Research On FDSS For Manufacturing And Restrictions On Application Of Artificial Intelligence Technology

DSS(Decision Support System) has been used in the manufacturing enterprises. But it is difficult for traditional DSS to adapt the new conditions coursed by the reformation of management thinking. It is exigent to analyzed the shortage of traditional DSS and find out the new way to direction the design and development.The analysis of the new conditions and traditional DSS are performed in this paper. The disregard of the uncertain requirement of DSS is considered the cause of problems in application. A new idea called "Flexible Support " and the "Flexible Decision Support System" are put forward in this paper.The difference between FDSS and traditional DSS is the latter only can deal with the divinable cases. FDSS depends on the mechanism but the model. According to FDSS, to provide a good decision-making condition to decision-maker is the best strategy. FDSS can deal with the uncertain cases which are always the important decision of the enterprise.This paper developed a FDSS based on the "order appraising" of a mechanical enterprise. The result proved that FDSS can satisfy user's needs better than the traditional DSS.Artificial Intelligence has been used in many fields, including the DSS. The relation of Al and FDSS is discussed in this paper. Though the important effect in applications is admitted, people must attach importance to the blindness and exaggeration in Al applications. It is important work to insure Al's development that to find out the restriction of the application of it.According to an in-depth study on the representation of problems and its reason , this paper studies the strategy of test samples selection in the process of Neural Network training. A main method to estimate the effect of Neural Network training depends on the forecast precision of the test samples, but there is no a proper method to select test samples from the sample set. A proportion of the adjacent samples' Euclidian distance based analytical method is presented in this paper. The result of the analysis can direct user to select the test samples and test the effect of the selection. "Attention" effect based strategy is proposed to select test samples. Examples studies show that the strategy can lead to a reasonable test set which can increase the veracity of estimate. This strategy can be applied easily and improve the effect of Al application observably. The strategy is a new approach to analyze the experimentation data in practicalities.Studies show that to attach importance to the restriction of application and find the strategy is an important and promising reach area.