In modern precision engineering, the design and manufacture of hypoid bevel gears demand increasingly high levels of accuracy, flexibility, and integration across the product development cycle. This work presents a comprehensive, integrated methodology where the tooth flank geometry is uniformly represented by Non-Uniform Rational B-Spline (NURBS) surfaces. This digital model serves as the foundational element, seamlessly connecting the stages of functional design, Computerized Numerical Control (CNC) machining simulation, and contact analysis. The core principle, often termed “function-oriented active design,” involves determining a conjugate pinion flank based on a prescribed gear flank, a defined path of contact, and transmission requirements. By employing a NURBS representation, we overcome the limitations imposed by traditional machine-setting-based flank forms (like the modified roll or tilt methods), enabling more direct control over meshing performance and compatibility with advanced digital manufacturing and inspection systems.
The proposed methodology is built upon several interconnected pillars: the digitization of the known gear tooth flank into a NURBS model, the application of active design theory using this model, the simulation of CNC machining to generate the pinion flank data, the fitting of this data into its own NURBS representation, and finally, the Tooth Contact Analysis (TCA) performed directly on the NURBS surfaces. This holistic approach ensures that the final design is not only theoretically conjugate but also manufacturable and verifiable, creating a closed-loop digital thread for hypoid bevel gears.
1. Digital Modeling of Hypoid Gear Flanks via NURBS
The first and critical step is obtaining an accurate digital model of the gear tooth surface. For a hypoid bevel gear designed via traditional methods, its flank can be described by a set of discrete points. These points can be sourced from the theoretical surface equations, physical coordinate measurements, or—as adopted here—from a virtual machining simulation. Simulation offers a controlled and precise method for data generation.
The simulation model is based on a generic 5-axis CNC hypoid gear generator architecture. The essential coordinate systems are illustrated below. The machine coordinate system is denoted as \( O-xyz \), where the cutter head translates along \( x, y, z \). The workpiece (gear) coordinate system is \( O_2-x_2y_2z_2 \), rotating about the \( z_2 \)-axis (A-axis). The workpiece tilt is represented by the B-axis, parallel to the machine \( y \)-axis and passing through point \( O_2 \).
To create a general and efficient simulation, the gear blank is discretized into a family of concentric circles within the workpiece coordinate system:
$$ x_2^2 + y_2^2 = r_i^2 $$
$$ z_2 = Z_i \quad (i = 1, 2, 3, …, n) $$
The cutting tool (cutter head) is simplified into two conical surfaces representing the inner and outer blade edges. In the cutter coordinate system \( O_c-x_cy_cz_c \), the inner cone surface equation is:
$$ x_c^2 + y_c^2 = z_c^2 \cdot \tan^2(\alpha_{01}^{in}) $$
Similarly, the outer cone surface equation is:
$$ x_c^2 + y_c^2 = (l – z_c)^2 \cdot \tan^2(\alpha_{01}^{out}) $$
where \( l \) is the distance between the apexes of the inner and outer cones, and \( \alpha_{01}^{in}, \alpha_{01}^{out} \) are the blade pressure angles.
The simulation mathematical model is formed by solving the intersection equations between the family of circles (the blank) and either the inner or outer cone (the cutter). A combination of point-by-point scanning and the bisection method is used for robust solving. Key to this process is validating solution feasibility based on cutter axial dimensions and spatial relationships. The accuracy of this simulation method has been verified, with calculated flank point errors consistently below \( 10^{-8} \) mm compared to points from analytical surface equations.
With a cloud of discrete points \( \mathbf{P}_{ij} \) obtained from the simulation, the next step is to reconstruct a smooth, continuous NURBS surface. A NURBS surface of degree \( p \) in the \( u \) direction and degree \( q \) in the \( v \) direction is defined by:
$$ \mathbf{S}(u,v) = \frac{\sum_{i=0}^{n} \sum_{j=0}^{m} N_{i,p}(u) N_{j,q}(v) w_{ij} \mathbf{P}_{ij}}{\sum_{i=0}^{n} \sum_{j=0}^{m} N_{i,p}(u) N_{j,q}(v) w_{ij}} $$
where \( \mathbf{P}_{ij} \) are the control points, \( w_{ij} \) are their corresponding weights, and \( N_{i,p}(u) \) and \( N_{j,q}(v) \) are the non-rational B-spline basis functions defined on knot vectors \( U \) and \( V \). The process of finding the control polygon and knot vectors that interpolate or approximate the given point cloud is known as NURBS surface fitting. A critical aspect is handling boundary conditions to ensure high accuracy across the entire surface, including edges. Employing the “non-uniform” condition for knot vector generation and specific end-condition equations for control points has proven effective, yielding principal curvature calculation errors in the range of \( 10^{-8} \) to \( 10^{-14} \) mm. This highly accurate NURBS model \( \mathbf{S}^{(1)}(u,v) \) of the gear flank becomes the input for the active design process.
2. Active Design Principles with NURBS Surfaces
The function-oriented active design theory for hypoid bevel gears is applied here. We assume the gear flank \( \Sigma_1 \) (Surface 1) is known and represented by the NURBS surface \( \mathbf{S}^{(1)}(u,v) \). Also given are a prescribed path of contact on \( \Sigma_1 \) and the transmission function \( \phi_1(\phi_2) \), where \( \phi_1 \) and \( \phi_2 \) are the rotation angles of the gear and pinion, respectively. The goal is to determine the conjugate pinion flank \( \Sigma_2 \).
The prescribed contact path on \( \Sigma_1 \) is a space curve, which can be defined by a relationship between its surface parameters, e.g., \( v = f(u) \) or \( F(u,v)=0 \). For each point on this curve, corresponding to a specific angular position \( \phi_2 \), we must calculate first-order (position, unit normal) and second-order (curvature) information. The NURBS representation facilitates these calculations.
| Quantity | Calculation Method (Based on NURBS Surface \( \mathbf{S}(u,v) \)) |
|---|---|
| Position Vector | \( \mathbf{r} = \mathbf{S}(u,v) \) |
| Partial Derivatives | \( \mathbf{r}_u = \frac{\partial \mathbf{S}}{\partial u}, \quad \mathbf{r}_v = \frac{\partial \mathbf{S}}{\partial v} \) |
| Unit Normal Vector | \( \mathbf{n} = (\mathbf{r}_u \times \mathbf{r}_v) / \|\mathbf{r}_u \times \mathbf{r}_v\| \) |
| First Fundamental Form Coefficients | \( E = \mathbf{r}_u \cdot \mathbf{r}_u, \quad F = \mathbf{r}_u \cdot \mathbf{r}_v, \quad G = \mathbf{r}_v \cdot \mathbf{r}_v \) |
| Second Fundamental Form Coefficients | \( L = \mathbf{r}_{uu} \cdot \mathbf{n}, \quad M = \mathbf{r}_{uv} \cdot \mathbf{n}, \quad N = \mathbf{r}_{vv} \cdot \mathbf{n} \) |
| Principal Curvatures \( k_1, k_2 \) | Roots of \( k^2 – 2Hk + K = 0 \), where \( K = (LN-M^2)/(EG-F^2) \), \( H = (EN-2FM+GL)/(2(EG-F^2)) \) |
Let the tangent vector along the contact path on \( \Sigma_1 \) be \( \mathbf{t}_1^{(1)} \). Its direction relative to the surface parameterization is given by \( \lambda = du/dv \). The normal curvature \( k_n \) and geodesic torsion \( \tau_g \) of \( \Sigma_1 \) along the \( \mathbf{t}_1^{(1)} \) direction are calculated as:
$$ k_n = \frac{L + 2M\lambda + N\lambda^2}{E + 2F\lambda + G\lambda^2} $$
$$ \tau_g = \frac{(EM – FL)\lambda^2 + (EN – GL)\lambda + (FN – GM)}{(EG-F^2)(E + 2F\lambda + G\lambda^2)} $$
The fundamental equations of gearing and the conditions for continuous contact (up to second order) are then applied. These equations, involving the relative velocity \( \mathbf{v}^{(12)} \) at the contact point, the surface normals, and the calculated curvatures, allow us to solve for the principal curvatures \( k_1^{(2)}, k_2^{(2)} \) and principal directions \( \mathbf{e}_1^{(2)}, \mathbf{e}_2^{(2)} \) of the unknown pinion flank \( \Sigma_2 \) at the corresponding contact point. This process is repeated for a discrete set of points along the prescribed contact path.
3. Generating the Pinion Flank via CNC Machining Simulation
With the second-order information (curvatures) of \( \Sigma_2 \) known along the contact path, the next task is to define the complete surface. This is achieved by determining the machine tool settings (cutting locations) that would generate such a surface. For our generic 5-axis CNC machine, the motion of the cutter center \( (x_{oc}, y_{oc}, z_{oc}) \), the workpiece rotation \( \phi_2 \), and the workpiece tilt \( \beta \) are all functions of the pinion rotation angle \( \phi_2 \). These functions are typically represented as polynomial relationships:
$$
\begin{aligned}
x_{oc}(\phi_2) &= a_0 + a_1\phi_2 + a_2\phi_2^2 + … + a_5\phi_2^5 \\
y_{oc}(\phi_2) &= b_0 + b_1\phi_2 + b_2\phi_2^2 + … + b_5\phi_2^5 \\
z_{oc}(\phi_2) &= c_0 + c_1\phi_2 + c_2\phi_2^2 + … + c_5\phi_2^5 \\
\beta(\phi_2) &= d_0 + d_1\phi_2 + d_2\phi_2^2 + … + d_5\phi_2^5
\end{aligned}
$$
The coefficients \( a_i, b_i, c_i, d_i \) are determined based on the solved conjugate flank conditions at discrete points, often through a fitting process. Once these polynomial functions are established, we employ the same virtual machining simulation methodology described in Section 1, but now from the pinion’s perspective. The pinion blank is discretized, and its intersection with the moving cutter path—defined by \( x_{oc}(\phi_2), y_{oc}(\phi_2), z_{oc}(\phi_2), \beta(\phi_2) \)—is calculated. This yields a new cloud of discrete points representing the pinion flank \( \Sigma_2 \). This point cloud is then fitted to create its own NURBS surface model, \( \mathbf{S}^{(2)}(u,v) \). Thus, both members of the hypoid bevel gear pair are now represented in a consistent, digital NURBS format.
4. Tooth Contact Analysis (TCA) on NURBS Surfaces
To verify the meshing performance of the designed pair, TCA is performed directly on the NURBS models \( \mathbf{S}^{(1)} \) and \( \mathbf{S}^{(2)} \). Traditional TCA solves for points where the surfaces are in continuous tangency, requiring the coincidence of position vectors and unit normals under the constraint of the gear ratio. For NURBS surfaces, which may exhibit slight numerical unevenness from the fitting process, a robust optimization-based approach is advantageous.
The core problem is to find a set of parameters \( (u_1, v_1, \phi_1, u_2, v_2, \phi_2) \) that satisfies the meshing equations. We formulate this as an unconstrained minimization problem. For two candidate contact points \( \mathbf{r}_1 = \mathbf{S}^{(1)}(u_1, v_1) \) and \( \mathbf{r}_2 = \mathbf{S}^{(2)}(u_2, v_2) \), and their unit normals \( \mathbf{n}_1 \) and \( \mathbf{n}_2 \), the objective function \( F \) can be constructed as follows:
$$ \min F(\mathbf{X}) = w_1 \| \mathbf{r}_1(\phi_1) – \mathbf{r}_2(\phi_2) \|^2 + w_2 \| \mathbf{n}_1(\phi_1) – \mathbf{n}_2(\phi_2) \|^2 + w_3 (\phi_1 – \phi_1^{input}(\phi_2))^2 $$
where \( \mathbf{X} = [u_1, v_1, u_2, v_2]^T \), \( \phi_2 \) is the independent variable, \( \phi_1 \) is derived from the transmission function, and \( w_1, w_2, w_3 \) are weighting factors. An alternative, more stable formulation replaces the strict normal coincidence condition with the condition that the vector connecting the two points \( (\mathbf{r}_1 – \mathbf{r}_2) \) should be perpendicular to the surface normal. This can be measured by the angle \( \alpha \) between this vector and the normal:
$$ \min f(\mathbf{X}) = \| \mathbf{r}_1 – \mathbf{r}_2 \| + | 1 – \cos \alpha |, \quad \text{where} \quad \cos \alpha = \frac{(\mathbf{r}_1 – \mathbf{r}_2) \cdot \mathbf{n}_1}{\|\mathbf{r}_1 – \mathbf{r}_2\|} $$
A direct search optimization algorithm, such as the complex method, is well-suited for solving this problem. The output of the TCA includes the transmission error \( \Delta \phi_1(\phi_2) \), the path of contact on both flanks, and the contact ellipse dimensions at various load points, validating the design against the initial specifications.
5. Application Example and Results
The integrated methodology was applied to design a hypoid gear pair. The known gear flank was defined with a straight line of contact in the tooth lengthwise direction. Key design inputs were:
- Path of contact angle \( \gamma = 60^\circ \).
- Transmission function: \( \phi_1 = \phi_{10} – 638\phi_2 \pm 0.0002\phi_2^2 \).
- Target contact ellipse length: 12 mm.
The geometric parameters of the gear pair are summarized in the table below.
| Parameter | Gear (Σ₁) | Pinion (Σ₂) |
|---|---|---|
| Number of Teeth | 38 | 6 |
| Spiral Angle at Midface | 36.982° | 50.000° |
| Mean Cone Distance | 168.619 mm | 171.059 mm |
| Face Width | 38 mm | 38 mm |
Through the steps of gear flank digitization, active design calculation, and pinion flank simulation, NURBS models for both members were successfully generated. The pinion flank, represented by its control point grid, is shown conceptually. The TCA was performed on these NURBS models. The primary output, transmission error, showed deviations from the theoretical function on the order of \( 10^{-5} \) degrees, confirming the high accuracy of the design and the fidelity of the NURBS representation. Key contact point parameters from the TCA are listed below for three mesh positions.
| Mesh Position | Pinion Angle \( \phi_2 \) (deg) | Gear Angle \( \phi_1 \) (deg) | Transmission Error \( \Delta \phi_1 \) (arc-sec) | Contact Ellipse (Length x Width) mm |
|---|---|---|---|---|
| Heel | -0.5 | 319.105 | -0.018 | 11.95 x 3.21 |
| Mid | 0.0 | 0.000 | 0.005 | 12.02 x 3.18 |
| Toe | +0.5 | -319.095 | +0.015 | 11.98 x 3.22 |
6. Conclusion
This work establishes a complete, integrated framework for the design and analysis of hypoid bevel gears using parametric NURBS surface representations. By combining function-oriented active design theory with digital surface modeling and CNC machining simulation, we have created a methodology that is both powerful and practical. The use of NURBS serves as a universal digital model, bridging the gap between design intent and manufacturing reality. It facilitates direct data exchange for CNC tool path generation and Coordinate Measuring Machine (CMM) inspection path planning, forming the backbone of a modern digital manufacturing workflow for complex gears. The proposed TCA method, based on optimization, proves robust even with fitted surfaces. The results from the application example demonstrate the accuracy and viability of this approach, paving the way for more advanced, performance-driven design of hypoid bevel gears in an Industry 4.0 context.