|
ABSTRACT: THIS PAPER DESCRIBES AND ALSO ANALYZES A MODEL THAT, WITH THE USE OF ARTIFICIAL INTELLIGENCE TOOLS, MORE SPECIFICALLY, OF PATTERN RECOGNITION TECHNIQUES, DETERMINE, IN AUTOMATIC WAY, BASIC SPECIFICATIONS OF INDUSTRIAL PIECE DESIGN. THE CONTROL OF THESE SPECIFICATIONS IS MADE BY THE ANALYSIS OF THE VARIATIONS ALONG THE USE OF THE PIECE (AS THE DEFECT OCCURRENCE IN THE PIECE BECAUSE IT SHOWS STAINS, FOR EXAMPLE). THE MODEL WORKS ON THE IMAGES OF THE PIECE UNDER STUDY. USING PATTERN RECOGNITION TECHNIQUES, THE MODEL ASSOCIATES, TO EACH COLLECTED IMAGE, A SIMPLES MATRIX STRUCTURE. CONSIDERING THE PROPERTIES OF THESE IMAGES, REPRESENTED BY SPECIFIC CHARACTERISTICS OF THE IMAGE REPRESENTATION STRUCTURE, IN THIS CASE, A MATRIX, THE MODEL DETERMINES PARAMETERS THAT CONTROL, IN A CONTINUOUS EVALUATION, THE VARIATIONS OF A GIVEN PROPERTY OF THE PIECE (AS THE COLOR OF A DRAWING IN A TILE, FOR EXAMPLE). THE MODEL CHECKS CHARACTERISTIC OF THE PIECE, MATCHING THEM TO SOME PATTERNS. THEN IT DECIDES IF THE VARIATIONS OBSERVED IN THE PIECE UNDER STUDY ARE CONSIDERED ACCEPTABLE, OR NOT. THE MODEL DETERMINES REFERENCE VALUES THAT SHOULD BE USED TO EVALUATE SUCH VARIATIONS. THE PRESENT PAPER DESCRIBES THE MODEL AND THE MOST USUAL SITUATIONS FOR ITS USE. THE PAPER DISCUSSES ALSO PRACTICAL CASES WHERE IT WAS USED AND A CRITICAL EVALUATION OF THE OBTAINED RESULTS IS DONE.
KEY-WORDS: PATTERN RECOGNITION, PRODUCT DESIGN, INDUSTRIAL QUALITY. |
