Application Of Knowledge Based Self Organizing Algorithm In Discrete Processing With Multiple Product And Small Batches

Hybrid manufacturing of rods, plates and other different types steel belong to the typical of discrete processing with multiple product and small batches. Hybrid manufacturing process of different types of steel requires switching process frequently, resulting in complex logistics and information flow of manufacturing system and high coupling of every product chain.As for the order allocation scheme design problems of most typical variable cross line in the production process, the traditional control algorithm is difficult to achieve satisfactory results. Currently, it needs to rely on the long-term experience of experts to make solutions, however manual ways exist inevitably alternatives less and the key performance index are difficult to count exactly. Therefore, the study on how to develop integrated production planning in iron and steel enterprises is of great significance to the realization production target of reducing energy consumption and improve the economic benefits.In view of the above problems, this paper proposes self-organizing algorithm(SOA) based on domain knowledge, and in the form of rules introduce expert knowledge into the model deduction and the candidate solutions in the search process, effectively solving the problem of steel processing and orders assignment. Specifically, the paper analyzed characters of steelmaking, continuous casting and rolling three production stages, established process model from smelting to rolling, realized the simulation deduction of any orders related to the logistics flows situation in the production. The constraints from the coupling relationship of devices and orders conflict lead to the feasible solution space division and the design of fitness function is difficult to conduct and also difficult to be mapped to the standard combinatorial optimization problems, thus introducing self-organizing optimization algorithm. Using the self-organized critical state to find the coupling and conflict contribute most of the variables in the solutions as an improved object to improve the value of the variable by the neighborhood search method, so as to enhance the overall fitness of the current solution. In order to enhance the efficiency of neighborhood search, introducing the manual experience of experts according to the business characteristics of steel production process and restricting the neighborhood construction direction through the development of rules make sure the number in violation of constraints monotonically decreasing during the search process.The simulation results show that the application scenarios difficult to quantify as a penalty term and added the fitness function, genetic algorithms and other traditional algorithms mutation strategy can’t distinguish viable and non-viable solution, however self-organizing optimization algorithm can make use of expert knowledge based on knowledge to construct neighborhood and find the viable solution and optimization. Refer to the production plan problems of the iron and steel enterprises to formulate strategies, combined with the existing expert knowledge and self-organizing algorithm, providing multiple alternatives meet the key performance indicators, and then bringing a new method for the model solution which is difficult to be mapped to the standard combinatorial optimization problems.