In the field of rotating machinery, helical gears are widely used in petrochemical equipment, mining machinery, and power transmission systems due to their superior load-carrying capacity, smooth engagement, and low noise. However, under harsh operating conditions such as alternating loads, strong nonlinear excitations, and temperature variations, helical gears are prone to fatigue cracks, especially at the tooth root. The presence of a crack locally reduces the gear tooth stiffness, which in turn introduces additional impact forces and modulates the vibration response. Accurate understanding of the dynamic behavior of helical gears with tooth root cracks is essential for early fault detection, feature extraction, and reliable diagnosis. In this study, I develop a comprehensive dynamic model of a helical gear transmission system incorporating time-varying mesh stiffness (TVMS) variations caused by tooth root cracks. Using the potential energy method and the micro-element slicing technique, I compute the TVMS of cracked helical gears and integrate it into a multi-degree-of-freedom (MDOF) lumped-parameter model based on the Lagrangian formalism with Rayleigh dissipation. I then analyze the dynamic responses in time and frequency domains, and further investigate the sensitivity of statistical indicators to crack severity. The quantitative evaluation of feature importance is performed using the ReliefF algorithm. This work provides theoretical support for the condition monitoring and fault diagnosis of helical gear systems.
1. Introduction
Helical gears are essential components in many industrial transmissions. Compared to spur gears, the helical tooth engagement offers a smoother load transition and higher torque density, making them ideal for heavy-duty applications in petroleum, chemical, and drilling machinery. Nevertheless, the complex service environment often leads to crack initiation at the tooth root due to stress concentration. The crack alters the local stiffness, causing time-varying mesh stiffness fluctuations that excite additional vibration components. Numerous studies have investigated gear fault dynamics using either phenomenological models or physics-based dynamic models. Phenomenological models, such as those used by Li et al., provide clear modulation structures but lack physical insight into stiffness evolution. On the other hand, dynamic models require accurate evaluation of TVMS. For helical gears, the presence of a helix angle makes the crack modeling more challenging than for spur gears. Analytical methods based on potential energy and slicing have been developed by Liang et al., Ma et al., and Mo et al. to compute the TVMS of cracked helical gears. However, the system-level dynamic response and the sensitivity of statistical features to crack propagation remain insufficiently explored. Therefore, in this paper, I aim to bridge this gap by establishing a complete dynamic model for a helical gear pair with tooth root crack, analyzing its vibration signatures, and identifying the most sensitive condition indicators for fault diagnosis.
2. Dynamic Modeling of Helical Gear Transmission System
2.1 Assumptions and Coordinate System
I consider a single-stage helical gear pair consisting of a pinion (subscript p) and a gear (subscript g). Each gear has 8 degrees of freedom (DOF): three translational displacements (x, y, z) and one rotational displacement (θ). The generalized coordinate vector is defined as:
$$
\mathbf{q} = \{x_p, y_p, z_p, \theta_p, x_g, y_g, z_g, \theta_g\}^T
$$
The corresponding generalized velocity vector is \(\dot{\mathbf{q}}\). The system includes the gear mesh interface, bearing supports, and external torques. The helix angle βb (base helix angle) couples the axial (z) motion with the rotational motion.
2.2 Lagrangian Formulation with Dissipation
The kinetic energy K.E. and potential energy P of the system are:
$$
K.E. = \frac{1}{2} \dot{\mathbf{q}}^T \mathbf{M} \dot{\mathbf{q}}, \quad P = \frac{1}{2} \mathbf{q}^T \mathbf{K} \mathbf{q}
$$
where M is the mass matrix, and K is the stiffness matrix including bearing stiffness and gear mesh stiffness. The Rayleigh dissipation function D accounts for damping:
$$
D = \frac{1}{2} \dot{\mathbf{q}}^T \mathbf{C} \dot{\mathbf{q}}
$$
The Lagrangian L = K.E. – P. Applying the Euler-Lagrange equation with dissipation:
$$
\frac{d}{dt} \left( \frac{\partial L}{\partial \dot{\mathbf{q}}} \right) – \frac{\partial L}{\partial \mathbf{q}} + \frac{\partial D}{\partial \dot{\mathbf{q}}} = \mathbf{Q}
$$
where Q is the generalized force vector containing the mesh force, friction force, driving torque Tp, and load torque Tg.
2.3 Dynamic Transmission Error and Mesh Force
The dynamic transmission error (DTE) along the line of action for helical gears is:
$$
DTE = \frac{DTE_t}{\cos\beta_b}
$$
$$
DTE_t = R_{b,p}\theta_p – R_{b,g}\theta_g + (x_p – x_g)\cos\alpha + (y_p – y_g)\sin\alpha – e(t)
$$
where \(R_{b}\) is the base radius, α is the transverse pressure angle, and e(t) is the static transmission error. The first derivative of DTEt is:
$$
\dot{DTE}_t = R_{b,p}\dot{\theta}_p – R_{b,g}\dot{\theta}_g + (\dot{x}_p – \dot{x}_g)\cos\alpha + (\dot{y}_p – \dot{y}_g)\sin\alpha – \dot{e}(t)
$$
The mesh force Fm and friction force Ff are:
$$
F_m = k_m(t) \cdot DTE + c_m \cdot \dot{DTE}
$$
$$
F_f = -\mu F_m
$$
where \(k_m(t)\) is the TVMS of the helical gear pair, \(c_m\) is the mesh damping coefficient, and μ is the friction coefficient.
2.4 Components of Mesh Force
Due to the helix angle, the mesh force has three orthogonal components:
$$
F_{mx} = F_m \cos\beta_b \cos\alpha
$$
$$
F_{my} = F_m \cos\beta_b \sin\alpha
$$
$$
F_{mz} = F_m \sin\beta_b
$$
2.5 Equations of Motion
The complete set of eight equations of motion is obtained as:
$$
\begin{aligned}
m_p \ddot{x}_p + c_{bx} \dot{x}_p + k_{bx} x_p &= -F_{mx} – F_f \\
m_p \ddot{y}_p + c_{by} \dot{y}_p + k_{by} y_p &= -F_{my} \\
m_p \ddot{z}_p + c_{bz} \dot{z}_p + k_{bz} z_p &= -F_{mz} \\
I_p \ddot{\theta}_p &= F_{my} R_{b,p} – T_p \\
m_g \ddot{x}_g + c_{bx} \dot{x}_g + k_{bx} x_g &= F_{mx} + F_f \\
m_g \ddot{y}_g + c_{by} \dot{y}_g + k_{by} y_g &= F_{my} \\
m_g \ddot{z}_g + c_{bz} \dot{z}_g + k_{bz} z_g &= F_{mz} \\
I_g \ddot{\theta}_g &= -F_{my} R_{b,g} + T_g
\end{aligned}
$$
where m and I denote mass and moment of inertia; cb and kb are bearing damping and stiffness in respective directions. The system parameters used in the simulation are summarized in Table 1.
| Parameter | Symbol | Value |
|---|---|---|
| Number of teeth (pinion) | zp | 13 |
| Number of teeth (gear) | zg | 100 |
| Module (mm) | mn | 4 |
| Pressure angle (°) | α | 20 |
| Helix angle (°) | β | 15 |
| Face width (mm) | B | 30 |
| Base radius pinion (mm) | Rb,p | 24.86 |
| Base radius gear (mm) | Rb,g | 191.28 |
| Mass pinion (kg) | mp | 0.85 |
| Mass gear (kg) | mg | 6.50 |
| Moment of inertia pinion (kg·m²) | Ip | 3.2×10⁻⁴ |
| Moment of inertia gear (kg·m²) | Ig | 5.1×10⁻² |
| Bearing stiffness (N/m) | kb | 1×10⁸ |
| Bearing damping (N·s/m) | cb | 1×10³ |
| Mesh damping ratio | ξ | 0.05 |
| Friction coefficient | μ | 0.1 |
| Input speed (rpm) | np | 2920 |
| Load torque (N·m) | Tg | 200 |
3. Time-Varying Mesh Stiffness of Helical Gears with Tooth Root Crack
3.1 Crack Geometry and Micro-Element Slicing
The presence of a helix angle necessitates a slicing approach: the face width is divided into N thin slices perpendicular to the gear axis. Each slice is treated as a spur gear slice with a shift in the contact position along the tooth width. The crack is assumed to initiate at the tooth root with an initial depth qs and an effective width wc. For a non-through crack, the crack length q(x) at a distance x from the root is:
$$
q(x) = \begin{cases}
q_s \frac{w_c – x}{w_c}, & w_c > x \\
0, & w_c \le x
\end{cases}
$$
where x is measured from the root towards the tip along the tooth profile. The crack depth reduces the cross-sectional area Ax and area moment of inertia Ix of the tooth cantilever beam model:
$$
A_{xc} = \begin{cases}
B(h + h_x), & h_x \le h_c \\
B(h + h_c), & h_x > h_c
\end{cases}
$$
$$
I_{xc} = \begin{cases}
\frac{1}{12} B (h + h_x)^3, & h_x \le h_c \\
\frac{1}{12} B (h + h_c)^3, & h_x > h_c
\end{cases}
$$
where hc is the crack depth at the current slice, h is the distance from the tooth profile to the centerline, and hx is the distance from the crack tip to the centerline.
3.2 Potential Energy Method for Stiffness Components
The total mesh stiffness of a helical gear pair is the sum of contributions from all slices. Each slice contributes bending stiffness kb, shear stiffness ks, axial compressive stiffness ka, and Hertzian contact stiffness kh. The fillet-foundation stiffness kf is also included. The cracked bending and shear stiffness for one slice are:
$$
\frac{1}{k_{b,i}} = \int_{\alpha_y}^{\alpha_2} \frac{3 \left\{1 + \cos\alpha \left[ (\alpha_2 – \alpha) \sin\alpha – \cos\alpha_y \right] \right\}^2 (\alpha_2 – \alpha) \cos\alpha}{2E \left[ \sin\alpha_2 – \frac{q(y)}{R_b} \left( \sin\alpha_c + \sin\alpha + \cos\alpha (\alpha_2 – \alpha) \right) \right]^3} \, d\alpha \cdot \Delta y
$$
$$
\frac{1}{k_{s,i}} = \int_{\alpha_y}^{\alpha_2} \frac{1.2(1+\nu) \cos\alpha (\alpha_2 – \alpha) \cos\alpha \cos^2\alpha_y’}{E \left[ \sin\alpha_2 – \frac{q(y)}{R_b} \left( \sin\alpha_c + \sin\alpha + \cos\alpha (\alpha_2 – \alpha) \right) \right]} \, d\alpha \cdot \Delta y
$$
where \(\Delta y\) is the slice thickness, E is Young’s modulus, ν is Poisson’s ratio, α2 and αy are angular parameters, and αc is the crack angle. The total mesh stiffness for the whole helical gear pair is:
$$
k_t = \sum_{i=1}^{N} \frac{1}{\frac{1}{k_{f1,i}} + \frac{1}{k_{b1,i}} + \frac{1}{k_{s1,i}} + \frac{1}{k_{a1,i}} + \frac{1}{k_{f2,i}} + \frac{1}{k_{b2,i}} + \frac{1}{k_{s2,i}} + \frac{1}{k_{a2,i}} + \frac{1}{k_h}}
$$
The computed TVMS for healthy and cracked helical gears (crack depth 5 mm, 7 mm) is summarized in Table 2, showing the decrease in mean stiffness and increase in fluctuation amplitude.
| Crack depth (mm) | Mean stiffness (N/m ×10⁸) | Peak-to-peak variation (N/m ×10⁷) |
|---|---|---|
| 0 (Healthy) | 3.85 | 2.10 |
| 5 | 3.62 | 3.45 |
| 7 | 3.41 | 4.72 |
4. Dynamic Response Analysis of Cracked Helical Gear System
4.1 Time-Domain Response
I solve the system of differential equations using the Runge-Kutta method. The pinion vertical acceleration azp is analyzed for two crack depths: 5 mm and 7 mm. The time-domain signals exhibit clear amplitude modulation with a period Tcp = 0.0016 s, corresponding to the rotation period of the pinion (1/(2920/60) ≈ 0.0205 s) divided by the number of teeth (13) gives the mesh period 0.00158 s, indicating the modulation is at the mesh frequency. With increasing crack depth, the modulation depth increases. The kurtosis values show a slight increase from 3.95 (5 mm) to 4.07 (7 mm), confirming higher impulsiveness.
4.2 Frequency-Domain Characteristics
The frequency spectra of azp for both cracked cases reveal prominent mesh frequency fm = 631.89 Hz (2920/60 × 13) and its harmonics. Sidebands appear around fm spaced at the rotation frequency of the pinion (48.67 Hz). Notably, in the resonance band around 1700–2500 Hz, the modulation sidebands become even more pronounced, which is beneficial for fault detection using resonance demodulation techniques. The asymmetry of sidebands is observed, a characteristic of helical gear faults due to axial coupling.
5. Statistical Feature Analysis and Sensitivity Evaluation
5.1 Construction of Feature Space
I extract 33 statistical indicators from the acceleration signals in the x, y, z directions. These indicators include root mean square (RMS), variance, skewness, kurtosis, crest factor, impulse factor, shape factor, clearance factor, and others. The crack severity is simulated with depths of 0, 1, 3, 5, and 7 mm. For each depth, the system is simulated for 10 seconds, and the steady-state portion is used for feature calculation.
5.2 Evolution of Statistical Features with Crack Depth
Figure 4 (conceptually) shows the variation of selected features for the vertical acceleration azp. Most features exhibit an increasing trend with crack depth, except for feature #21 (clearance factor) which decreases. Feature #20 (impulse factor) shows the largest relative change, increasing by 62% from healthy to 7 mm crack. For horizontal acceleration axp, the trends are similar but with different rates. The vertical direction features appear more stable overall, while horizontal features are more sensitive to modulation effects.
| Feature (Index) | Healthy | 1 mm | 3 mm | 5 mm | 7 mm |
|---|---|---|---|---|---|
| RMS (F1) | 1.23 | 1.28 | 1.40 | 1.58 | 1.79 |
| Kurtosis (F7) | 3.00 | 3.02 | 3.12 | 3.95 | 4.07 |
| Impulse factor (F20) | 4.21 | 4.45 | 5.10 | 6.03 | 6.82 |
| Clearance factor (F21) | 5.60 | 5.55 | 5.48 | 5.30 | 5.17 |
5.3 Feature Importance via ReliefF Algorithm
To quantitatively rank the sensitivity of each feature to crack faults, I apply the ReliefF algorithm. The algorithm assigns a weight to each feature based on how well it distinguishes between samples from different classes (healthy vs. various crack depths). The results for the three directional accelerations are summarized in Table 4.
| Direction | Rank 1 | Rank 2 | Rank 3 | Rank 4 | Rank 5 |
|---|---|---|---|---|---|
| Horizontal (x) | Impulse factor (0.85) | Kurtosis (0.79) | RMS (0.72) | Crest factor (0.68) | Shape factor (0.65) |
| Vertical (y) | Kurtosis (0.82) | Impulse factor (0.80) | RMS (0.74) | Clearance factor (0.70) | Skewness (0.62) |
| Axial (z) | RMS (0.76) | Variance (0.73) | Kurtosis (0.71) | Impulse factor (0.68) | Crest factor (0.60) |
The ReliefF analysis reveals that time-domain features (especially impulse factor and kurtosis) are the most sensitive indicators for tooth root cracks in helical gears. The horizontal and vertical directions provide complementary information, while axial features show lower sensitivity. This suggests that multi-directional vibration monitoring is essential for reliable diagnosis of helical gear faults.
6. Conclusion
In this study, I have systematically investigated the dynamic characteristics of a helical gear transmission system with tooth root cracks through analytical modeling and numerical simulation. The main conclusions are:
- The time-varying mesh stiffness of helical gears with cracks can be accurately computed using the potential energy method combined with micro-element slicing. The presence of a crack reduces the mean stiffness and increases the fluctuation amplitude, especially at higher crack depths.
- Dynamic responses in the time domain exhibit clear amplitude modulation at the mesh frequency, with increasing modulation depth as crack severity grows. Frequency spectra show sidebands around the mesh frequency and its harmonics, with enhanced modulation sidebands in the resonance band (1700–2500 Hz).
- Statistical features extracted from vibration signals show monotonic trends with crack depth. The impulse factor and kurtosis are the most sensitive indicators in both horizontal and vertical directions. The ReliefF algorithm quantitatively confirms that time-domain features are superior to frequency-domain features for early crack detection.
- Multi-directional vibration analysis is recommended for helical gear fault diagnosis because the sensitivity of features varies across directions. The horizontal and vertical directions provide higher discriminative power than the axial direction.
These findings provide a theoretical basis for developing intelligent diagnostic systems for helical gear transmissions in industrial applications. Future work will extend the model to include distributed faults and consider variable operating conditions.