Application Research On Intelligent Maintenance For Production Systems Bases On Ann Prediction

With the development of socioeconomics and innovation of science and technology, the manufacturing system is developing to the direction of integration, continuity, hi-speed and automation quickly in China. Therefore, maintaining the manufacturing system reasonably and ensuring the manufacturing system to operate smoothly has become an important factor in determining enterprises' core competitiveness. Now many enterprises adopt the method of regular maintenance. This not only cannot decrease the problems in manufacturing system, but also enterprises need to pay the expensive economic expenditure. Using advanced electronic computers and information technology and utilizing artificial intelligence can maintain the manufacturing system intelligently. This provide a shortcut to ensure the manufacturing system operate safely, efficiently and reliably.Intelligent maintenance is a systematic engineering that contains information collection hardware equipment, network information transmission; production performance prediction technology and decision support and optimize technology, etc. This paper is about the production performance prediction by establishing neural network technology model. Specific research content as follows:1) Through the review of the literature, investigation and the help of enterprise production management personnel, I analysis the main factors that affects manufacturing system performance and extract characteristic value.2) I establish corresponding self-organizing feature map (SOM) neural network model which includes subdividing the model's structure, process of learning algorithm and the parameters for the model. In Matlab7.1 software simulation run this model, I subdivide the existing manufacturing system state of Hannstar Display company.3) According to different characteristics of the system status, I study each state of manufacturing system and the association rules of the production performance by using BP identification model, and then construct the mature production performance prediction model. In Matlab, I run this prediction model simulate, and analyze the results of the experiment.Research shows that using the algorithm of SOM neural network can subdivide the existing production status; the pattern recognition algorithm of the BP neural network can forecast production system performance effectively. Experimental results of the testing samples are higher accuracy, so it can be a application model of production performance forecast. Therefore, applying the forecast production function of the neural network in the manufacturing system, the height of production performance can be predicted in high accuracy and the prediction results can be referred as well. The principle of 20/80 can be applied better and it can provide support for decision-making optimization more effectively.