Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance

CR Parikh, MJ Pont, Yuhua Li, NB Jones

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    This paper focuses on the development of neural-based condition-monitoring and fault-diagnosis (CMFD) systems. Specifically, we consider the impact of the limited availability of `faulty' training data in real CMFD applications. Where limited data are available we demonstrate two ways in which performance may, in some circumstances, be improved: (1) by using fewer training data made up of roughly equal numbers of,normal' and `fault' samples; or (2) by using a `duplicate-data' training algorithm.
    LanguageEnglish
    Title of host publicationUnknown Host Publication
    Pages237-243
    Number of pages7
    Publication statusPublished - 1999
    EventCONDITION MONITORING `99, PROCEEDINGS -
    Duration: 1 Jan 1999 → …

    Conference

    ConferenceCONDITION MONITORING `99, PROCEEDINGS
    Period1/01/99 → …

    Fingerprint

    Condition monitoring
    Failure analysis
    Classifiers
    Neural networks
    Availability

    Keywords

    • neural networks
    • condition monitoring
    • fault diagnosis
    • software design

    Cite this

    Parikh, CR., Pont, MJ., Li, Y., & Jones, NB. (1999). Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance. In Unknown Host Publication (pp. 237-243)
    Parikh, CR ; Pont, MJ ; Li, Yuhua ; Jones, NB. / Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance. Unknown Host Publication. 1999. pp. 237-243
    @inproceedings{799a0ef055f04cf594375e75d1e9ceec,
    title = "Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance",
    abstract = "This paper focuses on the development of neural-based condition-monitoring and fault-diagnosis (CMFD) systems. Specifically, we consider the impact of the limited availability of `faulty' training data in real CMFD applications. Where limited data are available we demonstrate two ways in which performance may, in some circumstances, be improved: (1) by using fewer training data made up of roughly equal numbers of,normal' and `fault' samples; or (2) by using a `duplicate-data' training algorithm.",
    keywords = "neural networks, condition monitoring, fault diagnosis, software design",
    author = "CR Parikh and MJ Pont and Yuhua Li and NB Jones",
    note = "International Conference on Condition Monitoring, SWANSA, WALES, APR 12-15, 1999",
    year = "1999",
    language = "English",
    pages = "237--243",
    booktitle = "Unknown Host Publication",

    }

    Parikh, CR, Pont, MJ, Li, Y & Jones, NB 1999, Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance. in Unknown Host Publication. pp. 237-243, CONDITION MONITORING `99, PROCEEDINGS, 1/01/99.

    Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance. / Parikh, CR; Pont, MJ; Li, Yuhua; Jones, NB.

    Unknown Host Publication. 1999. p. 237-243.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    TY - GEN

    T1 - Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance

    AU - Parikh, CR

    AU - Pont, MJ

    AU - Li, Yuhua

    AU - Jones, NB

    N1 - International Conference on Condition Monitoring, SWANSA, WALES, APR 12-15, 1999

    PY - 1999

    Y1 - 1999

    N2 - This paper focuses on the development of neural-based condition-monitoring and fault-diagnosis (CMFD) systems. Specifically, we consider the impact of the limited availability of `faulty' training data in real CMFD applications. Where limited data are available we demonstrate two ways in which performance may, in some circumstances, be improved: (1) by using fewer training data made up of roughly equal numbers of,normal' and `fault' samples; or (2) by using a `duplicate-data' training algorithm.

    AB - This paper focuses on the development of neural-based condition-monitoring and fault-diagnosis (CMFD) systems. Specifically, we consider the impact of the limited availability of `faulty' training data in real CMFD applications. Where limited data are available we demonstrate two ways in which performance may, in some circumstances, be improved: (1) by using fewer training data made up of roughly equal numbers of,normal' and `fault' samples; or (2) by using a `duplicate-data' training algorithm.

    KW - neural networks

    KW - condition monitoring

    KW - fault diagnosis

    KW - software design

    M3 - Conference contribution

    SP - 237

    EP - 243

    BT - Unknown Host Publication

    ER -

    Parikh CR, Pont MJ, Li Y, Jones NB. Neural networks for condition monitoring and fault diagnosis: the effect of training data on classifier performance. In Unknown Host Publication. 1999. p. 237-243