快速熱處理之模擬，參數判別和溫度控制

Abstract

快速熱處理是一種用在半導體製程中熱處理的新型的機器。其具有快速 的加熱速度及極低的熱預算，然而，有一些瓶頸限制了此項技術應用在半 導體工業上，其中溫度的控制及均溫皆是需要突破的關鍵技術。我們採用 了模型基礎的控制方法去做快速熱處理的溫度控制，先用一些系統判別的 方法，再利用判別出來的模型去做控制器。然而有些情況，系統判別出來 的模型並不精確，這時我們採用類神經網路的方式，去做控制，並利用多 輸入多輸出的方式，做均溫的控制。 An integrated model is given for the components that make up a rapid thermal processing (RTP) system. These components are the lamp power dynamics, ray-tracing model, and the wafer thermal dynamic model. The models for the components are integrated in a numerical code to give a computer simulation of the complete RTP system. The simulation can be used to investigate the interaction of the furnace, lamp contour, and the control system. Of the various techniques for controlling the temperature of the RTP system,model-based control has the greatest potential for attaining the best performance in the field of conventional control, when the model is accurate. In this thesis, some identification methods are introduced to help obtain more accurate models from measured input-output data. At first, identification techniques are presented to estimate the parameters (time constant and gain) of step response. Then show how to use the input-output measurements to obtain a black-box model (autoregressive exogenous mo del) of the RTP system, which turns out to have better prediction capability. The identified models are useful not only to control law design but also to analysis of system characteristics.We apply the tracking system technique that is known in the digital control literature for the model-based control, and propose a controller-switching rule throughout the nonlinear operating region. To overcome ill performance of the tracking system due to the inaccuracy of the identified model, the neural fuzzy network for inverse control is proposed. The key advantage of neural fuzzy approach over traditional ones lies on that the approach doesn't require a mathematical description of the system while controlling. The neural fuzzy network that we use is the Self-cOnstructing Neural Fuzzy Inference Network (SONFIN) proposed by Junag and Lin. Temperature control involves maintaining a spatially uniform temperature distribution across the wafer while tracking a temp orally varying temperature trajectory such as a rapid ramp to processing temperatures. The combination of lamp added RTP system and the MIMO SONFIN is shown to be able to track a temporally varying temperature trajectory and maintain the uniformity of the spatial temperature distribution simultaneously.