The Improvement Of Differential Evolution Algorithm And Its Application In 2.5D Integrated Circuit Test

Optimization issues are ubiquitous in scientific research and engineering applications and many researchers need powerful optimization algorithms to solve these issues that are critical to their daily working life.As a relatively new global optimization algorithm,differential evolution algorithm is based on the principle of "survival of the fittest,survival of the fittest" as the guiding ideology to imitate the process of biological evolution for algorithm construction.Since its introduction in 1995,the differential evolution algorithm has become one of the hottest research directions in the field of artificial intelligence today due to its simple,effective,robustness,and other advantages.The interposer based 2.5-dimensional integrated circuit,as the practical application object of the propsed improved algorithm in this paper,overcomes As the practical application object of the improved algorithm in this paper,the interposer-based 2.5-D integrated circuit with through silicon via structure to overcome the problem of delay and power consumption of internal connection in integrated circuits.It is important to ensure the reliability of 2.5-D integrated circuit and its wide application to design and solve the testing cost.In this paper,based on the introduction of the basic mechanism and principle of differential evolution algorithm,the improvement of the algorithm is studied according to the advantages and disadvantages of the control parameters,the variation process,and the crossover process.The improved algorithm is applied to test optimization plan design of 2.5D integrated circuit.The main research contents are as follows:1.Based on the analysis of the classical differential evolution algorithm's bivariate cross population diversity and low optimization inefficiencies,an improved adaptive differential evolution algorithm based on multi-angle search crossover strategy is proposed.This not only effectively overcomes the shortcomings of the classic DE crossover operation,but also increases the diversity of the population and greatly improves the optimization efficiency.The experiment s on the CEC2013 international standard test dataset verifies the versatility and superiority of the rotating crossover strategy proposed in this paper.The performances of the improved algorithms are improved by 57%~96% compared to original algorithms.2.For the classical DE mutation operator mutation parameter selection sensitivity and differential operator search step larger defects,from the perspective of enhanced algorithm local search ability to propose an adaptive mutation operation based on the individual adaptation value difference and distance to replace the original mutation Factors and differential operations enhance local search capabilities while eliminating the effects of parameter sele ction on optimization results.At the same time,based on the AP clustering algorithm,the automatic classification and division operation of the population enables each sub-population to operate in parallel and increase the optimization speed of the popul ation.Experiments on the CEC13 test set show that the optimization efficiency of the optimization algorithm can be rised dramatically.What's more,we also combine the two improved algorithms to verify the experimental results,the experimental results show that the combined algorithm is better than the original algorithm,thus demonstrating the universality and effectiveness of the two algorithms.The performances of the improved algorithms are improved by 50%~88% compared to original algorithms.3.Finally,this paper studies the test package scan chain balance design and system-level test scheduling method for 2.5-dimensional SOC.According to the size and arrangement information of each IP core,combined with constraints such as power consumption and hardware overhead,a 2.5-D SOC test scheduling model was established,and the improved differential evolution algorithm proposed in this paper was used to solve the test to determine the test of each IP core.Order and grouping.Experiments on the ITC'02 standard test set show that the test scheduling method proposed in this paper can effectively reduce the test time of 2.5-dimensional SOCs under constraints.At the same time,adopting the improved differential evolution algorithm for the two-balanced design of the scan can achieve a compromise between the test time cost and the hardware cost,and obtain a more balanced test package scan chain.