Double-loss value network deep reinforcement learning KVFD model mechanical parameter global optimization method and system

Abstract

The invention discloses a double-loss value network deep reinforcement learning KVFD model mechanical parameter global optimization method and system, and the method comprises the following steps: S1, inputting a pre-obtained nanoindentation measurement curve into a trained prediction value obtaining network, and obtaining a parameter prediction value of the nanoindentation measurement curve; and s2, performing iteration by taking the parameter prediction value as an iteration initial value of a deep reinforcement learning algorithm to obtain approximation of a global parameter solution of a pre-obtained nanoindentation measurement curve; and when the approximation of the global parameter solution reaches a preset convergence condition, outputting the approximation of the global parameter solution as a mechanical parameter of the KVFD model. According to the method, the parameter prediction value is introduced in iteration for parameter guidance, and the global optimal solution can be better approached.

Description

technical field [0001] The invention belongs to the technical field of mechanical parameters of nano-indentation measurement data, relates to the field of KVFD model multi-parameter function fitting and global parameter approximation, and in particular relates to a method and system for global optimization of mechanical parameters of a KVFD model with double-loss value network deep reinforcement learning. Background technique [0002] At present, in the process of obtaining the mechanical parameters of the measured material through the nanoindentation measurement data, for simple function fitting, the least square method is mostly used, and the function parameters are adjusted iteratively one by one to reduce the least mean square between the fitted curve and the real curve. error. This method is fast and effective for fitting simple functions, but it often performs poorly for complex functions and multi-parameter functions. [0003] For the multi-parameter optimization of ...

AI Technical Summary

Problems solved by technology

This method is fast and effective for fitting simple functions, but it often performs poorly for complex functions and multi-parameter functions.

[0003] For the multi-parameter optimization of complex functions of the KVFD model, the common greedy algorithm, gradient descent algorithm, and simulated annealing algorithm cannot obtain a better global optimal solution; among them, the greedy algorithm and gradient descent algorithm can find The local optimal solution of , has certain applicability to the multi-parameter optimization of complex functions, but whether to find the global parameters has a lot to do with the given initial value, and it is difficult to approach the global parameters; the simulated annealing algorithm adopts a certain probability to accept new The parameter solution of , has the ability to jump out of the local optimal trap. For optimization problems with many local optimal solutions, it shows better optimization capabilities. There is a certain probability that a solution near the global parameter solution can be found, but simulated annealing The probabilistic nature of the algorithm also makes it unable to approach the global parameter solution every time, and the reliability is poor

[0004] To sum up, at present, for the KVFD model complex function multi-parameter optimization problem, it is difficult to effectively approach the global optimal solution

Images