Process Industry Oriented Raw Material Supply Programming Model With Intelligent Decision And Its Application

The research of raw material supply is very significant for process plants because of the vital importance of supply reliability for continuous production and the enormous financial cost on raw material supply activity. The optimal decision in the activity will bring obvious benefit and save great cost for process plants. However, the decision making for raw material supplying activity is challenged by, the first, various supply sources of diverse distributions and different qualities and the second, the risks of uncertainty in the environment which is not very appropriate for agile manufacture pattern effective in discrete manufacturing industry. This thesis focuses on modeling and optimal decision-making in inventory and purchasing process of raw material supply for process plants, dealing with the uncertainty in supplying activity and studying on new and efficient optimization algorithms. All of this are contributed to build an integrate raw material supply programming (RMSP) system which is derived from the idea of supply chain management (SCM) with the producer as a chore in the role framework.In this thesis, RMSP model is discussed in chapters with idea and concepts, framework design, uncertainty description, inventory decision-making and rolling optimization, purchasing order optimal distribution and system implementation with application.First the concepts and framework of RMSP is proposed according to the supply problem of producer under an environment of decentralized supply chain. RMSP is an integration of following functions and objects to be implemented: ensuring the producer sufficient and qualified material, coordinating the relationships between producer and supplier in purchasing activity, reducing the uncertainty risks in supplying procedure and minimize the total cost in raw material supplying channel. It is leaned from the architecture and flow chart that RMSP is a synthetic technical program to deal with inventory and purchasing optimization problem based on intelligent evolutionary algorithms and a rolling decision-making mode designed according to the continuous andsimultaneous raw material supply/consuming procedure. With regard to relationship coordination between producer and suppliers, the thesis proposed the idea of preparative supply and a series of related concepts by referring to the idea of predictive control. The essence of preparative supply is a preparative message about future supply sent by the producer to the suppliers. This information helps the suppliers prepare future supply without demand prediction and estimation.An uncertainty measurement method of set pair information entropy (SPIE) is developed in the thesis deriving for quantified description of uncertainty in supply process from the theory of Shannon entropy and set pair analysis. In SPIE theory, the acquired information about objects is described by connection degree of same-indefinite-contrary, in which cognitive uncertainty and immanent uncertainty are distinguished. The discrete entropy is involved for mathematical description by distribution interval. Furthermore, to estimate the risks of decision, the concept and computation method of set pair risk entropy (SPRE) is proposed with the introduction of effective coefficient.Inventory control strategy is the key of RMSP system which is to determine the best inventory scheme for least cost and safe supply with proper preparative supply information provided to suppliers. The thesis proposed an inventory control strategy based on genetic programming (GP) algorithm for optimization of the object. The object includes synthetic cost of inventory system and the risks cost described by SPRE with the conditions of demand of raw material within a certain period. As a comprehensive solution of inventory cost and security, the result of GP optimization is a best order scheme in related period. In GP evolution, the individual is designed as a chain structure with inventory control pulses, as leafs, linked by uncertain number of temporal operators. Each individual of the evolutionary population is a sequential inventory scheme with variables of uncertain dimensions. This architecture not only simplifies the definition of searching process but also provides convenience for rolling optimization of inventory control. The thesis also discussed three rolling modes of stable running and burst event state. By rolling optimization, RMSP inventory system ensures both the validity ofpreparative supply information and quick reaction capability for burst events.Based on the gross purchasing raw material computed by inventory system, an optimization model is set up for indent distribution among multi suppliers, in which the composition restriction, minimum ordering quantity and fixed costs are taken into account. To solve the problem with discontinuous variables, a chaotic migration parallel genetic algorithm based on constraints is proposed. The characteristics of the algorithm are, the first, to constrain the genes of individuals by variable bounds through the function of Gaussian mapping; the second, to transmit the inequations of constraints into a part of energy function by an improved adaptive penalty function and the last, to introduce a chaotic sequence to pilot the migration of individuals among parallel subpopulations so as to reduce premature convergence. A practical example is given to show the feasibility and effectiveness of the computation model and optimization strategy for raw material purchasing problem.At last, based on the proposed model and algorithm an optimal management system for raw material supply is developed and tested in practical application. The system fulfills functions of inventory control decision-making, purchasing optimization, purchasing early warning and information management.