Research On Intelligent Optimization Algorithm For Full-Link Intelligent Manufacturing

With the rapid development of new-generation information technologies such as artificial intelligence,the Internet of Things,and big data,intelligent manufacturing has become a crucial driver for the transformation and upgrading of the manufacturing industry.Intelligent optimization algorithms are widely applied in various complex optimization problems in the field of intelligent manufacturing due to their high flexibility,strong robustness,and excellent global search ability.This paper focuses on three key problems in different stages of the fulllink intelligent manufacturing,including the Supply Chain Network Design(SCND)within the smart supply chain,resource allocation and transportation optimization within smart logistics,and Fourth-party Logistics Routing Problem(4PLRP)and product quality prediction within intelligent manufacturing.These optimization problems have complex constraints,large data sizes and high requirements for solving speed,making them difficult to solve by traditional optimization methods.According to the characteristics of these problems,this paper proposes different intelligent optimization algorithms to effectively solve them.The specific research content includes the following three aspects:1)A closed-loop supply chain model is designed and modelled with consideration for generalizability,aiming to match a wider range of application scenarios and maximize corporate profits by optimizing transport routes and product delivery volumes in the supply chain network.In order to efficiently solve this model,a Genetic Algorithm with Two-step Rank-based Encoding(GA-TRE)was proposed.The first stage encoding of GA-TRE is used for planning feasible transportation routes,and predicting their feasibility based on relevant constraint conditions.The second stage encoding uses a greedy method to rank the routes and make decisions on the corresponding delivery volumes.Simultaneously,genetic operators are improved by incorporating the characteristics of the SCND problem and the encoding method of GA-TRE.Additionally,an adaptive population disturbance method is proposed to enhance population diversity.2)A 4PLRP model is designed and modelled which considers multiple transportation modes and unfixed quantities of third-party logistics providers.The proposed model also considers the corresponding delivery distance and speed for different transportation modes.To effectively solve the model,a hybrid Iterative Local Search Ant Colony Optimization(ILSACO)algorithm was proposed.Firstly,based on the characteristics of the proposed 4PLRP model,the heuristic information and pheromone update rules are improved in the ACO algorithm.These improvements enable the algorithm to consider transportation costs,distance,and transportation modes,making it more suitable for solving such problems.Secondly,based on the ILS optimization algorithm framework,three different local search operators are proposed to address the different problems and enhance the global search capability of the ILSACO.Finally,a bidirectional repair method is proposed to improve the efficiency of ILS-ACO by repairing illegal solutions.3)An Evolutionary Strategy-based Neural Architecture Search(ESNAS)algorithm is designed to predict product quality in the steel industry.The ESNAS algorithm adopts a variable-length encoding strategy to simultaneously optimize the structure of the neural network and its hyperparameters.The ESNAS algorithm improves the exploration capability of the classical(μ+λ)-ES algorithm by using various types of mutation operators.Additionally,after performing crossover and mutation operations,the selection operation is used to screen out individuals with inferior fitness,and an elite strategy is used to preserve highquality individuals in the parent population.