Abstract: The comprehensive quality evaluation of 3D CAD software needs to consider several types of indexes such as function, performance, compatibility, usability, and so on. The sample data of various indexes are acquired by different methods, leading to different sample sizes and different uncertainties. The existing methods for evaluating software quality neither fully consider the characteristics and uncertainties of index data in the mathematical modeling of various indexes nor assign the weights of indexes according to the uncertainties and credibility of their index data, and thus can hardly ensure the credibility of comprehensive quality evaluation results. Therefore, a comprehensive quality evaluation method considering multi-source uncertainties is proposed for 3D CAD software. The systematic quality indexes are determined according to the characteristics of 3D CAD software. Considering the uncertainties of software performance itself and the unstable interference of the test environment contained in the abundant sample data of performance indexes obtained from massive repeated automated tests, a self-subtractive reverse cloud generator is proposed to subdivide the two types of uncertainties and avoid the probable failure for the existing reverse cloud generators in calculating the hyper entropy. With the full consideration of the subjective preference and the global representativeness in the limited sample data of usability indexes, a mathematical modeling method for usability indexes is proposed based on the extended interval number. On these bases, a unified representation of multi-type quality indexes based on the cloud model is established for 3D CAD software. By considering the credibility of each index, the variable weight analytic hierarchy analysis is implemented to achieve the reliable comprehensive quality evaluation results of 3D CAD software.

Key words: 3D CAD software, quality evaluation, uncertainty, cloud model, extended interval number, credibility