In modern railway systems, the reliability of traction motors is critical for safe and efficient operations. The JD160 traction motor, used in HXD1C locomotives, has been prone to gear shaft and sleeve failures, which can lead to severe operational disruptions and safety hazards. These failures often occur within a short window of about seven days, making timely detection challenging. Moreover, the vibration signals associated with these faults exhibit low decibel (dB) values, typically below the standard alarm threshold of 51 dB, further complicating accurate diagnosis. This paper presents an intelligent diagnostic approach that integrates on-board running gear quality monitoring (AT1) subsystem data analysis with ultrasonic non-destructive testing to enhance fault detection accuracy and prevent operational risks.
The gear shaft in the JD160 traction motor is a critical component that transmits torque and withstands significant mechanical stresses. Failures often initiate at stress concentration points, such as oil grooves and radial oil holes, where inadequate surface finishing or geometric transitions can lead to crack initiation and propagation. For instance, in double-groove gear shafts, the first oil groove located in the bearing load-bearing area is particularly susceptible. Stress concentration factors, denoted by $K_t$, can be expressed as:
$$K_t = 1 + 2\sqrt{\frac{a}{\rho}}$$
where $a$ represents the notch depth and $\rho$ is the root radius. In cases where machining marks or grooves are present, the effective stress intensity factor $K_{eff}$ increases, accelerating fatigue crack growth according to Paris’ law:
$$\frac{da}{dN} = C(\Delta K)^m$$
Here, $da/dN$ is the crack growth rate per cycle, $\Delta K$ is the stress intensity factor range, and $C$ and $m$ are material constants. Initial cracks often form at the intersection of oil grooves and radial holes, propagating under cyclic loading until fracture occurs. This failure mechanism underscores the need for a proactive diagnostic strategy.
The intelligent diagnostic scheme leverages the AT1 subsystem, which monitors vibration signals from sensors positioned at specific locations on the motor assembly. Based on structural analysis, sensors are installed at positions 2 and 4, relative to the D-end motor bearing centerline. Position 4, with a longer cantilever, captures slightly stronger signals (1–3 dB higher) than position 2, making these points ideal for detecting gear shaft anomalies. The diagnostic process involves automated screening followed by manual verification and ultrasonic testing.
Automated screening uses AT1 ground analysis software with a dedicated patch for gear shaft fault identification. The quantitative criterion sets the abnormal vibration level $Y$ for positions 2 and 4 within the range:
$$45 \leq Y < 51 \text{ dB}$$
Qualitatively, the frequency domain spectrum must show peaks corresponding to the gear mesh frequency and its first three harmonics. If both conditions are met, the system flags a “concern” for potential gear shaft issues. For example, the vibration energy $E$ related to gear faults can be modeled as:
$$E = \int_{f_1}^{f_2} |X(f)|^2 df$$
where $X(f)$ is the Fourier transform of the vibration signal, and $f_1$ to $f_2$ cover the gear mesh harmonics. This automated step efficiently narrows down cases requiring detailed analysis.
Upon receiving a “concern” alert, engineers perform manual verification by examining time-domain waveforms. The key criterion is the presence of “pinion cycle lines,” where the ratio of fault-indicative black vertical lines to total gray pinion cycle lines exceeds 80%. This is calculated as:
$$R = \frac{N_{\text{fault}}}{N_{\text{total}}} \times 100\% > 80\%$$
Here, $N_{\text{fault}}$ is the number of fault lines, and $N_{\text{total}}$ is the total pinion cycles. Based on this, faults are classified into two categories:
- Gear Shaft Spectrum Anomaly Type II Fault: Abnormal patterns simultaneously at both positions 2 and 4.
- Gear Shaft Spectrum Anomaly Type I Fault: Abnormal patterns at either position 2 or 4 alone.
Type II faults warrant immediate motor replacement, while Type I faults trigger further inspection, including visual checks and ultrasonic testing of the gear shaft and sleeve.
Ultrasonic non-destructive testing is employed to detect internal cracks in the gear shaft. Using a digital ultrasonic flaw detector with a 4P20 3.5° longitudinal wave probe, the inspection focuses on areas 245 mm to 260 mm from the gear shaft end face. The process involves:
- Equipment Calibration: The probe is calibrated on standard blocks (e.g., CSK-1) to set reference values for velocity, zero offset, and probe angle. The detection range is set to 400 mm.
- Sensitivity Adjustment: The initial sensitivity is set by maximizing the reflection from a 2 mm deep artificial notch at 254.5 mm in a reference shaft, adjusted to 80% full screen height.
- Scanning: The probe scans the end face in a zigzag pattern at speeds below 50 mm/s, with couplant ensuring acoustic coupling. Defects are quantified by comparing reflected wave amplitudes to the reference in dB.
The ultrasonic wave propagation and reflection can be described by the wave equation:
$$\frac{\partial^2 u}{\partial t^2} = c^2 \nabla^2 u$$
where $u$ is the displacement and $c$ is the wave speed. Defects like cracks cause reflections, and the amplitude ratio $A_r$ for a defect of size $d$ is approximated by:
$$A_r = A_0 e^{-\alpha x} \left( \frac{d}{\lambda} \right)^2$$
where $A_0$ is the initial amplitude, $\alpha$ is the attenuation coefficient, $x$ is the distance, and $\lambda$ is the wavelength. Defects exceeding thresholds lead to gear shaft replacement.
The effectiveness of this intelligent diagnostic approach is demonstrated through practical applications. Below is a summary of fault cases identified using the AT1 data analysis and ultrasonic testing integration:
| Case No. | Locomotive ID | Axle Position | Analysis Date | Motor Disassembly Findings | AT1 Data Characteristics |
|---|---|---|---|---|---|
| 1 | HXD1C-0601 | 2 | 2021-01-05 | Gear shaft crack, sleeve crack | Signals at positions 2 and 4, Y < 51 dB |
| 2 | HXD1C-0606 | 5 | 2021-01-05 | Gear shaft crack, sleeve crack | Signals at positions 2 and 4, Y < 51 dB |
| 3 | HXD1C-0656 | 4 | 2021-05-18 | Gear shaft crack | Weak signal at position 4 |
| 4 | HXD1C-6198 | 1 | 2021-05-20 | Gear shaft crack | Signal at position 4 |
| 5 | HXD1C-0655 | 2 | 2021-05-20 | Gear shaft crack, sleeve crack | Signal at position 4 |
| 6 | HXD1C-0623 | 4 | 2021-06-24 | Gear shaft crack, sleeve crack | Weak signal at position 4 |
| 7 | HXD1C-0619 | 2, 6 | 2021-06-25 | Gear shaft crack, sleeve crack | Signals at positions 2 and 4 |
| 8 | HXD1C-6218 | 5 | 2021-06-28 | Gear shaft crack, sleeve crack | Signals at positions 2 and 4 |
| 9 | HXD1C-0615 | 4 | 2021-08-31 | Sleeve crack | Signal at position 2 |
| 10 | HXD1C-0906 | 4 | 2021-08-31 | Sleeve crack | Weak signal at position 4 |
| 11 | HXD1C-6002 | 2 | 2021-09-11 | Gear shaft crack, sleeve crack | Signals at positions 2 and 4 |
Implementation of this diagnostic scheme significantly improved fault detection accuracy. Prior to adopting the AT1 data analysis分级标准, only 13.2% of suspected cases were confirmed as actual gear shaft faults. After integration, the accuracy reached 100%, as validated by ultrasonic testing and disassembly findings. This approach prevents operational risks by ensuring faulty gear shafts are identified and replaced promptly. The combination of AT1 monitoring and ultrasonic testing leverages the strengths of both methods: continuous data analysis for early warning and precise defect characterization for confirmation.
In conclusion, the intelligent diagnostic framework for JD160 traction motor gear shaft failures effectively addresses the challenges of low-intensity vibration signals and short fault windows. By employing AT1 data analysis with quantitative thresholds and qualitative waveform checks, followed by ultrasonic non-destructive testing, the system achieves high reliability in detecting gear shaft anomalies. This methodology not only enhances railway safety but also optimizes maintenance workflows, demonstrating the value of integrated technologies in modern rail systems. Future work could explore machine learning algorithms to further refine fault prediction models for gear shaft integrity.