Research On Production Planning And Scheduling Under Uncertainty

Production planning and scheduling are not only the important elements for enterprise production and management, but also the effective basis of production decision. Production planning and scheduling involve a lot of uncertainties in extend and depth, for instance, production demand, sale price, production cost, utility, processing time, production capacity, due time, and so on. All above uncertainties play serious effects on production planning and scheduling, which may make the optimized solution not optimal or not infeasible, and further to affect production effectiveness and profit. The production planning and scheduling problem under uncertainties will be studied in this thesis. The considered uncertainties include product demand, product price, cost and utility. According to the time scale, uncertainties are categorized and dealt with respectively in planning and scheduling layer. Compared to production planning, production scheduling involves more constraints, and the modeling methods for production scheduling are more various. So production scheduling under uncertainty is one of the important contents in this thesis. Furthermore, utility uncertainty is closely associated with production process. It has a serious impact on the optimality and feasibility of scheduling result, but that is always neglected. The research on production scheduling problem is studied deeply with the center of utility uncertainty in the terms of modeling method and process category. In order to reduce the effects of utility uncertainty on production and profit of enterprise, a vari-period scheduling strategy, fuzzy relationship between utility and production rate, a hybrid bi-level continuous time model, planning and scheduling integrated model are proposed. The main research works and innovation points of this thesis are as follow:1) The production demand, sale price and cost are considered in planning model. Based on discrete time modeling method, a MILP model is formulated. Chance constraint programming method is utilized to describe the above three uncertainties. Except basic constraints, the utility, production rate fluctuation and inventory reference limitation are also introduced. In case study, the effects of confidence level on optimized solution together with the reason are analyzed. The formulated model can deal with multi uncertain parameters, and make the constraints with uncertain parameters feasible under a certain confidence level. The optimized solution can supply the basis for production decision.2) Based on discrete time modeling method, formulate two MILP models for both batch process and continuous process, to manage the effects of utility disturbance on production process. In scheduling model of batch process, introduce a 4-index 0-1 variable, which can enable tasks to continue over multi scheduling periods, and effectively solve the inaccurate problem of utility consumption. In scheduling model of continuous process, a vari-period scheduling strategy is proposed to address the mismatching problem between duration time of utility disturbance and scheduling period, as well as the computational complexity problem. When utility is without disturbance, the proposed model can accurately count the utility consumption; when utility is with disturbance, the proposed model can tackle and respond to the utility disturbance, and also can reduce the effects of utility disturbance on optimized solution. Regards to the continuous process, when the utility disturbance appears, the vari-period scheduling model are used to manage the utility disturbance, which obtains a better optimal solution.3) Based on Unit-specific Event-Based Continuous Time modeling method, formulate a MINLP model considering utility uncertainty and the uncertain relationship between the utility and production rate. Fuzzy theory is utilized to describe the utility uncertainty and the uncertain relationship. The uncertain model is transformed into deterministic model by mathematical transformation. The effectiveness of proposed model is verified, and the effects of some important parameters on optimized solution, as well as the value selection method are analyzed.4) Due to the defects of unit-specific event-based continuous time model, that are over-strict and double counting, a unit-specific event-based and slot-based continuous time hybrid bi-level model is proposed. In upper layer, USCTM modeling method is used to formulate a MILP model, in which the utility constraint is not included. In lower layer, slot-based continuous time modeling method is utilized to formulate a MILP model, in which the utility uncertainty is considered. Fuzzy theory is introduced to describe the uncertainty of utility. The proposed UEB&SBCTM hybrid model is solved by iteration between the models in upper and lower layer. The UEB&SBCTM hybrid model not only has the advantage of USCTM, but also can address the defects of USCTM, and further supply the foundation for production decision under uncertainties.5) Considering multiple uncertainties in production planning and scheduling, a multi-period bi-level integrated planning and scheduling model is formulated. In planning layer, an uncertain discrete-time linear programming model is formulated based on discrete-time modeling method. Chance constraint stochastic programming is introduced to describe the demand uncertainty. The fluctuations of production rate and deviations from the reference limits of inventory are also considered in planning model. In scheduling layer, production scheduling problem with utility uncertainty is addressed. An uncertain continuous-time mixed-integer linear programming model related to multistage multipurpose batch process is formulated based on unit-specific event-based continuous-time modeling method. Fuzzy theory is utilized to describe the utility uncertainty in scheduling model. Finally, the uncertain planning and scheduling models are transformed into deterministic models by mathematical transformation, and solved iteratively by rolling horizon optimization strategy. The feasibility and effectiveness of the proposed model are illustrated through a benchmark example from literatures. Chance constraint programming can combine uncertain information to optimize the production yield, and can supply decision basis for decision maker, which are beneficial for increasing the reasonability and effectiveness of decision. The utility uncertainty is modeled by fuzzy theory, which can describe the utility uncertainty more actually. Through verification, the proposed model can increase the feasibility of scheduling solution, effectively reduce the effects of the two uncertainties and increase the utilization of machines and utilities.