Adaptive Damage Accumulation and Remaining-Service-Life-Prediction for Gearboxes
Final Report Abstract
Gearboxes surround us everywhere, from small applications, like the drive of a kitchen appliance, to gearboxes of electric vehicles and the transmissions of wind turbines or ships. Centerpiece of a gearbox are the gears, which in many cases are dimensioned for a limited service life. This enables an high resource efficiency, but the probability of a failure increases with proceeding service life. Especially if a failure would cause substantial economic or social damage, machinery is maintained or replaced earlier than necessary. This causes increased operating costs and an increased carbon footprint. Fatigue damages, like pitting and tooth root breakage, limit the operational life of gears. Usually, gear fatigue can only be noticed through inspection within a short period of time before failure. Therefore, theoretical methods for the calculation of the state of damage are of high importance for the prediction of the remaining useful life. The methodical basis for the calculation of the damage caused by variable loads date back to the first half of the 20th century. These methods are referred to as damage accumulation hypotheses and have proven as well applicable and reliable. But some important influence factors, like the load sequence, are not accounted for by these methods. Within this project theoretical and experimental investigations were conducted to create a methodical basis for the accuracy improvement of the remaining useful life prediction of gears. To predict the remaining useful life a prognosis of the expected future loads and an estimation of the actual state of damage are required. The project examines potentials for the accuracy increasement of both input factors. Regarding the calculation of the state of damage concepts to improve the accuracy of the load carrying capacity description and concepts to included information about the load sequence were analyzed. The potential of non-linear damage accumulation was investigated to enable a full consideration of the load sequence within the calculation. This approach showed a lot of potential, but the correct parameterization of these methods represents a challenge. A main emphasis of the project was the application of machine learning. Through the use of these methods’ potential for further development regarding the load carrying capacity description has been shown. The project was able to point out promising methodical approaches for the accuracy improvement of the calculation of the remaining useful life of gears.
Publications
-
Method for a cloud based remaining-service-life-prediction for vehicle-gearboxes based on big-data-analysis and machine learning. Forschung im Ingenieurwesen, 84(4), 305-314.
Vietze, Daniel; Hein, Michael & Stahl, Karsten
-
Non-Linear Analysis of Gear-Fatigue-Damage Under Variable Load - 23FTM08. AGMA Technical Paper.
Vietze, D.; Pellkofer, J. & Stahl, K.
