Tension is very closely related with fabric inspection quality, as not well controlled tension of gray cloth will lead to stretch, relax or fold of gray cloth, so that no sharpest image can be taken. Now that gray cloth tension and convey speed are related with taking sharp image, so this study attempts to design a gray cloth tension control module, develop the intelligent online dynamic gray cloth defect automatic detection system. Gray cloth tension control module makes direct regulation of structural tension of feed and wind rolls and conveys speed control module by load cell coupled with A/D, D/A signal conversion and two sets of inverters. This study utilized fuzzy control theory to design controller so as to keep surface tension consistency of gray cloth, improve recognition rate of gray cloth defect, so that the system can have high action efficiency. In previous fuzzy controller designs, membership functions were often designed by means of trial-and-error method, which usually cost much time. This study used Taguchi method to make membership function programming, and made main effect analysis to choose a group of optimal membership function combination. Through systematized, efficient experimental design, tension controller designed in this study could stabilize gray cloth tension very soon, and acquire sharp gray cloth image.