Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.664740
Title: Computational intelligence for fault diagnosis in gearbox systems
Author: Alkhadafe, H.
ISNI:       0000 0004 5365 4411
Awarding Body: Nottingham Trent University
Current Institution: Nottingham Trent University
Date of Award: 2015
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
Employing an efficient condition monitoring system in industrial applications is an important factor in improving the quality of production and increasing the operational life of machines by revealing machine faults at the earlier stage. Damage in gearbox system is one of the most catastrophic failures in machineries. Any defects related to a gearbox will influence the performance of an entire mechanical system. A reliable and efficient fault diagnosis system is required to reduce the maintenance cost and downtime, thereby preventing machinery performance degradation and failure. Many condition monitoring and fault diagnosis systems are investigated in the literature for gearbox fault detection and diagnosis. However, there are still many challenges to tackle mainly due to the complex nature of gearbox structure, limited access to the component to be monitored and the low signal-to-noise ratio experienced especially when operating machineries under fault conditions. The aim of this research is to develop a systematic methodology for the design of condition monitoring systems for gearbox faults by investigating sensor selection, sensor location, and sensory features to be able to diagnose a fault accurately. Therefore, the goal is to select reliable techniques at each stage in order to improve the reliability of the fault diagnosis system. Different sets of experiments based on gearbox conditions are conducted using several sensors including vibration, acoustic emission, speed, and torque. Measured signals are analysed using conventional and advanced signal processing and data analysis methods including time, frequency and time/frequency domains such as Fast Fourier Transform (FFT), Short Time Fourier Transform (STFT), and Wavelet analysis (WT). Several statistical and mathematical techniques have been proposed as features extraction methods to reduce the dimensionality and select appropriate information. For this research, a single stage gearbox system with two main type of faults (pitting and broken teeth) with various degrees of damage in helical gear are used to evaluate the proposed approach. This research investigated the relationship between sensor location and detecting the fault in gearbox system. A methodology has been proposed for locating indirect monitoring sensors such as acoustic emission and vibration on gearbox to obtain high quality information regarding the behaviour of machine condition. The methodology is designed to evaluate the optimum sensor positioning for detecting faults in the gearbox system. A novel gearbox monitoring approach named an Automated Sensor and Signal Processing Selection for Gearbox system (ASPSG) has been applied to select the most reliable and sensitive sensors, features and signal processing methods based on optimal sensor location. The ASPSG approach is based on simplifying complex sensory signals into a group of Sensory Characteristic Features (SCFs) and evaluating the sensitivity of these SCFs in detecting gearbox faults. The aim of this approach is to enhance the performance of monitoring system of gearbox fault detection and to reduce the number of sensors required in the overall system and reduce the cost. To implement the suggested ASPSG approach two strategies are proposed: automated system based on Taguchi's orthogonal arrays and stepwise system using (Linear Regression (LR), Fuzzy Rule Based System (FRBS) and Principal Component Analysis (PCA), techniques ). To evaluate both strategies, four dierent classication models are employed using supervised and unsupervised neural networks. Both strategies have been implemented to prove the capability of the suggested approach. A cost reduction is performed based on removing the least utilised sensors without losing the performance of the condition monitoring system. The results show that the ASPSG approach can provide a systematic design methodology for condition monitoring systems for gearboxes and that it is capable of detecting faults in a gearbox system with less cost and reduced number of experiments. Consequently, the ndings of this research prove that the sensor location could have signicant effect on the design of the condition monitoring system and its performance.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.664740  DOI: Not available
Share: