On the well-posedness and efficient approximation for the classical Melan equation in suspension bridges

1 Introduction

• Despite computer-based finite-displacement methods, deflection theory remains a main analytical method for preliminary design and practical calculations, helping designers assess parameters and validate complex simulations.Source: §1 · p. 2
• The deflection-dependent nonlocal term makes Melan’s equation difficult to analyze, while existing engineering approximations generally lack rigorous analysis, error estimates, and convergence proofs.The relationship between approximate and true solutions is rarely addressed.Source: §1 · pp. 3–4
• When live load is small relative to dead load, the less-stiff model approximates H + h(w) by H; numerical comparisons in suggested its use, but gave no rigorous model relationship or quantitative error estimate.Source: §1 · p. 4
• Gazzola et al. established existence results and uniqueness within a class of small solutions under stronger smallness assumptions, but their iterative scheme appeared unstable for some practical bridge parameters.Source: §1 · p. 5
• The open questions listed in include conditions for well-posedness, constructive approximation, convergence of iterative schemes, and algorithms that handle stable and unstable cases.Source: §1 · p. 5
• A recent study derived an explicit expression for the unique analytical solution of the simplified less-stiff model.Source: §1 · p. 5
• Earlier parameter restrictions limited results to short-span, high-stiffness bridges, whereas practical medium- to long-span bridges with flexible girders may exhibit inflection points under live loads.The introduction motivates extending well-posedness and approximation results to those engineering settings.Source: §1 · pp. 5–6
• Under an explicit downward-live-load condition, the classical Melan problem has a unique solution with nonnegative integral, without a separate smallness assumption on span length L.The corresponding deflection curve is uniquely determined and may have inflection points.Source: §1 · pp. 6–7
• The simplified-model solution seeds an iteration with geometric convergence and explicit error estimates; it is practically acceptable as the first iterate, while the second is considerably more accurate.Numerical examples suggest the method may remain effective even for very heavy live loads beyond the explicit condition.Source: §1 · pp. 6–7
• The error estimates and numerical results establish a quantitative connection between the simplified linear nonlocal model and the original nonlinear nonlocal model.They support the simplified model as an acceptable approximation to the classical Melan model.Source: §1 · p. 7
• Using integral properties of the simplified model, the authors derive an a priori estimate and prove contraction on a closed ball, yielding global uniqueness in the physically meaningful class.The operator is built from the simplified model’s analytical solution and associated integral properties.Source: §1 · p. 8
• The operator accommodates integrable, potentially piecewise distributed loads in L1[0, L], and actual-bridge examples show the explicit condition holds for realistic small and moderate downward live loads.The examples also suggest effectiveness for very heavy live loads beyond that condition.Source: §1 · pp. 7–8

Section 3

• A priori estimates support existence and uniqueness for the original nonlinear model under a concise condition on the live load.These results concern problem (1.4),(1.2).Source: §1 · p. 8

Section 4

• A fixed-point iterative method with error estimates is proposed, and its applicability and computational efficiency are demonstrated on two actual bridges.Source: §1 · p. 8
• Calculations quantify the nonlinear nonlocal term’s influence and the difference between the two models, while explaining several engineering observations.Source: §1 · p. 8

Section 5

• The final section summarizes the conclusions and proposes related open problems.Source: §1 · p. 8

2 The analytical solution for the “less stiff” model (1.5), (1.2) and global properties of the solution

• The simplified model has a unique solution represented explicitly using its Green function, providing an exact benchmark for numerical methods.The representation is derived from the Green functions of two second-order operators.Source: §2 · pp. 9–11
• The explicit solution gives a definitive criterion for evaluating numerical methods developed for the simplified model.Source: §2 · p. 11
• The solution’s integral admits an a priori estimate that is nonincreasing in parameter a; for φ ≥ 0, its integral is nonnegative without parameter restrictions.The section distinguishes nonnegative integral from pointwise positivity, which requires a parameter condition.Source: §2 · pp. 11–12
• The solution also satisfies an L∞ estimate proportional to ∥φ∥1, with a bound D(a) that is nonincreasing in a.Source: §2 · pp. 12–13
• Nonnegative φ alone does not ensure w′′ ≤ 0, so the simplified model’s deflection curve need not be free of inflection points.A sufficient parameter condition for w′′(x) ≤ 0 throughout the interval yields a curve with no inflection points.Source: §2 · p. 14

3 Existence and uniqueness of solutions to the classical Melan equation (1.4) with (1.2)

• The nonlinear boundary-value problem is equivalent to a fixed-point equation for the operator A, linking its solutions to fixed points in L1[0, L].The correspondence works in both directions: a solution gives a fixed point, and a fixed point determines a solution.Source: Proof. Let · p. 16
• Under hypothesis (H), the associated operator has a unique fixed point, yielding a solution of the classical Melan boundary-value problem.The theorem also provides estimates for the solution and its second derivative.Source: §3 · p. 14, Proof. Let · pp. 16–17

3.1. If ew ∈W 4,1[0, L] is any solution of the BVP (3.1),(1.2) satisfying Z L

• Among solutions with nonnegative integral, the solution is unique whenever it exists, without a smallness restriction on the solution.This uniqueness class is physically relevant for usual downward traffic loads, and the result can also apply to some sign-changing loads if the solution remains in the class.Source: §3.1 · pp. 18–19
• For p ≥0, hypothesis (H) ensures the constructed solution has nonnegative integral and is therefore the unique solution in that class.The class captures the physically meaningful solution for usual downward traffic loads.Source: §3.1 · pp. 19–20
• Strict global uniqueness is not established because large negative-integral solutions outside the fixed-point ball may exist.Such solutions could reach or cross the first L2, causing Green-function sign changes or resonance; their existence remains a separate mathematical question.Source: §3.1 · p. 21
• The results improve existence and uniqueness under explicit condition (H), without separate smallness restrictions on main span length or additional bridge-parameter restrictions.Earlier results cited in the passage imposed restrictions on span length and other structural parameters.Source: §3.1 · p. 21
• After customary lane-load reduction and vehicle-spacing effects, condition (3.16) holds for practical bridges under realistic small and moderate downward live loads.The stated loads are representative of ordinary service traffic.Source: §3.1 · p. 21
• Under those practical load conditions, the corresponding deflection curve is uniquely determined.For the original classical Melan problem, condition (H) is equivalent to a condition on the original bridge parameters and actual live load.Source: §3.1 · p. 21

• Under hypothesis (H), the iterates converge geometrically to the fixed-point solution, with an error bound that decreases by powers of q.When p ≥ 0, the limit is the unique solution with nonnegative integral.Source: §4 · p. 22
• For the 460 m bridge, hypothesis (H) requires ||P||_1 < 5744.859 kN; computations consider a centered 10 m coach (10 kN/m) and 230 m freight train (20 kN/m).Source: §4 · p. 23
• Although w1 is expressed piecewise, it is twice continuously differentiable and symmetric about the bridge center x = 230.Source: §4 · p. 25
• On the 460 m bridge, mean absolute errors of w1 and w2 do not exceed 0.0003033508 and 4.72693 × 10^-7 m, respectively.The corresponding global error bounds are 0.1395414 for w1 and 0.0002174388 for w2.Source: §4 · p. 28
• The global contraction constant is conservative: the observed convergence ratio is nearly an order of magnitude smaller.Source: §4 · pp. 28–29

4 Iterative method for classical Melan equation and its application to actual bridges

• For w2, q = 0.8007 accompanies a mean absolute error bound of 0.08705874 m; the qloc error estimate is about 1.85 × 10−2 m and is not a rigorous upper bound.The qloc estimate is small compared with w2’s displacement amplitude, indicating that w2 provides a reasonable engineering-level approximation.Source: §4 · pp. 32–33
• In both centered-load cases, displacement curves are symmetric about midspan, with maximum downward deflection at midspan and upward extrema and inflection points symmetrically placed.The second iterate has smaller amplitude than the first in both cases.Source: §4 · p. 33
• Across both centered loads, w1 tends to overestimate the original model’s displacement; relative mean errors are 1.34% and 2.08 × 10−3% for the narrow load, and 6.33% and 4.12% for the wider load.For the narrow load, the percentages use the conservative theoretical global estimate; for the wider load, they use the observed local contraction factor. In both cases the errors are listed for w1 and w2, respectively, relative to the maximum downward displacement amplitude max w2.Source: §4 · pp. 33–34
• For the wider load, global bounds are rigorous but conservative, while the observed local contraction factor is about one order of magnitude smaller and gives sharper, nonrigorous estimates.The gap reflects the difference between worst-case theoretical bounds and the iteration’s observed numerical convergence.Source: §4 · p. 34
• The Golden Gate Bridge example has a 1280 m main span and 145 m sag, and its live load must satisfy ∥P∥1 < 2.682132 × 10−5 EI under the stated condition.The section begins considering three live-load cases, starting with 30 kN/m.Source: §4 · pp. 34–35

Case 2: central-half-span loading

• The 30 kN/m left-half-span case has a total live load of 19,200 kN and satisfies Condition (H); the central-half-span case preserves the intensity and loaded length.The central-half-span case places the load symmetrically about midspan, leaving the total load and global contraction condition unchanged.Source: §4 · pp. 35, 39

Case 1: left-half-span loading

• The global-contraction estimate bounds the mean absolute errors of w1 and w2 over the main span by 152.56413.The estimate uses the global contraction constant q = 0.6616083.Source: §4 · p. 38
• The first two iterates have maxima of 1.95174 m for w1 and 1.87705 m for w2, and minima of −1.42751 m and −1.35406 m, respectively.The figure plots w1 and w2 for the 30 kN/m left-half-span load.Source: Figure 4 · p. 37, §4 · p. 38
• With qloc, ∥w−w1∥1 ≤ 61.88862 + 32.10631 = 93.99493 and ∥w−w2∥1 ≤ 32.10631; these are not rigorous upper bounds in the global-contraction sense.Source: §4 · p. 39

30 kN/m central-half-span load

• 84.61501 is the observed-local-factor estimate for the error of w1, versus the global-contraction bound of 176.54895; the local estimate is not a rigorous upper bound.The observed factor yields smaller a posteriori numerical estimates, indicating that the global estimate is conservative in this case.Source: §4 · pp. 43–44

Case 3

• 38,400 kN and 3.549041 × 10^-5 exceed the respective admissible bounds of 29,020.19 kN and 2.682132 × 10^-5, so the 60 kN/m load does not satisfy hypothesis (H).Existence, uniqueness, and rigorous global error estimates under (H) are therefore not guaranteed for this benchmark load.Source: §4 · p. 44
• 3.90348 m for w1 and 3.60477 m for w2 are the maximum downward displacements, near the main-span quarter point and reasonably close to ’s 3.81 m reference.The comparison uses the same 60 kN/m half-span live load.Source: §4 · p. 47
• The two observed successive-difference ratios are much smaller than 1, indicating contraction in the 60 kN/m iteration, although no global contraction is theoretically established.The sufficient condition for global contraction is not satisfied in this case.Source: §4 · p. 47

Remark 4.3

• Under the two left-half-span Golden Gate loading cases, maximum downward deflection lies slightly right of the main-span quarter point, maximum upward deflection near the unloaded-side three-quarter point, and an inflection point between them, matching.Source: §4 · pp. 47–48
• For the Golden Gate Bridge load cases, the estimates are about 3.91% for w1 and 1.34% for w2 in Case 1, and 5.36% and 3.26% in Case 2; the second iterate improves further.The first iterate is an acceptable first approximation in all three cases, and Case 3 agrees well with the engineering reference.Source: §4 · p. 48

Remark 4.4

• The left-half-span 30 kN/m load over half the main span produces greater maximum downward deflection than the centered half-span load: max w2 ≈1.87705 m in Case 1 versus 1.23413 m in Case 2.Source: §4 · p. 48

Remark 4.5

• In Cases 1 and 2, global contraction constants are both about 0.6616, while observed local factors are about an order of magnitude smaller, indicating faster convergence than the worst-case estimate.Source: §4 · pp. 48–49
• For the very heavy 60 kN/m left-half-span Case 3 load, condition (H) is not satisfied and the global factor exceeds 1; nevertheless, qloc ≈ 8.8417 × 10^-2, and the computed iteration exhibits fast local convergence.Source: §4 · p. 49

5 Conclusions and open problems

• A priori estimates from the simplified model extend classical engineering approximation methods with rigorous analysis and go beyond Wang’s short-span, high-stiffness parameter range.The authors also contrast their analysis with earlier methods that generally lacked convergence proofs.Source: §5 · pp. 49–50
• The paper reports the first rigorous well-posedness proof for the classical Melan equation in the engineering sense, under an explicit condition that uniquely determines the deflection curve.The result concerns solutions with nonnegative integral under downward live loads.Source: §5 · p. 49
• The proposed iterative method uses the simplified “less stiff” model as its first iterate and has a convergence proof for practical bridges of common span ranges under ordinary live loads.The paper describes it as the first approximation algorithm with a convergence proof for such practical bridges.Source: §5 · p. 49
• The work establishes a theoretical and computational connection between the classical and simplified models and analyzes the nonlinear nonlocal term’s influence.The numerical examples show the simplified model can approximate the original classical model effectively.Source: §5 · p. 50
• The first iterate gives a practically acceptable approximation but tends to overestimate displacement, while the second iterate substantially improves accuracy.The analysis provides quantitative error estimates for these numerical observations.Source: §5 · p. 50
• Examples on two actual bridges across different spans and live-load cases show applicability and efficiency, with computed load effects consistent with engineering experience.The examples examine live-load magnitude, loading interval, and loading position.Source: §5 · p. 50
• Whether well-posedness and convergence can be proved for loads beyond condition (H) remains open, although the method may still be effective for very heavy loads.The paper poses existence, uniqueness, and convergence under such loads as questions for further study.Source: §5 · p. 50