Design Of Surrogate-based Optimization Platform With Python And Its Application

In engineering problems,Optimization design involves various fields.But traditional simulation optimization always needs long period and high cost,which is intolerable.Surrogate model that is based on small data set is a mathematical model,which is time-consuming and low cost.With a satisfactory precision accuracy,the design cycle of products is greatly shortened,it provides a reliable and fast methods for subsequent optimization and related analysis,making it widely be used in various fields.Because surrogate model is a mathematical model,the traditional code files are difficult for non-specialist users to understand and limit application for surrogate model by nonspecialist users.Although existing commercial MDO software have been widely used,these optimized software must be spent too much time to study and have complex interface,which is disadvantage of non-specialist users to apply surrogate model to solve problems in a short time.Based on the python surrogate model to build a general,concise optimization platform that suitable for non-specialist users in this paper.The main contents of this paper is summarized as below.Firstly,the optimized platform is divided into some modules: DOE modules,Surrogate model modules,Optimization modules.Then based on the Python language written design of experiment,surrogate model algorithm,optimization algorithm,analyze the required parameters of all code files.In-depth study of PyQt's basic properties,learning about the use methods of widgets,and select the type of widget required for each module.Secondly,the software platform requirements are analyzed,and the overall layout of the software is made according to the above selected parameters and the characteristics of each module.In each module,the functional area is specifically divided to make the interface as simple as possible under the premise of ensuring clear logic.Then in each modules apply PyQt to layout the widgets at the function block,the code written Python language implement each function of each function block,such as 3D map,the data of model accuracy,optimization results,etc.Then,a standard test function library is built to test the correctness of the platform's code and the accuracy of the data stream transmission.In order to compare the correctness between platform and code,some two-dimensional functions are selected for verification,and dozens of standard functions are selected in the test function library for build models,and the model precisions of software platform modeling and code modeling are compared.Make sure the software platform adapts to multidimensional problems and guarantees basic stability.Finally,for the verified of software platform based on surrogate model.A numerical optimization problem and an actual engineering problem are applied to evaluate the performance of the platform.The numerical example is a linear constrained optimization example,which is use to compare the deviation between the results of different surrogate models and the real result.The engineering example is to optimize the laser cutting parameters by used this platform,from design of experiment to optimization,to verify that this platform can be used to solve practical problems.