AbstractThe reliability of epicyclic gearboxes is critical in today’s industrial world due to their large transmission capacity and non-linear characteristics. Epicyclic gearboxes are widely used in the automotive, aerospace and energy sectors. Condition monitoring and prognostics of epicyclic gearboxes are necessary to reduce downtime, increase production uptime and prevent catastrophic failures. Managing asset health using condition monitoring technique is pervasive in the modern world. A complex mechanical system like epicyclic gearboxes are exposed to many distortions. These are often challenging to measure for reliably diagnosing the fault in the gearbox without a physical inspection. A reliable condition monitoring technique is vital to reduce downtime. Failure of these systems and their subsystems will have serious ramifications. Many exploratory studies have conducted in the past in the field of condition monitoring. However, very few have validated their results and explored the accuracy of the developed tool. In this research, a diagnostic and prognostic tool can detect the fault, identify the faulty component and estimate the remaining useful life of the gearbox. Several computational techniques are proposed in this research that manifest the effectiveness of the developed tool.
This research aims to develop a diagnostics and prognostic tool for the planetary gearbox, which comprises five development stages; numerical modelling, Finite element modelling, fault detection and diagnostics, fault severity and classification, and prognostics. The numerical model developed using Newtonian and Lagrange equations that have investigated the backlash effect, change in stiffness of tooth, modal and resonant frequencies. This study applied the perturbation method to solve the implications of the system. The dynamic modelling also deals with the gear mesh frequencies and time-varying mesh stiffness in healthy and non-healthy conditions. The gear mesh harmonics have also been studied using the dynamic model. The change in stiffness due to crack and backlash have an impending impact on the system’s vibration. The gearbox’s finite element model with similar parameters has been developed for studying the modal and harmonic features. The effect of tooth flank fracture or pitting on modal frequencies and acceleration of the system has studied. Interestingly the rotational modal frequencies obtained from the numerical model are matching with a slight deviation. These results have validated the mode shape of the system.
The experimental research test rig under a constant load has constructed to detect the fault and diagnose the fault type present in the gearbox. An autoregressive continuous wavelet function has been developed to study different fault conditions. The vibration signatures extracted from the gearbox validated against the results obtained from the finite element model. In addition, the AR-CWT tool developed shown the fault type and the component affected during the investigation. The pitting, tooth flank fracture and more severe cases have been studied and diagnosed using the developed tool. Additionally, as a continuation of tool development, classification of the fault type and their severity have been performed using the K-means clustering algorithm. The tool developed have favourably shown the results by clustering the fault groups. Moreover, the K-means tool reliably estimated the crack depth level of fault when the crack propagates over a certain threshold.
The life prediction of the planetary gearbox is as essential as diagnostics from the reliability point of view. As an encompassed objective of this research, a prognostic tool using an artificial intelligence convolution neural network (CNN) and long-term short memory (LSTM) algorithm has been developed to predict the remaining useful life of the planetary gearbox. The experimental life and the theoretical life of the gearbox has been discussed. The developed tool reliably estimated the remaining useful life of the gearbox with greater accuracy. Furthermore, the results accomplished from the life prediction are validated against the gearbox’s real-life from the manufacturer.
The diagnostics and prognostics tool developed in this research using several computational techniques has quantified and proved to be an effective tool for a planetary gearbox for condition monitoring.
|Date of Award||2022|
|Supervisor||Faris Elasha (Supervisor) & David Trepess (Supervisor)|