Details
| Original language | English |
|---|---|
| Pages (from-to) | 312-317 |
| Number of pages | 6 |
| Journal | Procedia CIRP |
| Volume | 88 |
| Publication status | Published - 13 Jun 2020 |
| Event | 13th CIRP Conference on Intelligent Computation in Manufacturing Engineering, CIRP ICME 2019 - Naples, Italy Duration: 17 Jul 2019 → 19 Jul 2019 |
Abstract
The performance of a process monitoring system is determined by the information available to it. Existing methods for selecting relevant process information (features) work offline with data of faulty processes that is often unavailable or neglect random disturbances. This increases the risk of choosing non-sensitive features. Hence, this paper investigates whether a non-sensitive feature is detectable online in an initial selection of features presumed to be sensitive. A method for quantifying and assessing trends in features online is described. In the validation with turning and drilling processes, a single non-sensitive feature was detected successfully in seven out of eight test cases.
Keywords
- Feature selection, Online, Tool condition monitoring
ASJC Scopus subject areas
- Engineering(all)
- Control and Systems Engineering
- Engineering(all)
- Industrial and Manufacturing Engineering
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Procedia CIRP, Vol. 88, 13.06.2020, p. 312-317.
Research output: Contribution to journal › Conference article › Research › peer review
}
TY - JOUR
T1 - Wear curve based online feature assessment for tool condition monitoring
AU - Denkena, Berend
AU - Bergmann, Benjamin
AU - Stiehl, Tobias H.
N1 - Funding information: This research was funded by the Deutsche Forschungsge-meinschaft (DFG, German Research Foundation) – 313912117.
PY - 2020/6/13
Y1 - 2020/6/13
N2 - The performance of a process monitoring system is determined by the information available to it. Existing methods for selecting relevant process information (features) work offline with data of faulty processes that is often unavailable or neglect random disturbances. This increases the risk of choosing non-sensitive features. Hence, this paper investigates whether a non-sensitive feature is detectable online in an initial selection of features presumed to be sensitive. A method for quantifying and assessing trends in features online is described. In the validation with turning and drilling processes, a single non-sensitive feature was detected successfully in seven out of eight test cases.
AB - The performance of a process monitoring system is determined by the information available to it. Existing methods for selecting relevant process information (features) work offline with data of faulty processes that is often unavailable or neglect random disturbances. This increases the risk of choosing non-sensitive features. Hence, this paper investigates whether a non-sensitive feature is detectable online in an initial selection of features presumed to be sensitive. A method for quantifying and assessing trends in features online is described. In the validation with turning and drilling processes, a single non-sensitive feature was detected successfully in seven out of eight test cases.
KW - Feature selection
KW - Online
KW - Tool condition monitoring
UR - http://www.scopus.com/inward/record.url?scp=85089079028&partnerID=8YFLogxK
U2 - 10.1016/j.procir.2020.05.054
DO - 10.1016/j.procir.2020.05.054
M3 - Conference article
AN - SCOPUS:85089079028
VL - 88
SP - 312
EP - 317
JO - Procedia CIRP
JF - Procedia CIRP
SN - 2212-8271
T2 - 13th CIRP Conference on Intelligent Computation in Manufacturing Engineering, CIRP ICME 2019
Y2 - 17 July 2019 through 19 July 2019
ER -