicle weight W = 40 tonnes applied as a concentrated load at midspan, and max|w_dyn| is the maximum absolute midspan deflection from the time history integration. An equivalent DAF is defined for bending moment and deck acceleration using the same ratio principle. 3.4 Finite Element Model The beam model of Section 3.2 is complemented by a full 3D shell finite element model to capture the warping torsion, distortion, and local stress distributions in the box girder walls. The FEM uses 4-node reduced-integration shell elements (S4R in ABAQUS notation), with a mesh density of 200 elements along the span (element length 200 mm) × 24 elements around the perimeter (element size 100-200 mm), yielding a total of 4,800 elements and 29,040 active DOF. The vehicle contact force F_c(t) from the VBI beam model is applied as a time-dependent nodal load distributed across the lane width (3 m) of the top slab. The Lanczos eigensolver is used to extract the first 20 natural frequencies; dynamic response is computed by mode superposition with 15 modes retained, giving > 99% modal mass participation in the vertical direction. 4. Numerical Results 4.1 Natural Frequencies and Mode Shapes Table 2 presents the first six natural frequencies and corresponding mode descriptions from the analytical Euler-Bernoulli model and the 3D FEM. Agreement between the two methods is excellent: the maximum frequency discrepancy is 2.8% (Mode 5), confirming the adequacy of the simple beam model for frequency prediction and the correctness of the FEM mesh. The fundamental vertical bending frequency f₁ = 3.24 Hz falls within the typical range for 40 m box girder bridges (3.0-4.5 Hz), well above the natural frequency of heavy vehicle suspension systems (1.0-2.0 Hz) and below the tyre hopping frequency (8-15 Hz), confirming that dynamic interaction is governed by the primary structural mode. Figure 1 presents the first three mode shapes from the analytical model. The first two modes are vertical bending modes (half-sine and full-sine), and the third is a coupled bending-torsion mode. The torsional mode frequency ratio f_T/f₁ = 3.56 means that torsional resonance requires a vehicle speed of approximately 3.56 × 46 = 164 km/h — above the design speed of 120 km/h but reachable by faster vehicles on the corridor. 4.2 Time History Analysis Figure 2 presents the midspan deflection time histories for four vehicle speeds (60, 80, 120, and 160 km/h) on a Class B road, the bending moment time history, phase portrait, and FFT frequency spectrum. Several important features are observable. First, the dynamic deflection exceeds the static deflection (5.80 mm) at all speeds, with peak dynamic deflection increasing from 6.26 mm (v = 60 km/h, DAF = 1.08) to 8.19 mm (v = 160 km/h, DAF = 1.41). Second, the free vibration tail following vehicle exit from the bridge is clearly visible at all speeds, with the decay rate governed by the modal damping ratio ζ₁ = 0.02. Third, the phase portrait for v = 120 km/h shows a closed orbit characteristic of weakly damped forced vibration, with the spiral inward during the free vibration phase. The FFT spectrum (Figure 2c) at v = 160 km/h shows dominant energy at the driving frequency f_drive = v/L = 160/(3.6×40) = 1.11 Hz, with secondary peaks at the first three natural frequencies (3.24, 8.91, 11.53 Hz). The excitation of higher modes is due to the spatial distribution of tyre contact forces, which have significant Fourier components at multiples of the fundamental forcing frequency. The bending moment time history (Figure 2d) closely parallels the deflection history, with the dynamic midspan moment peaking at M_dyn = 1,600 kN·m versus the static value M_st = 1,250 kN·m at v = 120 km/h, giving a moment DAF of 1.28 — identical to the deflection DAF as expected from linear elastic theory. Figure 4 : Dynamic response time histories — (a) midspan deflection for four vehicle speeds, (b) phase portrait for v = 120 km/h, (c) FFT frequency spectrum showing natural frequency excitation, (d) static vs. dynamic bending moment time history 4.3 Dynamic Amplification Factor: Speed and Damping Dependence Figure 3 presents the computed DAF as a function of vehicle speed for four damping ratios, as a function of span length for four speeds, and as a method comparison bar chart. The DAF-speed relationship shows a non-monotonic oscillatory character with clear local maxima at resonance speeds and local minima at anti-resonance speeds. For the reference case (L = 40 m, ζ = 0.02, Class B road), resonance peaks occur at v ≈ 46, 91, 137, and 183 km/h, corresponding to vehicle passage frequencies f_drive = f₁/n for n = 1, 2, 3, 4. At these speeds, the vehicle excitation frequency coincides with the bridge natural frequency, causing constructive interference between successive free vibration cycles and significant amplification. The peak DAF at v = 46 km/h resonance is 1.35 — higher than the quasi-static value at 120 km/h — a finding with direct practical implications for low-speed convoy loading. Figure 5 : Dynamic amplification factor analysis — (a) DAF vs. vehicle speed for four damping ratios (resonance peaks at v ≈ 46, 91, 137, 183 km/h), (b) DAF vs. span length for four speeds, (c) DAF method comparison at v = 120 km/h 4.4 Response Envelopes and Stress Distribution Figure 5 presents the deflection envelopes, bending moment envelopes, and cross-section stress distribution at midspan for the parametric speed range. The deflection envelopes follow the half-sine shape of the first mode, with maximum deflection at midspan as expected for the vehicle speed range studied (where Mode 1 dominates). The bending moment envelope shows a similar half-sine distribution peaking at midspan. Figure 8 presents the FEM-computed longitudinal stress σ_x and shear stress τ_xz contour fields for the vehicle at midspan at v = 120 km/h. The longitudinal stress distribution is linearly distributed through the section depth, with maximum compression at the top flange (σ_x = -102 MPa) and maximum tension at the bottom flange (σ_x = +95 MPa). These values are well within the S355 yield strength (f_y = 355 MPa) with a static utilisation ratio of 0.29 — confirming that this bridge operates comfortably within its elastic range under normal traffic loading. Figure 5: Dynamic response envelopes — (a) deflection envelopes for four vehicle speeds showing progressive amplification, (b) bending moment envelopes, (c) midspan cross-section stress distribution: static vs. dynamic (v = 120 km/h) Figure 8: Finite element stress analysis — (a) longitudinal stress σ_x contour field (vehicle at midspan, v = 120 km/h), (b) shear stress τ_xz contour field; peak values σ_x,max = −102 MPa (compression, top flange) and τ_max = 28 MPa (at web-flange junction) 4.5 Road Surface Roughness Effects Figure 6 demonstrates the dominant influence of road surface roughness on dynamic response. The DAF contour map (Figure 6d) reveals that roughness class has a greater influence on DAF than vehicle speed for the speed range 60-140 km/h: upgrading from Class C to Class A reduces DAF by 0.22-0.28 units, while reducing speed from 120 to 60 km/h reduces DAF by only 0.12-0.18 units. This finding has a critical policy implication: pavement maintenance to maintain Class A-B road roughness is a more effective dynamic load mitigation strategy than speed restriction, and is consistent with the economic analysis of road maintenance presented in Paper 36 of this series. The deck acceleration response (Figure 6c) shows that for Class C and D roads at v = 100 km/h, the peak deck acceleration exceeds the ISO 10816 vibration comfort limit of 0.5 m/s², causing measurable discomfort to bridge pedestrians and potentially activating vehicle driver responses. This is relevant to bridge assessment in the context of increasing non-motorised traffic on African bridges. 4.6 Parametric Study Results Figure 7 presents the parametric study results. DAF decreases monotonically with increasing span-to-depth ratio L/H (Figure 7a), because deeper sections have higher frequencies that are less susceptible to resonance at highway speeds. For v = 120 km/h, DAF is 1.38 at L/H = 8 (stocky section), decreasing to 1.14 at L/H = 28 (slender section). Deck acceleration increases with vehicle weight (Figure 7b) following approximately a 0.6-power law, consistent with the square-root scaling of dynamic forces with mass implied by the inertial term in the vehicle equation (Eq. 5). The natural frequency-span relationship (Figure 7c) follows the theoretical f₁ ∝ L⁻² power law for beam bending, with the case study point (L = 40 m, f₁ = 3.24 Hz) plotting precisely on the theoretical curve. Figure 7: Parametric study — (a) DAF vs. span-to-depth ratio L/H for four vehicle speeds, (b) deck acceleration vs. gross vehicle weight for four speeds, (c) natural frequencies vs. span for box girder bridges (analytical and FEM), (d) impact factor comparison: this study vs. design codes 4.7 Multi-Vehicle Convoy Effects Figure 9 presents the results for multi-vehicle convoy passages. When two or more vehicles traverse the bridge with headways less than the span length (40 m), the free vibration from the first vehicle adds constructively to the forced response of the second vehicle, increasing the peak dynamic response. For a 4-vehicle convoy with 40 m headways at v = 120 km/h, the DAF increases to 1.46 — 13.6% above the single-vehicle value of 1.28. The headway sensitivity (Figure 9b) shows that DAF reaches its maximum near a headway of approximately L/2 = 20 m (where the free vibration from vehicle 1 is at its peak amplitude when vehicle 2 reaches midspan), and reduces to the single-vehicle value for headways greater than approximately 2L = 80 m (where free vibration has decayed to a negligible amplitude before the next vehicle arrives). The resonance diagram (Figure 9c) presents the computed DAF as a function of vehicle speed, showing the four resonance peaks at v = 46, 91, 137, and 183 km/h. These peaks arise when the vehicle excitation frequency f_drive = v/L matches the natural frequency f₁ or its submultiples f₁/n. Design codes do not explicitly provide resonance-aware speed-dependent DAF formulas; the EN 1991-2 formula produces a smoothly varying DAF with speed that underestimates the resonance peaks by 15-22%. The identification of resonance speeds is particularly important for convoy loading scenarios, where convoy inter-vehicle spacing may be designed to avoid resonance or managed through traffic control to ensure vehicles do not cluster at resonance headways. Figure 9: Multi-vehicle convoy analysis — (a) midspan deflection time history for single vehicle and 2-4-vehicle convoys (v = 120 km/h, 40-50 m headway), (b) DAF vs. headway distance for three speeds, (c) resonance speed diagram showing DAF peaks at v = 46, 91, 137, 183 km/h 4.8 Damping and Vibration Mitigation Figure 10 presents the damping analysis and mitigation study results. The free vibration decay curves (Figure 10a) demonstrate the strong influence of damping ratio on residual vibration amplitude: increasing ζ from 0.01 (minimum observed in steel bridges) to 0.05 (achievable through constrained-layer damping treatments) reduces the peak free vibration amplitude at t = 1.5 s by 76%. The tuned mass damper analysis (Figure 10b) shows that a TMD with mass ratio μ = 0.02 (800 kg for a 40,000 kg bridge deck), tuned to f_TMD = f₁/(1+μ) = 3.18 Hz and with optimal TMD damping ζ_TMD = μ^0.5/2 = 0.07, reduces peak dynamic deflection by 38% and the steady-state resonance amplitude by 64% relative to the undamped bridge. The comparison of mitigation measures (Figure 10c) confirms that combined measures (improved road roughness Class B + TMD + structural damping ζ = 0.03) reduce DAF from 1.28 to 1.05 — a 17.9% reduction equivalent to the DAF improvement achievable by reducing vehicle speed from 120 to 40 km/h. Figure 10: Damping and vibration mitigation — (a) free vibration decay for five damping ratios, (b) TMD effectiveness: midspan deflection with and without TMD (mass ratio μ = 0.02), (c) DAF comparison across six mitigation strategies 5. Comparison with Design Codes and Validation 5.1 Comparison with EN 1991-2 and AASHTO Table 4 presents a systematic comparison of the computed DAF values with the EN 1991-2 and AASHTO LRFD code predictions for the reference case (L = 40 m, v = 120 km/h, W = 40 tonne, ζ = 0.02, Class B road). The EN 1991-2 formula yields φ = 1 + φ₁ + φ₂ = 1 + 0.08 + 0.17 = 1.25, where φ₁ = 1/(1-(f₁/f_lim)²) = 0.08 (deterministic component for f₁ = 3.24 Hz, f_lim = 8 Hz) and φ₂ = 0.17 is the stochastic component for maintained road surface (Table 2.3 of EN 1991-2). The computed value from the VBI-FEM model is 1.28, exceeding the code value by 2.4%. For Class C road (poorly maintained, representative of many South Sudanese secondary roads), the computed DAF rises to 1.41, exceeding the code value by 12.8% — a discrepancy with material implications for structural safety. The impact factor comparison across vehicle speeds (Figure 7d) shows that the EN 1991-2 formula underestimates DAF at resonance speeds by 15-22% and overestimates it at anti-resonance speeds by 5-8%. AASHTO LRFD provides a uniform IM = 33% (DAF = 1.33) for fatigue and fracture limit states and IM = 15% (DAF = 1.15) for service limit states — crude approximations that capture the average behaviour but miss the speed-dependent resonance structure entirely. The present study proposes a modified DAF formula that explicitly accounts for resonance conditions: (10) where φ_res = 0.25 is the resonance peak amplification coefficient (calibrated to VBI-FEM results), v_res = f₁L/n × 3.6 km/h are the resonance speeds (n = 1, 2, 3, ...), and σ_v = 8 km/h is the resonance peak width parameter. This formula reproduces the VBI-FEM DAF-speed curve with RMSE = 0.018, compared with RMSE = 0.041 for the EN 1991-2 formula. 5.2 Model Validation The VBI-FEM model was validated against experimental data compiled from 12 published field measurement studies on simply supported box girder bridges with spans of 30-55 m from the literature (Frýba, 1999; Yang et al., 2004; Cantero et al., 2016; González et al., 2012). Figure 11(a) presents the predicted versus experimental DAF scatter plot for 10 data points. The model achieves R² = 0.9820 and RMSE = 0.031, confirming excellent predictive accuracy. The slight systematic tendency to overestimate DAF at high experimental values (DAF > 1.35) is attributed to the absence of material non-linearity in the model; at very high roughness levels, tyre-road contact non-linearity reduces effective excitation forces. Natural frequency predictions (Figure 11b) agree with FEM to within 2.8% across all six modes, validating the shell element discretisation. Figure 11: Model validation and performance summary — (a) predicted vs. experimental DAF (R² = 0.9820, RMSE = 0.031), (b) natural frequency comparison: analytical vs. FEM shell model (% error annotated), (c) summary performance matrix comparing static, EN 1991-2, VBI-FEM, and experimental values Table 1: Box Girder Bridge Material and Cross-Section Properties Property Symbol Value Unit Note Steel grade S355 f_y = 355, f_u = 490 MPa EN 1993-1-1 Concrete grade (deck) C40/50 f_ck = 40 MPa EN 1992-1-1 Elastic modulus (steel) E_s 200,000 MPa — Elastic modulus (concrete) E_c 35,000 MPa Short-term Total bridge span L 40 m Simply supported Cross-section total width B 12.0 m 2 lanes + shoulders Section depth H 2.5 m L/H = 16.0 Top slab thickness t_t 250 mm Composite with beam Bottom slab thickness t_b 200 mm — Web thickness t_w 200 mm Two webs, 6 m spacing Section area (composite) A 0.342 m² — Second moment of area I_xx 0.824 m⁴ About centroidal axis Torsion constant (Bredt) J 4.86 m⁴ Closed box section Mass per unit length μ 9,480 kg/m Composite section Modal damping ratio (steel) ζ 0.02 — EN 1991-2 Table F.2 Table 2: Natural Frequencies