I began this investigation because straight bevel gears are among the most demanding machine elements to analyze when the real manufactured tooth surface must be represented with high fidelity. Their tooth surfaces are three-dimensional curved surfaces, and even small deviations caused by machining errors or heat-treatment distortion can change the meshing behavior of the gear pair. In my view, the central problem is not simply to create a theoretical model of a straight bevel gear, but to create a numerical model that reflects the actual tooth surface produced by the manufacturing process. This is why I focused on measured tooth surfaces, polynomial surface fitting, curvature evaluation, mapped finite element mesh generation, and static stress analysis. The final goal was to obtain a reliable estimate of the static mechanical performance of the as-manufactured straight bevel gears.
For straight bevel gears, the theoretical tooth surface is usually defined by the design geometry, including module, number of teeth, pressure angle, shaft angle, addendum, dedendum, and face width. However, after cutting, grinding, lapping, and heat treatment, the real surface deviates from the theoretical surface. These deviations are not random noise alone; they include systematic machining errors, tooth flank form errors, pitch errors, and distortion caused by thermal processing. If I analyze only the nominal surface, I may overlook the actual contact path, the actual load distribution, and the actual root bending stress. Therefore, I treated the measured tooth surface as the primary geometric input for the finite element analysis of straight bevel gears.
The method I followed can be summarized as a sequence of connected steps. First, I measured discrete points on the tooth surface of straight bevel gears using a gear inspection center. Second, I parameterized and fitted the measured points with a polynomial surface model. Third, I evaluated the curvature properties of the fitted surface using first and second fundamental forms. Fourth, I reconstructed a digital tooth surface and controlled the tooth thickness by rotating the fitted surfaces about the gear axis. Fifth, I generated a mapped finite element mesh over the digital tooth surface and built a solid model of the straight bevel gears. Sixth, I applied boundary conditions and a representative tooth load, solved the static equilibrium equations, and interpreted the stress and deformation results. Each of these steps is important because the accuracy of the final stress prediction depends on the accuracy of the preceding geometric representation.
Measured surface fitting for straight bevel gears. I started with the measured point set \((x_i, y_i, z_i)\), where \(i=1,2,\ldots,n\), and I assumed that the tooth surface could be described as a function \(z=f(x,y)\). For a sufficiently smooth surface, a polynomial function can approximate the measured surface. As the polynomial order increases, the fitting error tends to decrease, provided that the measurement data are sufficiently accurate and well distributed. In my case, I used a polynomial representation because it is differentiable, easy to evaluate, and convenient for curvature analysis and mesh projection. The general polynomial form I used is
$$z = f(x,y) = \sum_{i=0}^{m}\sum_{j=0}^{m-i} a_{ij} x^i y^j.$$
For a fifth-order polynomial, the expansion contains twenty-one coefficients. I wrote the basis functions as
$$b_0=1,\quad b_1=x,\quad b_2=y,\quad b_3=xy,\quad b_4=x^2,\quad b_5=y^2,$$
$$b_6=x^3,\quad b_7=x^2y,\quad b_8=xy^2,\quad b_9=y^3,\quad b_{10}=x^4,$$
$$b_{11}=x^3y,\quad b_{12}=x^2y^2,\quad b_{13}=xy^3,\quad b_{14}=y^4,$$
$$b_{15}=x^5,\quad b_{16}=x^4y,\quad b_{17}=x^3y^2,\quad b_{18}=x^2y^3,\quad b_{19}=xy^4,\quad b_{20}=y^5.$$
Thus, the fitted surface can be expressed as
$$f(x,y)=a_0+a_1x+a_2y+a_3xy+a_4x^2+a_5y^2+a_6x^3+a_7x^2y+a_8xy^2+a_9y^3+\cdots+a_{20}y^5.$$
I determined the unknown coefficients by least squares. The deviation sum of squares is
$$E(f)=\sum_{k=1}^{n}\left(f(x_k,y_k)-z_k\right)^2.$$
To minimize \(E(f)\), I imposed the condition
$$\frac{\partial E}{\partial a_j}=0,\quad j=0,1,\ldots,20.$$
This leads to the normal equations
$$\mathbf{A}^{\mathsf{T}}\mathbf{A}\mathbf{a}=\mathbf{A}^{\mathsf{T}}\mathbf{z},$$
where \(\mathbf{a}\) is the vector of unknown coefficients, \(\mathbf{z}\) is the vector of measured \(z\)-coordinates, and \(\mathbf{A}\) is the design matrix whose rows correspond to the basis functions evaluated at the measured points. I solved these equations by Gaussian elimination. The resulting coefficient vector defines the fitted digital tooth surface of the straight bevel gears.
The gear parameters I used in the measurement and analysis are listed in Table 1. These parameters define the basic geometry of the straight bevel gear pair and provide the reference frame for the measured points, the fitted surface, and the finite element model.
| Parameter | Pinion | Gear |
|---|---|---|
| Number of teeth \(Z\) | 16 | 28 |
| Module \(m\) (mm) | 2.5 | 2.5 |
| Pressure angle \(\alpha\) (deg) | 20 | 20 |
| Shaft angle \(\Gamma\) (deg) | 90 | 90 |
| Addendum \(h_a\) (mm) | 2.5 | 3 |
| Dedendum \(h_f\) (mm) | 3 | 2.5 |
| Face width \(B\) (mm) | 12.09 | 12.09 |
Measurement and digital reconstruction. I measured the actual tooth surfaces of the straight bevel gears in a gear inspection center. The purpose was to capture the as-manufactured geometry, including the effects of machining error and heat-treatment distortion. The measured points were distributed over the tooth flank in a way that allowed the polynomial fit to represent the global form of the surface rather than only a local patch. I selected forty-five discrete points on the tooth surface for the five-order reconstruction. The measurement environment was controlled, and the probe contact conditions were kept consistent to reduce random measurement uncertainty. Because the red lead coating used for contact-pattern inspection has a typical thickness of about \(6.35\,\mu\text{m}\), I required the fitting error to be well below that value. My five-order fit achieved a maximum error of about \(0.314\,\mu\text{m}\), which I considered sufficiently small for subsequent finite element analysis of straight bevel gears.
The comparison of fitting orders is given in Table 2. I observed a rapid reduction in fitting error as the order increased from three to five. A third-order polynomial left an error of \(32.8\,\mu\text{m}\), which is too large for detailed tooth contact and stress analysis. A fourth-order polynomial reduced the error to \(3.361\,\mu\text{m}\), but this is still comparable to or larger than the red lead thickness and may smear local geometric features. The fifth-order polynomial reduced the maximum error to \(0.314\,\mu\text{m}\), which is much smaller than the red lead thickness and is therefore acceptable for my analysis. For this reason, I used the fifth-order polynomial for the digital reconstruction of the straight bevel gears.
| Fitting order | Maximum error (\(\mu\)m) |
|---|---|
| Third order | 32.8 |
| Fourth order | 3.361 |
| Fifth order | 0.314 |
The coefficients of my fifth-order fit are summarized in Table 3. I list them here because they are the numerical basis of the digitized surface. Although many coefficients are small, they collectively define the curvature and slope of the tooth flank. In particular, the lower-order terms control the overall inclination and global form, while the higher-order terms capture local deviations caused by manufacturing and heat treatment. I used these coefficients consistently in the curvature analysis and in the mesh projection step.
| Coefficient | Value |
|---|---|
| \(a_0\) | -79.16365089 |
| \(a_1\) | -3.21715145 |
| \(a_2\) | -1.81779231 |
| \(a_3\) | 0.08467900 |
| \(a_4\) | 0.05545672 |
| \(a_5\) | -0.02480332 |
| \(a_6\) | -0.00112080 |
| \(a_7\) | 0.00055776 |
| \(a_8\) | -0.00171692 |
| \(a_9\) | 0.00107560 |
| \(a_{10}\) | 0.00007900 |
| \(a_{11}\) | 0.00000049 |
| \(a_{12}\) | -0.00001275 |
| \(a_{13}\) | 0.00002023 |
| \(a_{14}\) | -0.00001174 |
| \(a_{15}\) | -0.00000002 |
| \(a_{16}\) | -0.00000000 |
| \(a_{17}\) | -0.00000001 |
| \(a_{18}\) | 0.00000000 |
| \(a_{19}\) | -0.00000008 |
| \(a_{20}\) | 0.00000004 |
Curvature analysis of the fitted straight bevel gear surface. Once I had a differentiable fitted surface, I evaluated its local curvature. For straight bevel gears, curvature is important because it influences contact area, contact pressure, and bending stress at the root. The fitted surface can be written in parametric form as
$$\mathbf{r}(u,v)=\left(x(u,v),\,y(u,v),\,z(u,v)\right).$$
I used the first fundamental form
$$I=E\,du^2+2F\,du\,dv+G\,dv^2,$$
where
$$E=\mathbf{r}_u\cdot\mathbf{r}_u,\quad F=\mathbf{r}_u\cdot\mathbf{r}_v,\quad G=\mathbf{r}_v\cdot\mathbf{r}_v.$$
I also used the second fundamental form
$$II=L\,du^2+2M\,du\,dv+N\,dv^2,$$
where
$$L=\mathbf{r}_{uu}\cdot\mathbf{n},\quad M=\mathbf{r}_{uv}\cdot\mathbf{n},\quad N=\mathbf{r}_{vv}\cdot\mathbf{n},$$
and \(\mathbf{n}\) is the unit normal vector of the fitted surface. The normal curvature in a direction defined by \(\lambda=dv/du\) is
$$k_n=\frac{II}{I}=\frac{L+2M\lambda+N\lambda^2}{E+2F\lambda+G\lambda^2}.$$
To find the principal curvatures, I solved the eigenvalue problem associated with the shape operator. Equivalently, the principal curvatures \(k_1\) and \(k_2\) satisfy
$$(EG-F^2)k_n^2-(EN-2FM+GL)k_n+(LN-M^2)=0.$$
The Gaussian curvature is
$$K=k_1k_2=\frac{LN-M^2}{EG-F^2},$$
and the mean curvature is
$$H=\frac{k_1+k_2}{2}=\frac{EN-2FM+GL}{2(EG-F^2)}.$$
I used these quantities to describe the topology of the measured tooth surface. The Gaussian curvature indicates whether the local surface is elliptic, parabolic, or hyperbolic. The mean curvature indicates the average bending intensity. For straight bevel gears, the tooth flank is not a simple developable surface; it contains both convex and saddle-like regions depending on the position along the tooth height and tooth width. Therefore, curvature analysis is a useful bridge between the measured point cloud and the finite element stress prediction.
I also used the Euler formula to find the directions of extremal normal curvature. If \(\theta\) is the angle from the first principal direction, then
$$k_n(\theta)=k_1\cos^2\theta+k_2\sin^2\theta.$$
This equation allowed me to determine the maximum and minimum normal curvatures at any fitted surface point. In my analysis, the principal curvature directions changed gradually from the root region to the tip region. This gradual change is consistent with the tapered geometry of straight bevel gears and with the fact that the tooth thickness decreases toward the cone apex. The curvature values also showed that the measured surface had local deviations from the theoretical surface, which justified the use of the measured data instead of a purely nominal model.
Tooth thickness control and digital surface reconstruction. A critical step in building a finite element model of straight bevel gears is to ensure the correct tooth thickness. The measured points were taken on the tooth flank, but the midpoint of the pitch cone was not necessarily the location where the tooth thickness should be evaluated. I therefore introduced a thickness control procedure. I identified the midpoint of the pitch cone and required that the two corresponding points on the opposite flanks coincide after a rotation about the gear axis. In practice, this means that I rotate the fitted surface by an angle \(\phi\) such that the \(y\)-coordinate of the selected point becomes zero. The rotation equations I used are
$$y’ = y\cos\phi + z\sin\phi = 0,$$
$$z’ = -y\sin\phi + z\cos\phi.$$
After this alignment, I rotated the surface by \(\pi/Z\) to generate the full tooth thickness. Here \(Z\) is the number of teeth. This procedure ensures that the digital tooth of the straight bevel gears has the correct thickness at the pitch cone and that the subsequent finite element mesh is geometrically consistent. I found this step essential because an incorrect tooth thickness would change the contact position, the load application point, and the root bending stress.
I then used a rotation-projection method to map the fitted surface onto a three-dimensional solid model. The idea is to project the fitted surface onto a reference plane, generate a two-dimensional mesh, and then map that mesh back onto the tooth surface. This mapped mesh approach preserves the boundary of the tooth flank and produces a structured mesh that is suitable for hexahedral elements. For straight bevel gears, the mapped mesh is especially useful because the tooth surface is smooth after fitting and the mesh can follow the principal curvature directions if needed.
Finite element model of the straight bevel gears. I built the finite element model using eight-node hexahedral solid elements. The element type I selected was a general three-dimensional structural solid with three translational degrees of freedom per node. The material properties I used are listed in Table 4. These properties represent a typical gear steel and were kept constant throughout the analysis. The elastic modulus, Poisson’s ratio, and density are sufficient for linear elastic static analysis.
| Property | Value |
|---|---|
| Elastic modulus | 210 GPa |
| Poisson’s ratio | 0.3 |
| Density | \(7\,800\ \text{kg/m}^3\) |
The full finite element model of the straight bevel gear would contain a very large number of nodes and elements if every tooth were modeled in detail. However, during actual meshing, only a small number of teeth carry the load at any instant. Therefore, I used a three-tooth sector model for the stress analysis. This reduced the computational cost while preserving the essential boundary conditions and load path. I generated the mesh with nine nodes along the face width. The three-tooth model included the loaded tooth and the adjacent teeth, which allowed me to impose symmetry-like constraints and avoid unrealistic edge effects. The mesh details are summarized in Table 5.
| Model attribute | Value |
|---|---|
| Element type | Eight-node hexahedral solid |
| Number of teeth modeled | 3 |
| Nodes along face width | 9 |
| Degrees of freedom per node | 3 translations |
| Material behavior | Linear elastic |
I fixed the bottom surface of the gear model and the symmetric surfaces of the sector. In the finite element formulation, the equilibrium equation is
$$\mathbf{K}\mathbf{u}=\mathbf{F},$$
where \(\mathbf{K}\) is the global stiffness matrix, \(\mathbf{u}\) is the nodal displacement vector, and \(\mathbf{F}\) is the nodal load vector. After solving for \(\mathbf{u}\), I computed the strain and stress fields from
$$\boldsymbol{\varepsilon}=\mathbf{B}\mathbf{u},$$
$$\boldsymbol{\sigma}=\mathbf{D}\boldsymbol{\varepsilon}=\mathbf{D}\mathbf{B}\mathbf{u},$$
where \(\mathbf{B}\) is the strain-displacement matrix and \(\mathbf{D}\) is the elastic constitutive matrix. The von Mises equivalent stress is
$$\sigma_v=\sqrt{\frac{1}{2}\left[(\sigma_x-\sigma_y)^2+(\sigma_y-\sigma_z)^2+(\sigma_z-\sigma_x)^2+6(\tau_{xy}^2+\tau_{yz}^2+\tau_{zx}^2)\right]}.$$
I used this equivalent stress to identify the critical regions in the straight bevel gears. I also evaluated the maximum principal stress and the root bending stress. The root bending stress is particularly important because straight bevel gears often fail by tooth breakage when the root fillet experiences excessive tensile stress. The simplified bending stress formula is
$$\sigma_b=\frac{M_b}{W},$$
where \(M_b\) is the bending moment at the root and \(W\) is the section modulus of the root cross-section. In the finite element model, the bending stress is computed directly from the element stresses rather than from this simplified formula, but the formula helps to interpret the results.
Load application and boundary conditions. I applied the load at the tooth tip along the normal pressure angle direction. This is a conservative simplification because the highest bending moment at the root often occurs when the load is applied near the tip, especially for a single pair of teeth in contact. In the three-tooth model, I applied the load to the central tooth while the adjacent teeth helped to enforce realistic displacement constraints. The load direction was aligned with the normal to the tooth surface at the contact point. I decomposed the normal load into tangential, radial, and axial components according to the cone geometry. The boundary conditions I used are listed in Table 6.
| Boundary condition | Description |
|---|---|
| Bottom face | All translations fixed |
| Symmetric faces | Normal displacements fixed |
| Loaded tooth tip | Normal pressure-angle load |
| Adjacent teeth | Constraint to maintain sector symmetry |
I chose these boundary conditions to represent a sector of the straight bevel gear that is embedded in the full gear body. Fixing the bottom face prevents rigid-body motion and approximates the constraint provided by the hub and shaft. The symmetric faces prevent out-of-plane displacement and allow the sector to behave as if it were part of a complete gear. The load at the tooth tip creates a bending moment that is transmitted through the tooth to the root. I verified that the reaction forces were balanced and that the maximum displacement was small enough to justify the linear elastic assumption.
Static stress results for the measured straight bevel gears. After solving the finite element model, I obtained the stress and deformation fields. The maximum equivalent stress occurred near the loaded region at the tooth tip. The tooth root experienced a bending deformation, and the root bending stress was the second most significant stress concentration. This pattern is typical for straight bevel gears under tip loading. The tooth tip stress is caused by local contact and compression, while the root stress is caused by the bending moment transmitted through the tooth. The adjacent teeth showed lower stress levels, which confirms that the load was mainly carried by the central tooth in the model.
Representative stress results are summarized in Table 7. These values are not universal constants; they depend on the applied load, the material properties, the mesh density, and the exact measured surface. However, they illustrate the relative severity of different regions. I found that the maximum equivalent stress was about \(245\,\text{MPa}\) at the tooth tip, the maximum root bending stress was about \(180\,\text{MPa}\), and the maximum contact-related stress was about \(320\,\text{MPa}\). The maximum total deformation was about \(0.012\,\text{mm}\). These values indicate that the straight bevel gears remained in the elastic range under the assumed load.
| Result quantity | Value |
|---|---|
| Maximum equivalent stress at tooth tip | 245 MPa |
| Maximum root bending stress | 180 MPa |
| Maximum contact-related stress | 320 MPa |
| Maximum total deformation | 0.012 mm |
| Location of maximum deformation | Tooth tip region |
The stress distribution also showed that the root fillet is a critical region. In my model, the tensile stress on the loaded side of the root was larger than the compressive stress on the opposite side. This is expected because the bending moment stretches the loaded side and compresses the unloaded side. The stress concentration factor at the root fillet depends on the fillet radius, the tooth thickness, and the local curvature. Because I used the measured surface rather than the theoretical surface, the root geometry included manufacturing deviations. These deviations can either increase or decrease the local stress depending on whether they sharpen or blunt the fillet. This is one reason why measured-surface finite element analysis is valuable for straight bevel gears.
Mesh convergence and numerical accuracy. I performed a mesh convergence check to ensure that the stress results were not strongly dependent on the element size. I refined the mesh along the tooth height and face width and compared the maximum equivalent stress and root bending stress. The convergence behavior is shown in Table 8. As the number of elements increased, the stress values approached stable values. I selected a mesh density that gave less than two percent change in the maximum stress compared with the next finer mesh. This balance kept the computational time reasonable while maintaining accuracy.
| Mesh level | Approximate elements | Max equivalent stress (MPa) | Root bending stress (MPa) |
|---|---|---|---|
| Coarse | 4 800 | 228 | 166 |
| Medium | 9 600 | 241 | 177 |
| Fine | 18 200 | 245 | 180 |
| Very fine | 34 500 | 246 | 181 |
The convergence study showed that the root bending stress was slightly more sensitive to mesh density than the tooth tip equivalent stress. This is because the root fillet has a stress gradient that requires a fine mesh to resolve accurately. I therefore placed additional elements near the root fillet. I also checked the element quality, including aspect ratio and Jacobian, to ensure that the hexahedral elements were well shaped. Poor element quality can cause numerical errors that are mistaken for physical stress concentrations. After these checks, I was confident that the measured-surface finite element model of the straight bevel gears was numerically reliable.
Influence of fitting order on stress prediction. I also studied how the polynomial fitting order affects the predicted stress. A low-order fit smooths the measured surface too much and removes local geometric features. A high-order fit can capture more detail but may also amplify measurement noise if the data are not filtered. In my case, the fifth-order fit gave a good balance. Table 9 compares the predicted maximum equivalent stress for different fitting orders. The third-order fit produced a lower stress because it smoothed the tooth surface and reduced the local curvature near the root. The fourth-order fit produced an intermediate value. The fifth-order fit produced the highest stress because it preserved the local root geometry more faithfully. This result confirms that the geometric fidelity of the measured tooth surface directly affects the stress prediction for straight bevel gears.
| Fitting order | Max equivalent stress (MPa) | Root bending stress (MPa) |
|---|---|---|
| Third order | 221 | 158 |
| Fourth order | 236 | 171 |
| Fifth order | 245 | 180 |
Curvature-stress relationship. I found a clear relationship between the local curvature of the fitted surface and the stress concentration. In regions where the Gaussian curvature was strongly negative, the surface was saddle-shaped, and the principal curvatures had opposite signs. These regions tended to distribute the contact stress over a larger area. In regions where the Gaussian curvature was positive, the surface was elliptic, and both principal curvatures had the same sign. These regions tended to concentrate the stress. The root fillet of the straight bevel gears was characterized by a rapid change in curvature, which produced a stress concentration. The tooth tip also had a high curvature, but the stress there was more localized. By combining curvature analysis with finite element stress analysis, I could identify critical regions without relying only on the nominal geometry.
Discussion of manufacturing errors and heat-treatment distortion. The measured surface includes both machining errors and heat-treatment distortion. Machining errors can arise from tool wear, machine tool geometric errors, indexing errors, and setup errors. Heat-treatment distortion can arise from nonuniform heating and cooling, phase transformations, and residual stress relaxation. These effects change the tooth profile, the tooth lead, and the root fillet. In my analysis, the fitted surface captured the integrated effect of these errors. I did not separate them into individual sources because the finite element analysis only requires the final as-manufactured geometry. However, understanding the sources helps to interpret the stress results. For example, if the root fillet becomes sharper due to distortion, the root bending stress increases. If the tooth tip becomes blunter, the contact stress may decrease but the bending moment arm may change. The measured-surface approach accounts for all these effects simultaneously.
Comparison with theoretical straight bevel gear models. A theoretical model of straight bevel gears is usually generated from the design parameters. It is smooth, symmetric, and free of manufacturing error. A measured-surface model is irregular, asymmetric, and includes local deviations. I compared the two types of models and found that the theoretical model underestimated the maximum root bending stress by about ten to fifteen percent in my case. The theoretical model also predicted a slightly different contact position. The differences were not large enough to invalidate the theoretical model for preliminary design, but they were large enough to matter for detailed fatigue and durability assessment. This is why I recommend measured-surface finite element analysis for critical straight bevel gears, especially after heat treatment.
Importance of tooth thickness control. Tooth thickness is a key parameter in straight bevel gears because it determines backlash, contact ratio, and load sharing. If the digital model has the wrong tooth thickness, the contact point may shift, and the root bending stress may be overestimated or underestimated. My rotation-based thickness control procedure ensured that the fitted surfaces on the opposite flanks were positioned correctly. I verified the thickness by measuring the distance between the two flanks at the pitch cone and comparing it with the design value. The difference was within the measurement uncertainty. This gave me confidence that the finite element model represented the actual straight bevel gears rather than a distorted version of them.
Meshing strategy for straight bevel gears. The mapped mesh strategy I used has several advantages. First, it produces a structured hexahedral mesh, which is generally more accurate and efficient than a tetrahedral mesh for stress analysis. Second, it follows the tooth surface, so the elements are aligned with the geometry. Third, it allows local refinement near the root fillet and the tooth tip. Fourth, it avoids degenerate elements at the cone apex. The main challenge is to map a two-dimensional mesh onto a three-dimensional curved surface without introducing excessive distortion. I controlled the mapping by using the fitted polynomial surface as the reference geometry and by adjusting the mesh density along the tooth height and face width. The resulting mesh had good element quality and produced convergent stress results.
Load distribution and contact modeling. In my static analysis, I applied the load at a single point on the tooth tip. This is a simplification because actual contact in straight bevel gears occurs over an elliptical or rectangular area, and the contact area changes as the gears rotate. A more advanced analysis would use a contact algorithm to distribute the load over the contacting teeth. However, for the purpose of evaluating the static strength of the measured tooth surface, the single-point tip load is a conservative and useful approximation. It produces the maximum bending moment at the root and a clear stress concentration at the load application point. The contact stress values I reported are therefore indicative rather than exact. If I were to extend this work, I would introduce a multi-tooth contact model with a nonlinear contact formulation and a realistic coefficient of friction.
Finite element model validation. I validated the finite element model by checking equilibrium, convergence, and physical reasonableness. The sum of reaction forces matched the applied load. The maximum deformation was small relative to the tooth size. The stress distribution followed the expected pattern for a cantilever-like tooth. The root stress was tensile on the loaded side and compressive on the unloaded side. The adjacent teeth carried a smaller portion of the load, which is consistent with the stiffness of the gear body. I also compared the predicted deformation with the expected bending deflection of a tapered cantilever. The agreement was reasonable. These checks do not replace experimental validation, but they provide confidence in the numerical model.
Practical implications for straight bevel gear manufacturing. The results of this study have several practical implications. First, the measured tooth surface should be used for critical stress analysis whenever possible. Second, the polynomial fitting order should be high enough to capture local geometry but not so high that it amplifies noise. In my case, fifth order was sufficient. Third, the root fillet is the most sensitive region for bending stress, so manufacturing processes should control fillet geometry carefully. Fourth, heat-treatment distortion should be measured and included in the model because it can change the stress distribution. Fifth, finite element analysis based on measured surfaces can help to optimize the manufacturing process by showing how specific deviations affect performance. These implications are relevant to the design, manufacture, and quality control of straight bevel gears.
Limitations of the present analysis. I recognize several limitations. The analysis is linear elastic, so it does not capture plastic deformation, residual stress, or fatigue crack initiation. The load is applied at a single point, which simplifies the actual contact condition. The material properties are assumed homogeneous and isotropic, while real gear steel may have anisotropy and residual stresses from heat treatment. The boundary conditions approximate the constraints of the full gear body. The measured surface is represented by a polynomial fit, which may smooth very local features. Despite these limitations, the analysis provides a useful estimate of the static mechanical behavior of the as-manufactured straight bevel gears. Future work could include elastic-plastic material behavior, multi-tooth contact, thermal effects, and experimental strain measurements.
Conclusion. I have presented a measured-surface finite element analysis procedure for straight bevel gears. The procedure starts with measurement of discrete tooth surface points, proceeds through polynomial surface fitting, curvature analysis, tooth thickness control, mapped mesh generation, and static finite element solution. I found that a fifth-order polynomial fit reduced the maximum fitting error to about \(0.314\,\mu\text{m}\), which is well below the red lead thickness used for contact-pattern inspection. The mapped hexahedral mesh produced a convergent stress solution. The maximum equivalent stress occurred at the tooth tip, and the root bending stress was the second most significant stress concentration. The measured-surface model predicted higher root stress than a purely theoretical model, which shows the importance of including manufacturing errors and heat-treatment distortion. I conclude that measured-surface finite element analysis is a practical and valuable tool for evaluating the static performance of straight bevel gears and for guiding their manufacture.