Rescheduling Models And Algorithms For Disruption Response In Production Scheduling

In the production process of manufacturing enterprise, due to the impact of disruptions such as new task arrival arising from the market demand uncertainty and stochastic machine breakdown arising from process uncertainty, the original optimal schedule with respect to production cost and resource allocation is no longer optimal or even infeasible. Disruptions increase the production cost, as well as cause deviation of the actual schedule from the original one, disturbing the original schedule dependent activities such as delivery time quotation, workers shift arrangement and raw material procurement. How to make a quick and effective response to the disruptions to minimize the production cost and disturbance simultaneously becomes urgent for the manufacturing enterprise. This is as well the key and core problem in the research field of disruption management.In the existing researches of disruption response in production scheduling, the deterioration effect of tasks'processing times caused by workers'operating fatigue, equipment aging and so on is often neglected. The controllability of tasks'processing times, usage of machine breakdown probability and machine maintenance activities and rejection of some tasks which can substantially improve the response flexibility are also not considered. Due to the high complexity and multi-objective nature of the response strategy, the existing theories and methods are difficult to take multiple stakeholders" interests into account and rapidly formulate an effective response strategy which simultaneously optimizes the operational cost and disturbances. In order to improve the scientific nature of the disruption response strategy, this dissertation constructs the response model for new task arrival and stochastic machine breakdown, respectively, with the real production environmental charateristics and the production factors under consideration, and designs rapid and effective solution algorithms. Finally, application research is done on the disruption response problem in a container manufacturing enterprise in Dalian to examine the efficiency of the proposed models and algorithms, and the application value of this paper. The main contents of this dissertation are as follows:(1) Researches on the response model and the solution algorithm for new tasks arrival with deterioration effect and controllable processing time. The deterioration effect and controllable processing time are emphasized to effectively respond to new tasks arrival. The'"predictive-reactive" strategy is adopted to construct the Pareto optimization model with the objective of minimizing operational cost and disturbance. Combing the model characteristics, a hybrid multi-objective evolutionary algorithm based on a priori knowledge, multi-objective simulated annealing (AMOSA) and NSGA-? is designed for large-scale problems. The algorithm makes full use of the advantages of NSGA-? in convergence and diversity, and introduces the multi-objective simulated annealing process into NSGA-II framework to reduce the probability of getting stuck in local optima. The structural properties of the model are also used to guide the search process to further improve the convergence and the quality of the algorithm.(2) Researches on the response model and the solution algorithm for new tasks arrival based on rescheduling integrated with production factors. On the basis of content (1), to effectively respond to new tasks arrival, the "predictive-reactive" strategy is adopted to construct the integrated Pareto optimization models with the objective of minimizing operational cost and disturbance, where the preventive machine maintemance and rejection of some tasks are embedded in the model, respectively. Combing the model characteristics, a hybrid multi-objective evolutionary algorithm based on a priori knowledge, differential evolution, tabu search and NSGA-? is designed to solve large-scale problems. The algorithm introduces the differential evolution into NSGA-? framework to improve the convergence speed and global search ability, and introduces a tabu search strategy to reduce the probability of getting stucking in local optima. The model structural properties are used to guide the search process to further improve the convergence speed and the quality of the algorithm.(3) Researches on the response model and solution algorithm for stochastic machine breakdown with robustness under consideration. To effectively respond to stochastic machine breakdown, the "robust" non-buffering strategy is adopted to construct the Pareto optimization model, where the probability information of stochastic machine breakdown is incorporated. In the designed multi-objective evolutionary algorithm, a support vector regression based surrogate model is constructed to approximate the schedule robustness and to reduce the expensive computational cost arising from simulation based fitness evaluation, which effectively improves the convergence speed to meet the real-time response requirement. At the same time, the saved computational cost is used to research the global and local solution space, which improves the quality of the obtained Pareto front.(4) Researches on the application. In order to verify the efficiency of the proposed disruption response model and solution algorithms, the methods are applied in a container manufacturing enterprise in Dalian and a corresponding software system for disruption response is designed. Based on the actual environment and related parameters in factory shopfloor, the proposed model and algorithm are called to generate the response strategy for new tasks arrival and stochastic machine breakdown, which outperforms the common manual response strategy, and improves the scientific and real-time decisions in manufacturing process.This research is the intersection and penetration between operational research, industrial engineering and artificial intelligence. It explores the disruption response issues in production scheduling and the construction of more scientific and practical response models and solution algorithms, which provides a new idea and method to generate the complex disruption response in production scheduling. The decision-making of manufacturing enterprise is efficiently improved through the theory-based decision aided system.