微波催化酯化反應SVM模型的優化

化工論文編號:HG199  論文字數:10847,頁數:24

 目  錄
中文摘要 I
英文摘要 II
目錄 III
1. 緒論 1
2. 微波催化酯化反應建模 4
 2.1 建模的背景 4
 2.2 微波催化酯化反應模型分析 4
 2.3 支持向量機 5
 2.3.1 支持向量機原理  6
 2.3.2 核函數   8
 2.3.3 最小二乘支持向量機   8
 2.3.4 支持向量機主要特點   9
 2.4 利用LS-SVM對水楊酸乙酯微波催化合成建模 10
 2.4.1 數據預處理   10
 2.4.2 交叉驗證   10
 2.4.3 模型的建立   10
3. SVM模型的優化 13
 3.1  優化方法簡介 13
 3.2  粒子群優化算法 13
 3.2.1 粒子群優化 13
 3.2.2 隨機粒子群 14
 3.2.3 算法性能測試   14
 3.3  水楊酸乙酯微波催化合成SVM模型的優化  16
4．總結與展望 18
致謝 19
參考文獻 20

摘 要：本論文針對微波催化酯化反應過程表現出較強的非線性，影響因素眾多，催化機理復雜的特點，充分利用支持向量機在處理小樣本方面的優勢，使用最小二乘支持向量機（Least Square, LS-SVM）基于實驗數據對水楊酸乙酯的微波催化合成反應進行建模。通過交叉驗證，模型的擬合誤差平方和為2.93%,。標準粒子群優化（Standard Particle Swarm Optimization）算法結構簡單易于實現,但處理高維復雜問題容易陷入局部極值。為此，對標準粒子群進行改進，首先構建隨機粒子群算法，通過標準測試函數驗證，表明其在運行速度和尋優能力上都好于標準粒子群，但可能會將找到的最優值又拋出。最終本文實現一種保證全局收斂的算法，并用于優化水楊酸乙酯的微波催化合成的SVM模型，得到最優條件為：酸醇比1/11，功率402W，催化劑用量3.00mL，反應時間43min，對應的產率為81.91%。

關鍵詞：微波；催化酯化；支持向量機；粒子群優化

Abstract: This article built SVM model of synthesis of ethyl salicylate under microwave on the experiment data ,by using LS-SVM toolbox .Nonlinearity and complex mechanism is showed in the microwave catalytic esterfication, but this I advantage of SVM as well as dealing with small sample.,. By cross validation the sum of square error of the model is 2.93%.The standard PSO optimization algorithm is easy on structure and prone to be carried out ,but may sink into local best when conducting complex problems of higher-dimension .For this reason, PSO is improved in as stochastic PSO. By verifying on test function, the latter algorithm is better in speed and optimization ability ,but may throw the found best solution .Eventually a styptic algorithm in the global region is carried out to optimize the svm model of synthesis of ethyl salicylate under microwave .The best condition is :acid alcohol ratio of 1/11,the power of 402W, catalyst of 3.00 mL and reaction time of 43 min. The corresponding yield is 81.91%.
 
Keywords：Microwave; Catalytic esterification; LS-SVM;PSO