Abstract: Six widely-used methods for stability analysis of simulated data with multi-morphological features and different levels of stability were theoretically compared, and the adaptability for statistical assessment was further studied to suggest optimal selection of these methods, which include range method, variance method, deviation degree method, p-value method, information entropy, and CV method. After summarizing the five different cases of multi-morphological data, an integrated strategy was proposed to simulate the variation of these data and then strictly evaluate the objectivity of these methods used for stability analysis. It shows that the range method, deviation degree method and p-value method are not suitable for the stability analysis of industrial production process data with high complexity. In contrast to these three methods, the performance of variance method and information entropy method is improved for the analysis of some types of data. Fortunately, the CV method has high robustness and wide adaptability to all the five cases of multi-morphological industrial data. The analysis of real data obtained from cigarette processing process further validate the results and conclusions introduced above. The results reported in this study provide an efficient way for stability analysis of industrial data with high complexity.

Key words: industrial production process, data stability analysis, adaptability study, coefficient of variation, cigarette production process