The Drilling Quality Inspection Of The Multi-holes Part Based On Features Fusion Of Multi-sensor Signals

The drilling quality consistency is key factors to measure and ensure the hole parts processing quality and work performance. In this dissertation, to realizing the consistency detection of holes drilling quality, we conducted studies on drilling experiment platform establishment, singularity detection and extraction of multi-sensors signals, feature extraction, feature fusion, pattern recognition and others.(1) Drilling experiment platform establishment and signal acquisition. According to the characteristics of drilling, the drilling experimental platform was established. In this experiment, hall current sensor was used to monitor the variation of motor spindle load, acoustic emission sensor was devoted to monitor the work-piece material fracture and deformation, and the three-axis vibration acceleration sensors were applied to monitor vibration of work-piece.(2) The preprocessing and singularity detection of monitoring signals. First, spectral subtraction method was adopted to remove noise. Then, according to the variations of the drilling force and spindle power signals when the drill bit touched the work-piece and left the work-pieces, the catastrophe points were extracted with the RMS analysis and variation rate calculation of spindle power signals. Finally, according to these points, the signals which associated with the drilling quality were split out from the all sensor signals. These signals were the foundation for the drilling quality analysis.(3) The features extraction and analysis of monitoring signal. Because drilling monitoring signals always have the characteristics of non-linear and non-Gaussian and non-stationary, Hilbert Huang transform and high-order spectrum were employed to extract the marginal spectrum and bi-spectrum feature of all holes’ drilling monitoring signals. And then analysis of the organic relationship between change of each hole’s features and drilling quality were carried out in this chapter.(4) The feature fusion and clustering based on principal component analysis. The marginal spectrum and bi-spectrum feature were employed to build feature matrix, and then principal component analysis was used to reduce the feature matrix dimension. Afterward, according to the first and the second principal component of feature matrix, k-means clustering is employed to obtaining the holes drilling quality distribution, and is compared with the artificial detection results to verify the effectiveness of clustering results. Finally, other drilling experiment with changing the parameters of drilling was applied to verify the validity of this method.The analysis and experimental validation results showed that, the study method can effectively realize holes drilling quality consistency assessment, rapid analysis and discern the abnormal hole with bad drilling quality. The method effectively overcomes the missed detection of drilling quality and so on. The method provides a theoretical basis for realizing the drilling quality consistency detection and control.