Optimized ultrasonic assisted flavonoids extraction from an annual desert plant of Agriophyllum squarrosum and its implications for natural antimicrobial agent

The strength characteristics of ice materials are crucial for the analysis of the interaction between ice and structure in ocean engineering and ice engineering. In this investigation, six machine learning methods were utilized to predict the strength of the envelope surface of ice materials. The database for the ice strength was first established by collecting 1,481 testing data reported in the previous literatures. A quadric strength criterion was adopted to describe failure behaviors of ice materials under different conditions of material property and laboratory. Three model parameters in this strength criterion were forecasted by using six machine learning methods. The prediction capacities of six machine learning methods were evaluated by three statics indices, and the integrated simulation ability of six machine learning methods was arranged. Three machine learning algorithms were selected to be improved and optimized, and the simulation capacity of the three algorithms was further explored. The optimization results indicate that the improved potential of the Ensemble algorithm is much higher than that of the SVM algorithm and the GPR algorithm for predicting the ice strength. • The database for the ice strength was established by collecting 1481 testing data reported in the previous literatures to investigation strength characteristics and damage mechanisms under different conditions. • Six machine learning methods were utilized and extended in the prediction of the strength envelope surface of ice materials under different conditions of material property and laboratory. • The prediction abilities of six machine learning algorithms were assessed by three important statistic indicators. • The optimization procedures were respectively conducted on three algorithms to further probe their simulating and evaluating capability.