The Self-Adaptive Dynamic Organization And Scheduling For Knowledgeable Manufacturing System

To reduce the repetitive research on the common parts of the advanced manufacturing systems, as a new concept, the knowledgeable manufacturing system (KMS) incorporates various kinds of advanced manufacturing knowledge converted from manufacturing mode, and achieves the complementary advantages of various manufacturing modes. The self-adaptation is one of the important characteristics of KMS, which enables the manufacturing system to rapidly respond to the dynamic changes of the production environment; meanwhile, it can adjust the structure and scheduling strategy as needed, and improve the competitiveness of the manufacturing enterprises. To this end, the self-adaptive dynamic organization and scheduling for KMS is studied in this dissertation. Firstly, aiming at the matching feature between current knowledgeable manufacturing mode and dynamic environment factors, the matching classification method based on the nonlinear fuzzy weight-support vector machine (NFW-SVM) is proposed; Secondly, with respect to the unmatched classification result of the manufacturing mode and dynamic production environment, the model of dynamic knowledge mesh and the corresponding dynamic organization method of KMS structure are presented, which can realize the structure adjustment and improvement of manufacturing system so as to adapt the changing requirement of production environment; Then, the evaluation method based on the fuzzy data envelopment analysis/assurance region (FDEA/AR) is put forward to evaluate the efficiency of the manufacturing modes which are generated during the dynamic organization process and contained inherently in KMS, and to select the most effective one for self-adaptive decision-making; Finally, aiming at the uncertainty of production environment in KMS, a dynamic scheduling strategy based on improved (Q-learning is proposed to instruct the self-adaptive selection on the scheduling strategy under the dynamic production environment.The main contents of this dissertation are introduced in detail as follows:1. To judge the matching category between the current knowledgeable manufacturing mode and dynamic environment factors, and provide the bases for rapid response of manufacturing enterprise, a model of nonlinear fuzzy weight-support vector machine (NFW-SVM) is proposed, in which fuzzy inputs and imbalance of the different matching categories of samples are considered. Considering the vagueness and uncertainty of dynamic environment factors, the triangular fuzzy number is introduced to describe the vague factor. Concerning the imbalance characteristic of the data sample in different categories, the diverse penalty factors are set up in the model to reduce the fault proportions of small samples. The nonlinear separable problem with fuzzy and imbalance characteristics is transformed into the solution problem of the fuzzy chance constrained programming, and the clear equivalent programming equation of the fuzzy chance constrained programming is deduced. The mutation operator and dynamic inertia weight with constriction factor are introduced to the standard particle swarm optimization algorithm. To enhance the classification accuracy, the model parameters are optimized through the improved particle swarm optimization algorithm. The classification method is given based on NFW-SVM to judge the matching category between current manufacturing mode and dynamic environment factors. Finally, an example demonstrates the effectiveness and feasibility of the proposed method.2. In regard to the unmatched classification result, the knowledge mesh describing the advanced manufacturing mode is improved for the dynamic self-adaptive function. Thus, a new concept named by dynamic knowledge mesh is presented based on the theory of multiple set. The model structure of the dynamic knowledge mesh consisting of the static and dynamic parts is built. To realize the effective decision-making during the dynamic organization process, the concept of static sub-knowledge mesh is defined, and its matching measuring method characterized by the information matching degree, functional matching degree and perfection degree is presented, in which the amount of information and function factors of the static sub-knowledge meshes are considered; meanwhile, the monotonicity and boundedness of the amount of information are proved. The dynamic organization algorithm based on the matching degree is conducted. The self-adaptive dynamic organization of KMS is implemented, and the corresponding enabling tool is developed. Finally, the feasibility and practicability of the presented dynamic knowledge mesh model and organization algorithm are verified by an application example.3. With respect to the various manufacturing modes which are possiblely generated during the dynamic organization process and contained inherently in KMS, an evaluation method is proposed based on fuzzy data envelopment analysis/assurance region (FDEA/AR). The most effective manufacturing mode in the evaluation results is selected to guide the production to meet the demand of self-adaptation characteristic in KMS. The knowledgeable manufacturing modes are taken as decision making units (DMUs), taking into account their complexities, the FDEA/AR model which can evaluate the efficiency of manufacturing mode is established. Considering the uncertainty and vagueness of the inputs and outputs, the triangle fuzzy number is introduced. The assurance region derived from experts’opinions is applied to avoid the influence of traditional Non-Archimedean infinitesimal on the evaluation results. To solve the fuzzy linear programming problem, a-cut is introduced to calculate the bounds of the fuzzy inputs and outputs. At the confidence level a, the upper bound and lower bound of the fuzzy efficiency score are deduced. The ranking approach is provided on the basis of the fuzzy effective upper bounds and lower bounds at all a-cuts. The validity of the proposed approach is verified by an example.4. Aiming at the uncertainty of the production environment in KMS, a knowledgeable dynamic scheduling simulation system based on the multi-agent is built, in which the improved contract negotiation mechanism is adopted and the job to be processed bids for the machine agent with idle time. To ensure that the machine agent can select the appropriate bid job according to the current system status, the improved Q-learning based on clustering & dynamic search (CDQ) algorithm is proposed to guide the adaptive selection of dynamic scheduling strategy in the dynamic production environment. Considering the system status space is too large, the dynamic scheduling strategy adopts the sequence clustering method to reduce the dimension of system status. The system studies according to the status difference degree and dynamic greed search strategy. The convergence of the algorithm is proved and the complexity analysis is conducted. By the simulation experiments, the effectiveness and adaptability of the proposed dynamic scheduling strategy are certified.