My doctoral research was driven by the need to enhance the efficiency and quality of spiral bevel gear production using modern CNC machine tools. Traditional methods for selecting cutting parameters, often based on experience or data for older mechanical machines, fail to leverage the full potential of CNC technology. Furthermore, a lack of fundamental research into the physical cutting mechanics of gear cutting, such as dynamic cutting forces and their relationship to surface finish, hinders optimal machine design and process control. This work bridges that gap by developing models for cutting forces, optimizing cutting parameters, and establishing a link between process variables and surface quality in CNC spiral bevel gear cutting.
1. Dynamic Cutting Force Modeling for Form Milling of the Gear
The initial phase of my research focused on the form milling process, typically used for roughing the larger gear (gear) member. Establishing an accurate dynamic cutting force model is foundational for machine design, power verification, and subsequent process optimization.
1.1 Geometric and Kinematic Foundations
The model is built upon the precise geometry of the form gear cutting process. The cutter head, equipped with multiple indexed inserts, and the gear blank are positioned according to calculated machine settings (Machine Root Angle $\delta_M$, Horizontal Center $H$, Vertical Center $V$). The fundamental relationship defining the cutting engagement is the intersection between the cutter surface and the gear blank’s front cone in the machine coordinate system $S_m$:
$$ \mathbf{L}_M(u, \theta, v, \psi, z_j) = \mathbf{r}_m^{(a)}(v, \psi, z_j) – \mathbf{r}_m^{(t)}(u, \theta) = 0 $$
where $(u, \theta)$ are parameters for a point on the cutter insert surface, $(v, \psi)$ are parameters for a point on the gear front cone, and $z_j$ is the infeed position of the workpiece slide.
1.2 Calculation of Instantaneous Undeformed Chip Geometry
At any given time $t$, the engagement condition for insert $j$ is determined. The instantaneous undeformed chip width $b(t)$ and thickness $h(t)$ for both inner and outer inserts are derived from the geometric solution of the engagement equation and the feed motion. For an outer insert, the chip width is:
$$ b_{ew}(t) = u_{ej}(\theta_j(t), z_j(t)) – u_{e0} $$
and the chip thickness is:
$$ h_{ew} = 2f_c \sin\alpha_e + \Delta h_{w,j} $$
where $f_c$ is the feed per tooth, $\alpha_e$ is the tool pressure angle, and $\Delta h_{w,j}$ accounts for insert radial runout.
1.3 Milling Force Model and Coefficient Calibration
A mechanistic milling force model was adopted, where the instantaneous tangential cutting force $F_{Tmj}$ on an insert is proportional to the uncut chip area. The force is composed of contributions from the side-cutting edge and the end-cutting edge (top land):
$$ F_{Tmj} = F_{Tmw} + F_{Tmt} = C_s \cdot (b_{mw}(t) h_{mw}) + C_s \cdot (b_{mt} h_{mt}) $$
The specific cutting pressure $C_s$ is modeled as a function of chip thickness and tool rake angle $\gamma$:
$$ C_s = C_{s1} h^{q-1} (1 – 0.01\gamma) $$
The radial and axial force components are derived from the tangential force using calibrated force ratios $f_{TR}$ and $f_{TA}$:
$$ F_{Rmj} = f_{TR} F_{Tmj}, \quad F_{Amj} = f_{TA} F_{Tmj} $$
A dedicated experimental setup with a dynamometer was used for orthogonal tests to calibrate the force coefficients $C_{s1}$, $q$, $f_{TR}$, and $f_{TA}$ for the specific workpiece material (20CrNiMo) and cutter (HSS). The calibrated values are summarized below.
| Workpiece Material | Tool Material | $C_{s1}$ (N/mm²) | $q$ | $f_{TR}$ (Outer) | $f_{TA}$ (Outer) | $f_{TR}$ (Inner) | $f_{TA}$ (Inner) |
|---|---|---|---|---|---|---|---|
| 20CrNiMo | HSS | 1856 | 0.25 | 0.346 | 0.135 | 0.362 | -0.126 |
1.4 Model Verification
A simulation program was developed, and cutting tests were performed on a YK2180 CNC gear cutting machine. The comparison between simulated and measured cutting forces in the X, Y, and Z directions showed a strong correlation in trend and magnitude. The relative error between simulation and experiment was within $\pm 15\%$, validating the accuracy of the proposed dynamic model for form gear cutting.
2. Cutting Force Modeling for Rough Cutting of the Pinion via the Modified Roll Method
The cutting process for the pinion (gear), especially using the modified roll method on a 4-axis CNC machine, is more complex due to the simultaneous rolling motion between the cutter and the workpiece. Modeling the cutting force here is crucial for predicting power consumption and optimizing the roughing cycle.
2.1 Processing Principle and Cutting Parameters
On a 4-axis CNC machine, the traditional mechanical roll motion is replicated by the coordinated linear movement of the X and Y axes ($v_{xy}$) coupled with the rotation of the workpiece axis ($C$-axis). The key cutting parameters for this gear cutting method are the cutter blade peripheral speed $v_c$, the generating speed $v_{xy}$, and the single-step interpolation angle $\Delta q$.
The angular velocity of the cutter head $\omega_t$ and the generating roll $\phi_c$ are:
$$ \omega_t = v_c / r_c, \quad \phi_c(t) = \phi_s + k \cdot (\phi_e – \phi_s) / N_c $$
where $k$ is the interpolation step index and $N_c$ is the total number of interpolation steps. The workpiece rotation $\phi_p$ is governed by a fixed ratio $m_{cp}$:
$$ \phi_p(t) = \phi_{p0} + \phi_c(t) / m_{cp} $$
2.2 Instantaneous Uncut Chip Area Calculation
The core of the force model lies in calculating the instantaneous uncut chip area $A_i$. The cutter surface, represented as a series of discrete cutting edges, and the pinion front cone surface are defined in a unified coordinate system. For a given interpolation step, the points of contact between a specific cutting edge $i$ and the workpiece surface are determined by solving the meshing equation and the surface intersection equation. The chip area removed by an outer insert is approximated by the area swept between its current position and the position of a following insert:
$$ A_{oi} \approx \frac{1}{2} s_i^2 \sin(\angle N_i M_i Q_i) $$
where $s_i$ is a distance parameter from the cutting edge to the contact point, and $\angle N_i M_i Q_i$ is the small angle between the normals of successive cutting edge positions. A similar approach is used for inner inserts.
2.3 Cutting Force and Power Model
Using the calculated chip area, the instantaneous tangential cutting force is:
$$ F_T(t) = C_s \cdot A_{ki}(t) \quad (k = i \text{ or } o) $$
The required cutting power $P_c(t)$ and the estimated main motor input power $P_m(t)$ are then:
$$ P_c(t) = F_T(t) \cdot v_c, \quad P_m(t) = \frac{P_c(t)}{\eta_m \eta} $$
where $\eta_m$ and $\eta$ are the motor and transmission efficiencies, respectively.
2.4 Indirect Experimental Validation
Direct force measurement during the complex rolling motion is challenging. Therefore, validation was performed by comparing the calculated main motor input power $P_m(t)$ with the output power signal from the machine’s spindle drive inverter, measured via a data acquisition card. The results for different sets of cutting parameters ($v_c$, $v_{xy}$) showed that the amplitude and timing of the calculated and measured power profiles were in strong agreement, confirming the model’s effectiveness for predicting loads in rolling-type gear cutting.
3. Realization of Variable Feed Rate Control Under Approximately Constant Cutting Force
A significant issue in form gear cutting with constant feed rate is the large fluctuation in instantaneous tangential cutting force as the depth of cut increases, leading to vibration, noise, and reduced tool life. To address this, I developed a variable feed rate control method based on maintaining an approximately constant cutting force.
3.1 Control Strategy
The method is an offline optimization applied to the NC code. For a given tool path (infeed distance $s$), the relationship between feed rate $f$ and cutting force $F_T$ at discrete points is approximated as linear. If $F_{1,i}$ and $F_{2,i}$ are the forces at point $i$ for two different constant feed rates $f_1$ and $f_2$, the feed rate $f_{lim,i}$ required to achieve a desired constant force $F_{lim}$ is calculated as:
$$ f_{lim,i} = \frac{f_2 – f_1}{F_{2,i} – F_{1,i}} (F_{lim} – F_{1,i}) + f_1 $$
This calculation is performed for points along the entire cutting path, and the resulting variable feed rates $f_{lim,i}$ are written into the NC program’s feed rate commands.
3.2 Simulation and Experimental Results
Simulations for a sample gear cutting job showed that while constant feed rates of 20 mm/min and 40 mm/min led to force increases from ~1250N to ~2500N and ~1900N to ~4250N respectively, the variable feed rate method successfully limited the force variation. For a target force $F_{lim}=2200$ N, the force remained around 2100N with significantly reduced fluctuation.
| Control Method | Feed Rate (mm/min) | Max $F_T$ (N) | Min $F_T$ (N) | Single-Tooth Cutting Time (s) |
|---|---|---|---|---|
| Constant Feed | 20 | ~2500 | ~1250 | 57.09 |
| Constant Feed | 40 | ~4250 | ~1900 | 28.55 |
| Variable Feed ($F_{lim}=1600$N) | Dynamic | ~1700 | ~1400 | 29.12 |
| Variable Feed ($F_{lim}=2200$N) | Dynamic | ~2250 | ~2100 | 21.75 |
Experiments on the YK2180R machine confirmed the simulation. The process with variable feed control was noticeably smoother, with reduced vibration and noise, while also reducing the single-tooth cutting time by 23.8% compared to the higher constant feed rate, demonstrating simultaneous gains in quality and gear cutting efficiency.
4. Optimization of Cutting Parameters for CNC Rough Gear Cutting
To systematically harness the capabilities of CNC gear cutting machines, I formulated and solved optimization problems for both form milling and pinion rough rolling, using the Harmony Search Algorithm (HSA).
4.1 Optimization for Form Milling
For form gear cutting, the objective was to minimize total batch processing time $T_t$. The optimization variables were cutter head rotational speed $n$ (related to $v_c$) and workpiece feed rate $f$.
Objective Function (Minimize):
$$ T_t = \frac{L_b}{v_{max}} + \sum_{i=1}^{Z_2} T_{si} + t_p + \frac{t_c}{N} $$
where $T_{si}$ is single-tooth machining time, $Z_2$ is tooth number, $t_p$ and $t_c$ are auxiliary and clamping times, and $N$ is batch size.
Constraints:
1. Cutting Speed: $v_{c,min} \le \frac{\pi n r_0}{30} \le v_{c,max}$
2. Feed Rate: $f_{min} \le f \le f_{max}$
3. Power: $\max(F_{Tmj}(t) \cdot v_c) / \eta \le P_{motor}$
4. Tool Life: $T = \frac{C_v}{v_c^m a_c^{x_v} f_c^{y_v} B^{p_v} z^{q_v}} \ge [T]$
Applying HSA with practical machine and tool data yielded optimal parameters that reduced processing time compared to handbook recommendations.
4.2 Multi-Objective Optimization for Pinion Rough Rolling
For the more complex pinion roughing process, a multi-objective optimization was conducted with variables $v_c$, $v_{xy}$, and $\Delta q$. The objectives were to minimize both unit processing time $t$ (maximize productivity) and unit cost $C$.
Normalized Combined Objective Function (Minimize):
$$ f(\mathbf{x}) = \lambda_1 t^*(\mathbf{x}) + \lambda_2 C^*(\mathbf{x}) $$
where $t^*$ and $C^*$ are normalized values of time and cost, and $\lambda_1 + \lambda_2 = 1$ are weighting factors. The unit cost model includes machine, tool, and labor costs:
$$ C = M t + \left( \frac{t_m}{T} \right) \left[ C_p + G(t_b + t_p) + \frac{C_0 + C_m}{k_3} \right] + G t_c $$
Constraints: included cutting speed, generating speed, interpolation angle bounds, power ($\max(F_T(t) \cdot v_c) / \eta \le P_{motor}$), and tool life.
The HSA was successfully applied. The table below shows sample optimization results for different weighting strategies, demonstrating the ability to trade between high efficiency and low cost in gear cutting.
| Optimization Goal | $\lambda_1 / \lambda_2$ | $v_c$ (m/min) | $v_{xy}$ (mm/min) | $\Delta q$ (°) | Theor. $t_m$ (min) | Measured $t_m$ (min) |
|---|---|---|---|---|---|---|
| Max Productivity | 1 / 0 | 47.62 | 75.52 | 0.104 | 25.79 | 25.83 |
| Min Cost | 0 / 1 | 37.64 | 69.62 | 0.173 | 27.57 | 27.62 |
| Balanced | 0.5 / 0.5 | 43.27 | 73.50 | 0.126 | 25.89 | 25.95 |
5. Influence of Cutting Parameters on Tooth Flank Residual Height in Finish Gear Cutting
For finish gear cutting, surface quality is paramount. Given the difficulty in directly measuring roughness on complex spiral bevel gear flanks, I established a model to predict the theoretical average scallop height $\bar{\delta}$, which serves as an indicator for surface roughness.
5.1 Mechanism and Calculation of Average Scallop Height
The scallop height originates from the discrete nature of cutter inserts and the interpolated rolling motion. The theoretical finished surface is an envelope of the continuous cutter surface. In reality, cutting is performed by discrete edges at discrete interpolation steps, leaving cusps between successive cutting passes.
The 3D problem is simplified by analyzing 2D scallop height in sections along the tooth profile direction. For a point $M_i$, which is the intersection of paths from two successive cutting edges, the scallop height $\delta_i$ is its normal distance to the theoretical flank surface. The average scallop height $\bar{\delta}$ over $N$ sampled points on the flank is:
$$ \bar{\delta} = \frac{1}{N} \sum_{i=1}^{N} \delta_i $$
An approximate surface roughness $R_s$ can be estimated as:
$$ R_s \approx \frac{1}{2} \bar{\delta} $$
5.2 Influence of Cutting Parameters
Analysis using the developed model for a specific pinion finish-cutting case revealed distinct influences:
- Cutter Speed ($v_c$): Increasing $v_c$ slightly reduces $\bar{\delta}$, but the effect is marginal.
- Generating Speed ($v_{xy}$): Increasing $v_{xy}$ slightly increases $\bar{\delta}$, as a faster roll reduces the density of cutting passes along the roll path.
- Single-Step Interpolation Angle ($\Delta q$): This parameter has the most significant impact. Reducing $\Delta q$ drastically increases the number of interpolation steps, making the cutting passes denser and substantially reducing the scallop height. The relationship is strongly non-linear; halving $\Delta q$ reduces $\bar{\delta}$ approximately fourfold.
| Cutting Parameter | Effect on Avg. Scallop Height $\bar{\delta}$ | Magnitude of Influence |
|---|---|---|
| Cutter Speed $v_c$ ↑ | Slight Decrease | Low |
| Generating Speed $v_{xy}$ ↑ | Slight Increase | Low |
| Interpolation Angle $\Delta q$ ↓ | Sharp Decrease | Very High |
5.3 Experimental Correlation
Finish-cutting tests were performed with different $\Delta q$ settings. While the actual measured roughness (via visual comparison with gauges) was higher than the calculated $R_s$ due to machine errors, vibration, etc., the trend was confirmed. A smaller $\Delta q$ (0.25°) produced a visibly finer surface finish compared to a larger $\Delta q$ (0.5°). Furthermore, gear inspection on a coordinate measuring machine showed that the finer interpolation led to smaller deviations of the actual flank from the theoretical design, confirming that controlling $\Delta q$ is a critical factor for achieving high precision in final gear cutting.
6. Conclusion and Outlook
This research provides a comprehensive framework for understanding and optimizing the CNC gear cutting process for spiral bevel gears. The developed dynamic cutting force models for both form and roll-gear cutting offer essential data for machine design, process monitoring, and parameter selection. The variable feed rate control method demonstrates a practical approach to stabilize cutting forces, improving surface quality, tool life, and efficiency. The implementation of the Harmony Search Algorithm for multi-objective optimization provides a systematic method to determine optimal cutting parameters, balancing productivity and cost. Finally, the analysis of the scallop height establishes a clear, predictable relationship between key CNC interpolation parameters and the theoretical surface quality, offering a valuable guideline for finish gear cutting.
Future work should extend the cutting force modeling to more complex gear cutting methods like the Modified Roll method for hypoid pinions. Integrating the optimization and surface prediction algorithms into a closed-loop, intelligent manufacturing system for spiral bevel gears would represent a significant step towards fully automated, optimal gear production. The foundational models developed in this thesis pave the way for such advanced intelligent gear cutting systems.