Open Access Open Access  Restricted Access Subscription or Fee Access

Research on Fault Feature Extraction Method Based on SVM of Printing Machine

YI-MING WANG, SHU-QIN WU, WEN-CAI XU, CHENG-WEN CHAI, XIN QIAO

Abstract


With the increase of the printing machine speed, once the printing failure is not solved in time, it will affect the normal production and cause a lot of waste. The overprinting fault is one of the common faults of various printing machines, and the modularization and complexity of various color groups make the fault depend mainly on the experience of the people. A fault feature extraction and diagnosis method based on One-class Support Vector Machines (One-class SVM) is proposed in this paper. According to the characteristics of typical faults of printing machines and three kinds of information obtained by built-in sensors, external sensors and printed materials, established typical fault characteristic parameter sets of printing machines. The One-class SVM fault diagnosis model of printing machine is established by One-class SVM method. According to the nonlinear and separable feature of printing machine faults, the kernel function suitable for typical faults of printing machines is determined. Taking a multi color lithography machine as an example, the dynamic signals are obtained by field test in view of the inconsistency of registration accuracy in different color sets. Through signal feature extraction and mesh parameter optimization, the recognition and classification of overprinting faults between color groups in modular multi-color printing machine are completed. The research shows that, according to the dynamic information collected by external sensors in the field, the fault diagnosis One-class SVM model of the printing machine is put forward to realize the identification of the overprinting faults in the modular color group. The method can be used for remote condition monitoring and fault diagnosis, and is suitable for most types of presses.

Keywords


One-class Support Vector Machine (One-class SVM);fault feature extraction; vibration test; printing machineText


DOI
10.12783/iapri2018/24434

Full Text:

PDF

Refbacks

  • There are currently no refbacks.