Planned Production Capacity For Dynamic Uncertainty Modeling And Prediction

This paper based on the research project comes from national natural foundation project"Dynamics uncertainty-oriented planned capacity forecasting and risk decision-making research", this project forecasting planned capacity though analysis the uncertainty factors during capacity planning, then estimate and control the risk of over or short planning. Inaccuracy of capacity parameters estimate in factory capacity planning model will lead to low reliability of planning solution. Aims at the problem, the content of this paper is as following:First, planned capacity concept and its forecasting system analysis: based on dynamic uncertainty of capacity planning in manufacturing, propose the concept of planned capacity and design the planned capacity forecasting system.Second, planned capacity modeling analysis: this part includes model building and simplification. Firstly, using time analysis in manufacturing and basic formulas of production performance metrics, to build two-step planned capacity model that composes of general model and industrial application model. After that, model simplification step includes following processes: 1) Classify the variables with degree of controllability and schedulability, to get controllable and uncontrollable independent variables. 2) Use two-step global sensitivity analysis which combines modified Morris method with Sobol'to identify weights of uncontrollable variables. 3) Use fixed coefficients instead of uncontrollable variables with low sensitivity and controllable variables to get planned capacity simplified model.Third, planned capacity forecasting method analysis: based on the simplified model, down-up planned capacity forecasting solution is used. Firstly, propose the improved nonparametric density kernel estimate for non-negative data analysis, and use it for uncertainty quantification of key variables forecasting. After that, use uncertainty propagation method based on Monte-Carlo to forecast planned capacity, to get planned capacity forecasting point estimation, uncertainty distribution and uncertainty interval with certain confidence level. Fourth, planned capacity modeling and forecasting case study: take capital-intensive chip assembly and testing factory, labor-intensive car overall assembly factory as examples for case study of planned capacity modeling and forecasting.Fifth, software design and development: based on Excel-VBA, which achieves modified Morris method, Sobol', improved nonparametric density kernel estimate and uncertainty propagation method based on Monte-Carlo.This paper aims at the problem that unaccuracy capacity estimate under dynamic uncertainty, propose concept of planned capacity, design planned capacity forecasting system and dicuss its key techniques. Take semiconductor manufacturing and car manufacturing as examples to validate the proposed system. Develop relative software tool and apply its in factoris to prove its validity and practicability. This planned capacity forecsting system increases planned capacity forecasting accuracy and cuts forecasting time consumption, in order to provide more accuracy capacity information for factory capacity planning model, and have great meaning for efficiency improvement of factory capacity planning decision-making.