New Approach for Fault Identification using Residual-based Fault Diagnosis Mr. Haedong JEONG, Bumsoo PARK, Seungtae PARK and Seungchul LEE (UNIST)
Manufacturing machineries are becoming more complicated, and breakdowns of machinery are related to not only reduced efficiency but also safety issues. Due to the needs for reliability within facility operation, various methods for maintenance are suggested as the importance of monitoring systems has increased. Among the various maintenance techniques, in this paper, a model-based fault detection and isolation (FDI) technique for the diagnosis of the machine state is introduced. In particular, we suggest new approach for the fault identification that extracts the information of the fault mode such as the magnitude or shape of the fault. The main idea of new methodology is defining the relation between fault mode and observer theory. For the proposed method, a numerical simulation is conducted to show the effectiveness of the fault detection and identification. Then a comparison between the proposed method and data-driven method is implemented to show the difference in aspect for fault diagnosis.
Graph Partition based on Dimensionless Similarity and Its Application to Fault Diagnosis Dr. Bo ZHENG, Hong-Zhong HUANG, Jie ZHOU and Yan-Feng LI (University of Electronic Science and Technology of China)
To improve the efficiency of fault diagnosis, a novel granular computing algorithm is developed to reduce computational cost. It is realized by extracting and partitioning on the complete graphs, and in the process of graph generation, the dimensionless similarity method is proposed to overcome the influence of attributes with different dimensions. The algorithm is named graph partition based on dimensionless similarity (GPDS). Moreover, similarity threshold determination method based on frequency distribution histogram is proposed to reduce the dependency on the experiences of experts. Meanwhile, a weighted relative error is proposed to measure quantitatively the distribution change of original data after being compressed. Finally, different characteristic data have been applied to verify the theories, and the experimental results have shown that the compressed training samples can maintain the classification accuracy. Furthermore, the elapsed time can be obviously reduced. Therefore, the GPDS method can be used in fault diagnosis properly.
Fatigue Life Prediction Based on Walker and Masson Models Mr. Jie ZHOU, Hong-Zhong HUANG, Bo ZHENG and Zhaochun PENG (University of Electronic Science and Technology of China)
It is known that mean stress has significant effects on fatigue life prediction, and various modifications have been developed to explain the mean stress effect, yet seldom accounting for mean stress sensitivity. The Smith-Watson-Topper (SWT) model is one of the most widely used models that can give satisfactory predictions, and it is viewed as a particular case of Walker model when the material parameter γ equals to 0.5. The Walker equation takes both the mean stress effect and sensitivity into count and can give accurate predictions in many fatigue programs. In this paper, a modified model accounting for the mean stress effect and sensitivity is proposed to estimate the fatigue life based on the Walker model and Masson Model. Three set of experimental data are used to validate the applicability of the proposed model. A comparison with the SWT model and Morrow model is also made. The results show that the proposed model has more accurate predictions than the others.