本論文對鋼板熱軋製程中之精軋區,從原廠提供之線上資料分析開始,以熱傳、物理學理建立含六個參數之熱模型,並以最佳化程序鑑別熱模型之參數。 由於使完軋溫度估測誤差降至最小的熱模型並非唯一解;以不同的熱模型參數可以得到相同的完軋出口溫度,因此,要得到正確的熱模型參數,必須藉助其他資訊。在本研究引入根據各軋機溫度所建立之軋延力模型:估算各軋機處之軋延力,再與實際量測之軋延力比較,藉以修正熱模型參數。 在建立精軋區熱模型時,需要精軋區入口溫度的資訊。惟線上之量測數據不足與原廠溫控系統之估測不正確,造成精軋區入口溫度資訊的缺乏,基此,本研究另根據粗軋區出口高溫計之量測資料,建立含兩個參數之保溫傳送區熱模型,以估算所需之精軋區入口溫度。 最佳化程序以線性化之溫度估測誤差為目標函數,配合權重函數、更新平滑係數,及適用於各鋼種之基準參數為最佳化程序之初始值,以連續適應性參數最佳化進行熱模型參數的鑑別後,對於完軋溫度估測誤差之均方根值可小於5℃,軋延力估測誤差之均方根值亦可低於5%。
This paper presents a study for establishing a thermal model, including six parameters of a hot strip mill during finishing mill process. This thesis starts with the focus on the analyzing of the original online data, then modeling the finishing mill process with physical and heat-transfer theory. The numerical optimization technique is used to identify parameters of the thermal model. The solution of the thermal model for minimizing the estimate exit temperature error of finishing mill section is not unique. This means the same estimate exit temperature of finishing mill section can occur from different parameters of the thermal model. For the correct parameter site of thermal model, the more information will be required. This thesis has a new idea from the force model, which is established by the entry temperature of each rolling mill. The parameter is determined by comparing the predicted force from the force model, with measured one. While modeling the finishing mill process, the information, neither from the on-line measurement data nor the on-line computer temperature prediction system, is not sufficient nor inexactitude. Establishing a thermal model, based on the on-line measurement data of the finishing temperature of roughing mill section, is required. This model, including two parameters of a hot strip mill during holding table, is for re-building the information of temperature of finishing mill process. Therefore, by optimization method with linearized estimated error and weighting function, this thermal model has well-performance on serial strings. The exit temperature of finishing mill section predicted by the model agrees to the measurement data with a RMS error, lower than 5℃, and the force model estimated error lower than 5%.