A Fabric Defect Valuation Method by Using Expert System

Abstract

Fabric inspector can recognize the fabric defects because they can observe not only the light and darkness but also the brilliance or gloss on the fabric surface. So, in order to distinguish the fabric defects by the machine, it is also necessary to deal with such information as gathered by inspectors. In this paper, a new method has been developed to distinguish the fabric defects. The laser light is irradiated on the fabric surface and then the diffused light is measured cross 360 degrees to create a reflection pattern.
Because there are many naps and twists on the fabric surface irregularly, so the distribution of diffused light will be varied according to the incidence angle of laser. We know that if the light is irradiated on the fabric surface with some suitable angle and the signal is detected at some suitable angle the informations about the fabric defect could be extracted at its maximum. The experiment, has shoun that the most suitable incidence angle is 25 degrees and the most suitable measurement angle is 75 degrees.
By using pinhole and filter in the optical path, we can normally detect the signal even if not in the darkroom. And when the pinhole diameter was 0.7mm, we detected the signal with averaging the ripple of the noises due to the thread naps and twists, and without losing the defect signal. Here the measured area (S) is given by the next equation.
S=πa²/4cosθ
where a is the pinhole diameter, and θ is the measurement angle.
In this experiment, we found that the reflection pattern on th fabric surface was dependent on the fabric construction. The pattern from normal part of the fabric was symmetric, while that from an abnormal part was distorted and the aspect of being distorted was due to the kind of defects. Also the similar reflection patterns can be detected from the same sort defects, but the identical one practically does not appear again even if from the same sort of defect.
To recognize the fabric defect with the reflection pattern, the table reasoning method was introduced. Firstly twelve characteristic parameters were sampled from the reflection pattern to investigate the relations between the refection patterns and the fabric defects, then an expert system has been developed to identify the fabric defects.