Monitoring of punch failure in micro-piercing process based on vibratory signal and logistic regression
| Autor: | Tsung-Liang Wu, Bor-Tsuen Lin, Delima Yanti Sari, Chia-Wei Chang |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
0209 industrial biotechnology
Engineering business.industry Mechanical Engineering Process (computing) Experimental data 02 engineering and technology Logistic regression computer.software_genre Stability (probability) Signal Industrial and Manufacturing Engineering Computer Science Applications 020303 mechanical engineering & transports 020901 industrial engineering & automation 0203 mechanical engineering Breakage Control and Systems Engineering Frequency domain Statistics Feature (machine learning) Data mining business computer Software |
| Zdroj: | The International Journal of Advanced Manufacturing Technology. 93:2447-2458 |
| ISSN: | 1433-3015 0268-3768 |
| Popis: | Tool online monitoring plays an important role in preserving the stability of the process and detecting the process anomalies. By early detection of tool failure, the corrective action can be taken to avoid subsequent damage and reduce scrap. In this present study, a vibration-based monitoring method for online detection of punch failure in the micro-piercing process is presented. The signal features are extracted from both time and frequency domain signal. The statistical overlap factor is utilized to select the best features. To evaluate the tool condition, a logistic regression model is used with the selected features as an input. The experimental data with three different clearances, i.e., 3, 5, and 9%, are used to examine the applicability of the proposed method. The logistic models are established by using the selected features separately and by incorporating all the features in the model. The results show that the developed logistic models can be used to estimate the state of tool condition for experiment with different clearances. The drop or fluctuation of the probability value before the punch breakage signifies the deterioration of tool condition which leads to the punch breakage. Thus, the breakage indication or the failure condition can be detected. Among the developed logistic models, the models which use the selected feature separately give better accuracy, i.e., over 99%. The validation of the model indicates that the proposed method is applicable and potential in the monitoring of the micro-piercing process. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|