Research On Early Fault Diagnosis Of Gearbox Based On Embedded Sensors

In the traditional faulty diagnosis method based on vibrational signal processing,the vibration signals are measured from sensors installed on the surface of a gearbox.The signal propagation path is composed of many parts and complex.When the weak fault on the tooth profile of gear such as crack,pitting take place,vibration signals caused by these early faults are measured from sensors installed on the surface of a gearbox through gear body,gear shaft,bearing and structure of gearbox.The frequency components in the vibration signals from the gear mesh point to the sensor are changed in the transmission path(just like passing mechanical filter).There is signal amplitude attenuation,the high-frequency component amplitudes attenuate more quickly than the low-frequency components,furthermore some of faulty information in signals is lost.So the early weak fault is very difficult to detect.In order to solve this problem,this paper suggest a novel test approach for the crack on tooth root of gear using an embedded piezoelectric accelerometer.This enhances the ability to extract useful fault information by shortening the transmission path and provide early fault detection in gear transmission systems.In order to verify the effectiveness of method,the faulty transmission systems with 2mm and 4mm crack on the tooth root of gear are studied in the gear dynamic simulation and gear running test.The two aspects are as follows:(1)Dynamic simulation: the healthy gear system and fault gear systems are simulated using both differential equations and explicit dynamics methods.(2)Gear running test: a healthy gear system and systems with cracked(2mm and 4mm)gear faults are experimented on test rig.In order to extract the weak fault information from the experimental data,we use EEMD+TFPF method to eliminate noise from the vibration signals.Main research contents and conclusions are as follows:(1)Time-varying mesh stiffness of gear is calculated by finite element method under the conditions of health and different length crack gears.It is concluded that the time-varying meshing stiffness curve change in mesh cycle.A single stage gear system dynamics model is developed according to the change of meshing stiffness to solve the dynamic properties of gear in different condition.(2)Dynamic simulations are done by using ANSYS/LS-DYNA for the healthy gear and crack fault gear(including 2 mm and 4 mm crack length).Acceleration signal analysis for some nodes selected from the gear body and the shaft and bearing are done in spur gear transmission systems with tooth root 2mm and 4mm crack.The results show that because of amplitude attenuation and frequency components losses in the vibration signal during the transmission process,when crack length is 2mm vibration signals from the shaft and gear body can detect effectively crack fault and vibration signals from the gearbox sensors cannot find crack fault features.When crack length is 4mm vibration signals from the shaft and gear body and gearbox can detect effectively crack fault.A new method for gearbox early weak fault diagnosis is developed.(3)The vibration signals a gearbox are complex,nonlinear and non-stationary.This paper proposes a method based on Time-Frequency Peak Filtering(TFPF)and ensemble empirical mode decomposition(EEMD)noise reduction method combining.In view of the TFPF algorithm being restricted in the window length selection problem,we improve the balance in two aspects of the signal noise suppression and signal fidelity by using the EEMD method.The results show that the EEMD+TFPF method can effectively extract crack fault feature from intensive background noise.(4)For the gearbox early weak fault,vibration signals occur the loss of the frequency components on transmission path,this paper presents a novel test approach for tooth root crack fault diagnosis of a spur gearbox using an embedded piezoelectric accelerometer.The proposed approach is tested by analyzing experimental data from a healthy gear system and systems with cracked gear faults.In order to extract the weak fault information from the experimental data,EEMD+TFPF is used to eliminate noise from the vibration signals,and then the cyclic autocorrelation function is used to extract the frequency components.The results suggest that the proposed approach can effectively detect 2mm and 4mm crack faults,while traditional methods can only detect 4mm crack faults.This would be consistent with the conclusion of the third chapter.So the embedded sensor measurement method is more suitable the early fault diagnosis of gearbox.