A Study On The Building And Applications Of Automatic Appraise System For Surface Quality Of Steel Balls

At present, the methods of traditional and artificial detection to steel balls' surface quality are still used at bearing production fields in our country. Its quality inspection doesn't parallel to the design standard of steel balls in condi-tions where the temperature is high and the velocity is faster than the average. With the wide application of digital image processing, pattern recognition and the gradual enhancement of economy benefit, especially the rapid improvement of apparatus characters such as CCD, there is an urgent need to employ automatic checkout equipment to evaluate surface quality of the steel balls correctly and efficiently. In this topic, from the following several aspects, we want to study the theory and application technology about the automated decision system of steel balls surface. Above all, the first step is about the choice to the pick-up camera used in digitizing image field and image card. Then we explore the structure feature and operating principle of expanding wheel of steel balls produced at SOMET in Czechoslovakia. Through the force analysis of the checked steel ball, we see that the steel ball occur side-rolling motion under the action of asymmetry side expanding wheel. This demonstrates the feasibility that the expanding wheel with such structure can expand steel balls surface to acquire the information about the whole surface of steel balls. During the process of this experiment, considering the complexity of optional drive unit of expanding wheel, we also design a kind of simple structure called rotating table for steel balls to substitute expanding wheel and come up with a experiment method about sampling system of steel balls whose resolution are marked in the experiment. We also combine the technology of digital image processing and pattern recognition with automatic inspection to steel balls' surface flaw and disposal to identification ways. Following this, a reasonable gray-scale threshold, which is determined by the method of double windows based on Gauss function, can clearly separate defect image from background to achieve binary algorithm. On the base of data preprocessing, we can do automatic tracking to the boundary of defect region, restoration and marking to the image of defect region, extraction to defect feature and identification to defect. The experiment demonstrates that adopting binary possesses the advantage of rapid calculating and reliable recognition. The scheme's excellent performance shows its more practicability. Utilizing stepwise regression, we extract two texture features, which express the steel balls' surface quality: angular second moment and entropy features, and explain their physical meaning. Traditional ways of characteristic extraction often choose a certain number of the best character by intuition, physical meaning of characteristic root or predecessor's experience .In contrast, we adopt explainable method to conduct the choose to character, thus it is more persuasive. We come up with a mathematical model forecasting vibration values of steel balls. Analysis and emulation experiment show that fit precision of this square non-linear prediction model is higher. The building of mathematical model establish a solid theoretical foundation on which the automatic identification to steel balls' precision can be done in computer and also make it possible for computer to identify the precision of steel balls automatically. First, we introduce the system construction of the applications software developed by myself and illustrate the usage of the checking software of the surface quality of the steel balls,then we verify the correctness and reliability of the theory algorithm proposed by author and the applications software using contrast experiment between checking by man and by computer. Through this contrast experiment, we draw the conclusion that the result checked by computer is better that that of sampling statistical checked widely adopted in current bearing industry. So we can say that the method of defect identification and grade classification of steel balls proposed by author is important to the application of the no-defect check of steel balls in bearing industry.