Author Description

The rapid growth of manufacturing industry impelled the automation in large-scale assembly lines. Consequently, it made the manufacturing process more vulnerable to various kinds of machine failures resulting in more frequent, complex and unexpected breakdowns which can cause a lot of damage. Condition-based maintenance (CBM) of machinery, which is utilizing condition monitoring techniques to monitor the machine health condition, makes it possible to evaluate the need for maintenance in advance and to plan the maintenance action without interrupting production process.Generally, a simple condition monitoring system can be divided into three general tasks: data processing (signal processing), feature extraction, and condition classification. Therefore, this thesis attempts to propose innovative and effective condition monitoring system from these three aspects. In signal processing, singularity analysis using wavelet transform is applied to identify the transient signals related to machine failure through wavelet modulus maxima. To investigate this method, modulus maxima distribution is proposed as a feature and two health indexes is defined and evaluated. Further, in order to provide decision information for CBM, a two-stage HMM-based classification system is presented using the feature extracted from wavelet modulus maxima. A variety of experimental evaluations demonstrate that the proposed system possesses remarkable advantages over well-accepted conventional techniques.