| Quick Response Manufacturing System (QRMS) based on Quick Response theory is raised to shorten the circle of product development and manufacturing, and met the customers'requirement as quickly as possible. The time analysis of product development process is of great significance for maintaining the system's quick response capability and promoting progress control and management ability. The thesis focuses on the research of time estimation and optimization for product development process in QRMS.First, some basic concepts of product development process are introduced. The task of product development process of QRMS and the design iteration problem exists in it are described, and based on meta-model approach, a time estimate program of product development process in QRMS is proposed. The analysis is mainly focused on the problem of design iteration,which is as the important feature…. Especially, depict the iteration description based on DSM in detail. In order to evaluate the time of product development process more scientifically and reasonable, the meta-model approach in computer simulation area is introduced. Meta-modeling procedures are proposed in line with the characteristics of problem of design activity time estimation, and meta-model fitting method and training data selection method are the points that focuses on.Secondly, in the light of meta-model fitting method, the product design activity time estimation method based on Gaussian Process meta-model is put forward. Divide the common meta-model approaches into parametric method and nonparametric method, and introduce them respectively., present the defects of nonparametric method and Gaussian Process meta-model can replaces several common nonparametric methods theoretically. Introduce the modeling principles of Gaussian Process meta-model and illustrate the covariance function setting and hyper-parameter optimization methods. Since it is one of the properties of design activity time estimation that the training data have linguistic variables, the concept of fuzzy distance is introduced. After linguistic variables fuzzed, covariance function is computed according to its fuzzy distance, and the modeling procedures are proposed. Case study shows the advantages of Gaussian Process meta-model approach in solving design activity time estimation problems.Thirdly, a data selection method based on optimal Latin Hypercube design is proposed for the problem of training data selection. The estimation accuracy and efficiency are greatly affected by training data selection, therefore, in order to cooperate with the application of meta-model approach in time estimation of product development process, the concept of experiment design is employed and optimal Latin Hypercube design is adopted for guiding the selection of historical data selection; Consider orthographic attributes, ? p criterion and Kullback-Leibler criterion; The theoretical high and low limits of Kullback-Leibler criterion are identified to construct objective function. The improvement is made on ILS algorithm for multi-objective optimization of LHD,Fourthly, the study is on time estimation and optimization of QRMS product development process in overlapping mode. To build the model for product development process in overlapping mode, the thesis suggested an information interactive strategy between upper and lower design activity based on the existing research along with the information interactive strategy and representation of every kind of iteration parameters appropriate for this mode. Assuming design activity time as the constant to construct the general model in overlapping mode, the time analysis model is reconstructed taking time uncertainty into account. An optimization method based on ant colony optimization is brought forward for setting the starting time of every design activity, and the reschedule of starting time can effectively reduce the overall time of product development process. |
PDF Full Text Request