Tadpon Kullawong. Transformation reliability engineering for maintenance management. Doctoral Degree(Production Engineering). King Mongkut's University of Technology North Bangkok. Central Library. : King Mongkut's University of Technology North Bangkok, 2015.
Transformation reliability engineering for maintenance management
Abstract:
Maintenance management at present suffers from problems referring to maintenance data and machine history accuracy, consistent and constant information resulting in inefficiency of planning, maintenance, diagnosis and prognosis of machinery disorders. For this reason, it is necessary to develop tools and procedures to ensure compliance with these problems. Thus, the objective of this research was to create models which modified the data to increase reliability of engineering maintenance management.
The framework of research consisted of 5 parts : (1) Fundament of Reliability-Centered Maintenance (RCM) for the machinery that has the data of Time To Fail;
(2) Integrating Mechanical Vibration and Statistical Forecasting Techniques with RCM for the machinery which does not have the data of Time To Fail; (3) Combining Modified Cost Engineering with RCM for both the machinery which has the data of Time To Fail and the machinery that does not; (4) Applying RCM with General Diagnosis of Machine Faults for industrial rotating machinery with Engineering Statistics and Engineering Vibration on each time to forecast the period of maintenance data; (5) Applying RCM with Complex Diagnosis of Machine Faults by Artificial Neural Networks (ANN) and Support Vector Machines (SVM) with signs of vibration. ANN was used to focus the overall error of all data in a classifying process. SVM and ANN have the same working principle but different classifying processes. SVM is better than the Artificial Neural Network (ANN) because SVM rarely has a problem with overfitting.
The results of this research found that Fundament of RCM and Integrating Mechanical Vibration and Statistical Forecasting Techniques with RCM showed that the main time between failures for the plant equipment and the probability of sudden equipment failures were decreased. Moreover, the research results of Combining Modified Cost Engineering with RCM was able to develop a methodology to determine maintenance costs which must be applied to subsets of the elements of a plant, grouped according to their criticality. In addition, the research results of Applying RCM with General Diagnosis of Machine Faults presented an approach for life prediction of machinery such as spindle CNC rolling bearing using a nonlinear regression analysis. Vibration data were also used to assist in the design of rolling bearing fault diagnosis strategies which were effective agents in life prediction and diagnosis. The research results of Applying RCM with Complex Diagnosis of Machine Faults by Artificial Neural Networks (ANN) and Support Vector Machines (SVM) showed that ANN was able to accurately diagnose about 75-80% of automatic motor fault diagnosis based on vibration signals, whereas SVM showed the highest efficiency accurately at 95-100%.