Rolling-k Δ Analysis in Rolling Mechanics

• Rolling-k Δ Analysis is a rigorous method that quantifies elastic deflection of rolling elements by linking applied force to mill stiffness through micro-bump and viscoelastic effects.
• It integrates experimental data, machine learning models, and finite-element analysis to accurately predict deflection and rolling resistance under dynamic conditions.
• Empirical findings validate the method by demonstrating dynamic mean pressure updates and precise load-deflection curves critical for industrial rolling and transportation applications.

Rolling-k Delta Analysis (Rolling-k Δ Analysis) provides a rigorous, physically grounded framework for quantifying the elastic deflection (Δ) of rolling elements—such as work rolls in strip rolling or wheels in vehicle applications—under applied loads and rolling resistance. The methodology integrates the mechanical responses arising from real contact geometry, dynamical material properties, and the explicit origins of rolling friction torque, yielding an accurate, predictive relationship between applied force, structural compliance, and system deflection. The "k–Δ" formulation refers to the proportionality between the applied rolling force (or separating force), a generalized mill stiffness (k), and the resulting roll deflection Δ, capturing effects of rolling resistance not addressed in idealized textbook treatments.

1. Governing Equations of Rolling Motion

Rolling-k Δ analysis is rooted in the coupled translational and rotational equations of motion for a rolling cylinder, extended to include friction torque terms that originate from both surface microgeometry ("micro-bump collisions") and bulk elastic deformation with stress–strain history (Sasaki et al., 2021). For a cylinder of mass $m$, radius $a$, and moment of inertia $I$, rolling without slip, the governing equations are:

• Translational (Newton’s Second Law):

$m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$

• Rotational (about center):

$I\,\dot\omega = M_{\rm ext} + M_R$

Under pure rolling ($F_{\rm slip} = 0$), these simplify and explicitly expose the role of rolling resistance torque $M_R$ in retarding motion and dictating deceleration.

2. Decomposition of Rolling Friction Torque and k–Δ Formulation

The rolling resistance torque $M_R$ is decomposed into two physically distinct, additive components:

• Micro-bump Collisions:

Small collisions due to surface roughness (characterized by "bump" angle $\beta$ and energy loss fraction $c$) yield

$a$0

With deformation depth $a$1, this term links to $a$2 as

$a$3

• Elastic History Effect:

Asymmetry in loading/unloading during high-speed contact gives rise to a history-dependent torque,

$a$4

with $a$5 (speed-proportional at practical velocities).

These are consolidated as

$a$6

where $a$7 and $a$8 encapsulate micro-geometry and viscoelastic response. The key $a$9 relationship is then

$I$0

linking macroscopic rolling resistance directly to microstructural and dynamical variables.

3. Rolling-k Δ Relation and Solution for Deflection

The Rolling-k Δ formulation equates the maximal roll or wheel deflection Δ to the applied load $I$1 over an effective stiffness $I$2, integrating updated mean pressure and contact mechanics:

$I$3

Here, $I$4 is typically constructed as $I$5, where $I$6 is the dynamically determined mean pressure acting on the roll, continuously updated to account for changing material properties and reduction stages (Lotfinia et al., 2021).

A rolling element subject to such torques does not admit a constant-velocity steady state unless micro-collision losses vanish ($I$7). Instead, integrating the ODE for speed evolution yields

$I$8

Forecasting deceleration and eventual cessation, this dynamical law bridges the microscopic origins of resistance with observed macroscopic motion.

4. Methodological Workflow: Integration of Materials Models, Pressure, and Finite Element Deflection

A modern high-fidelity Rolling-k Δ workflow links experimental, machine learning, numerical, and structural modeling tools (Lotfinia et al., 2021):

1. Flow-Stress Model:
   • Room temperature uniaxial tensile tests inform an artificial neural network (ANN), trained to predict true stress $I$9 as a function of true strain $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$0 and strain rate $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$1. The dataset ("Stress316L") contains 15,858 records.
   • The ANN estimator $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$2 outputs $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$3, directly reflecting work hardening (e.g., strain-induced martensite in ASS 316L).
2. Finite-Difference Equilibrium Solver:
   • The slab method equilibrium equations are discretized (forward-difference) across the roll entry/exit arc, with local yield strengths $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$4 updated using the ANN.
   • The neutral angle $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$5 is found, and mean pressure $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$6 is numerically integrated for each rolling pass.
3. Finite-Element Analysis (FEA) of the Roll:
   • The work roll is modeled as a cantilevered beam, loaded across a localized span corresponding to the contact region, and solved as a linear elastic system.
   • Maximum deflection $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$7 is extracted (at the load region’s midspan), with numerical agreement to analytic solutions (<5% divergence) in early passes.
4. k–Δ Law Application:
   • The updated $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$8 at each pass enables immediate construction of $m\,\dot v = F_{\rm ext}^{(x)} - \dfrac{M_R}{a}$9 vs. $I\,\dot\omega = M_{\rm ext} + M_R$0 curves, or pass-by-pass prediction of $I\,\dot\omega = M_{\rm ext} + M_R$1 given rollback stiffness.

5. Empirical Results and Practical Insights

The table below summarizes mean roll pressure $I\,\dot\omega = M_{\rm ext} + M_R$2 and maximum deflection $I\,\dot\omega = M_{\rm ext} + M_R$3 for 1–7 rolling passes on ASS 316L:

Pass	Mean pressure $I\,\dot\omega = M_{\rm ext} + M_R$4 [MPa]	$I\,\dot\omega = M_{\rm ext} + M_R$5 [mm]
1	121	0.010
2	236	0.019
3	335	0.027
4	418	0.033
5	498	0.038
6	575	0.043
7	658	0.047

These results demonstrate the critical role of dynamically updating mean pressure and work hardening in accurate $I\,\dot\omega = M_{\rm ext} + M_R$6 prediction, especially as the reduction proceeds and deviations from constant-stress assumptions become pronounced (Lotfinia et al., 2021).

In railway systems, the $I\,\dot\omega = M_{\rm ext} + M_R$7 term explains the nearly linear rolling resistance increase with speed, matching observations in high-speed rail, while the micro-bump/viscoelastic formulation rationalizes the low rolling resistance in pneumatic tires (due to increased $I\,\dot\omega = M_{\rm ext} + M_R$8 but low $I\,\dot\omega = M_{\rm ext} + M_R$9 via material/inflation effects) (Sasaki et al., 2021).

6. Extensions and Future Directions

Prospective advancements in Rolling-k Δ analysis include:

• Incorporation of thermal–mechanical effects (by expanding ANN inputs to include $F_{\rm slip} = 0$0), relevant for warm/hot rolling processes.
• Extension of finite-difference models to encompass frictional heating and higher-speed strain-rate dependencies.
• Replacement of 1D beam models with full 3D FEA to recover through-thickness and edge effects (e.g., roll crowning, lateral inhomogeneity).
• Inclusion of a variable mill stiffness $F_{\rm slip} = 0$1 accounting for work-roll bending, backup roll compliance, and integrated mill drive elasticity, permitting system-level optimization and control.

These extensions are expected to further bridge experimental, computational, and analytic models, enhancing predictive accuracy and operational performance in both industrial rolling and transportation applications (Lotfinia et al., 2021, Sasaki et al., 2021).