Yoshifuru Saito

Wavelet multi-resolution cross-correlation analysis of swirling gas-solid flow in a horizontal pipe

A wavelet multi-resolution cross-correlation analysis was developed and applied to experimental pressure-time signals in order to analyze the characteristics of swirling gas-solid flow in both Fourier and physical spaces. The experiment was carried out in a horizontal pipe with a length of 7.5 m and an inner diameter of 76 mm. The initial swirl number based on the total inflow was varied from 0.0 to 0.61, the mean gas velocity was varied from 6 to 28 m/s, and the solid mass flow rate was varied from 0.08 to 0.5. From the wavelet multi-resolution correlation analysis of the fluctuating pressure in the range of low air velocity, the characteristics of swirling gas-solid two-phase flows were extracted at various frequencies. Much stronger correlations were found in the range of low frequency, which implied periodic motion of dunes and sliding clusters. Additionally, it was revealed that the motion of a large cluster sliding flow contains two smaller clusters and the moving velocities of dunes were 1 m/s and 2 m/s, respectively. However, no correlation existed at smaller scales of correlation features, which indicated heterogeneous suspension flow.

Particle surface inspection with Fourier-wavelets transform

Products surface inspection has been needed for inferior products classification in, for example, pharmaceutical industry, agriculture and semiconductor industry. Mainly, the inferior particles are produced as chipping, scratching and sticking tablets in the pharmaceutical process. The normal system is only able to classify the inferior particles by means of image processing such as a threshold value method. Defective parts between two particle surface images have been clearly extracted by means of Fourier-wavelets transform methods without threshold values even though the two particles are differently located in a two-dimensional space. This concept is applicable to inferior products classification of a complicated image such as an IC pattern. The method consists of two steps; the first is to acquire the difference between the two particle surface images in Fourier space. The second is to extract the feature of the difference image by means of wavelets transform and multiresolution. The low wavelets level indicates the whole image of the defective part. The high level indicates the outline position of the defective part. This technique contributes to automation of products classification.