Hours Fixed Based On Neural Network Technology Research

With the development of advanced manufacture technologies and systems, such as CIMS, CE, AM, PDM, BPR, the application of information technology in manufacturing enterprises is greatly promoted and the informatization, integration and intelligentization is becoming a tendence.Informatization of man-hour management is an important issue of informatization in manufacturing enterprises. It is not only a necessary part for integration system of design (CAD/CAPP), manufacture (CAM) and information management (MIS/ERP), but also as an important basis for production planning, accounting and cost control in enterprises.Based on the design and development of Computer-aided Mould and Fixure Man-hour Quota System (CFMH), a method of calculating man-hour quota based on artificial neural networks (ANN) is proposed in this thesis and a sub-system of calculating man-hour quota based on ANN is also implemented. The main contributions of this thesis are as follows:1. By analyzing the principle and calculation methods of man-hour quota, the work flow and system framework of CFMH are considered and designed. Especially, the sub-system framework and funtion of calculating man-hour quota based on ANN is studed.2. A model for calculation of man-hour quota based on ANN is established. On the basis of analyzing the structure of standard quota table, an approach for constructing a man-hour quota ANN model using synthetical methods is proposed. By using Matlab script language, the man-hour quota calculation model is described accurately. The input parameters of the model are selected from the parameters of the standard quota table and the train set is getten from the data of the table. Meanwhile, the calculated results are post-corrected according to the applied part. Dealing with calculation of man-hour quota, a new way of inormalization is explored.3. An approach to extract text information from heading block of the CAD files for use of the sub-system and import them into database isproposed. Feature extracting method for text information is one of theemphasis of this thesis. The problem of identifying text from different format of title block is solved by using template and feature identifying technologies. So the text information in CAD files can be recognized and shared with different CAD systems.4. The development platform and the interface of this sub-system are detailed. Under a Visual C++ integrated environment, it is realized by using a development tool OpenDWG Toolkit and Matlab. Finally, some experiment and application results for the proposed approach and system are also presented.