| Summary: | 博士 === 義守大學 === 電機工程學系 === 101 === Due to the powerful learning and nonlinear mapping capabilities, neural network (NN) has been widely employed into many areas, including the areas of engineering, social science and business management. Its real applications in engineering area include various subjects, such as image recognition, pattern classification, signal processing, linear and nonlinear systems’ modeling and system’s control.
Generally, the supervised learning neural network is the most popular neural model used in the real applications and the error back-propagation (BP) algorithm based on steepest decent method is the common learning way adopted by NN model. The aim of NN’s learning is to search the optimal connections (weights) of network which can make the model accurately obtain the linear or nonlinear relationship between the input and output pairs of training data. Once the complex relationship between input and output data is obtained, then such a well-trained neural model can be used to do the estimation of signal.
In this thesis, the estimations of optical properties of touch panel decoration film, including the transmittance and chromatic aberration, are obtained by using neural networks. It is well known that the coating materials and the related control parameters of evaporator are the important influencing factors in obtaining the specific transmittance and chromatic aberration (L,a,b values). The relationships among the transmittance, chromatic aberration and their influencing factors are very complex and nonlinear. It is very difficult to use the certain mathematical model to find or describe these relationships. In this research, the neural network was employed to catch the relationships among transmittance, chromatic aberration and their all possible influencing factors. In other words, an efficiently and precisely automatic evaporation parameters decision system for the touch panel decoration film is expected to be developed, if the optical properties of film could be estimated accurately by using the well-trained neural model. Through such intelligent decision system, the evaporation process could be accomplished successfully and the optical properties of touch panel film could meet the customer’s request.
|
|---|
