智慧型可適性計算平台之發展及其於地下水流、熱流與污染傳輸耦合模擬之應用

Development of Intelligent and Adaptive Computation Framework and its Application to the Simulation of Groundwater Flow Coupled with Heat and Pollutant Transport

蔡瑞彬(Jui-Pin Tsai)

指導教授 ： 張良正

國立交通大學/工學院/土木工程系所/博士(2010年)

https://doi.org/10.6842/NCTU.2010.01115

摘要

為了解決傳統數值模式不易擴充及人工智慧方法難與數值模式深入整合等兩大問題，本研究提出一新的計算架構，智慧型可適性計算平台，突破傳統數值模式之開發方法，建立一系統化、自動化之數值建模平台，應用此系統所開發之模式，可隨著需求逐步增加模式模擬能力，並整合人工智慧方法於模擬模式核心，而使模式具有類「人工智慧」能力，如可以因應環境變化調整自身模擬範疇等。模式開發上分為兩階段進行，第一階段為開發一具可擴充性之建模系統，稱為可適性計算平台(ACF)，以ACF所開發之模式可容易的新增或修改模式所計算之方程式，第二階段再將ACF與人工智慧進行整合，而成為智慧型可適性計算平台(IACF)，其中在人工智慧的部分本研究整合了專家系統(expert system)及類神經網路(ANN)。
本研究將發展完成之計算平台應用於發展地下水流、熱流及污染物傳輸耦合模擬模式，為檢驗模式之正確性與適用性，本研究設計數個簡例，再應用IACF所發展模式與Tough2及HST3D兩模式進行模擬驗證。其中Tough2與HST3D以不同之方程式描述地下水流、熱流及污染傳輸問題之運動機制，由於IACF平台之擴充彈性，本研究可在短時間內完成與此兩模式之比較驗證。驗證結果顯示，IACF所發展模式與前述既有模式之模擬結果，具有高度一致性，此亦可證明了應用IACF建模相較於傳統建模方式更有效率。本研究亦將專家系統與ANN整合入以IACF所發展之模式中，由模擬結果顯示，考量專家系統之判斷規則，可因應環境變化而自動調整計算模擬，有效提高模擬效率。系統並成功的將類神經網路實作於地下水流模組中，證明了IACF具有與不同人工智慧方法整合之彈性。前述模擬驗證結果顯示了以IACF發展模式，可容易的修改模式所計算之方程式或新增原模式未考量之方程式，因此可提升模式之開發效率，協助研究者致力於問題本身機制之探討，減少程式開發或修改所費之時間與精力。

關鍵字

可適性計算 ； 人工智慧 ； 數值模式 ； 專家系統 ； 類神經網路

並列摘要

Difficult to modify or extend the computational functions of a numerical model and incapable of integrating artificial intelligent methods into the computation kernel are two important issues in the conventional numerical model development and maintenance. Therefore, this study proposed a novel numerical model framework to improve the numerical model developing efficiency. One can easily modify or increase the computational functions of a numerical model developed by the proposed framework. Furthermore, the model can integrate artificial intelligent methods into the computational kernel. The whole framework is denominated as intelligent adaptive computation framework (IACF) and its development was divided into two steps. The first step was to develop an adaptive computation framework (ACF). The ACF can facilitate the development of a model with the flexibility of adding new functions to simulate new modeling mechanisms easily. The second step is to integrate artificial intelligent methods, expert system and artificial neural network, into ACF and completed the IACF.
After developing the IACF, this study applied the IACF to develop a simulation model for solving groundwater flow coupled with heat and pollutant transport problem. To examine the accuracy of the simulation model, the model simulation results of several hypothesis cases were compared with those of Tough2 and HST3D. The Tough2 and HST3D have applied different equations to describe the flow, heat and pollutant transport problem. The IACF made the model comparison became possible within a limited time. The simulation results showed good consistence among the three models. The situation not only verified the model accuracy but also, on the other hand, demonstrated the capability of IACF to improve model development efficiency. A rule based expert system and ANN model were integrated into the simulation model. Simulation results show that the model can adjust the computations according to the system states and effectively improve the computation efficiency by activating the expert system. The embedding of ANN model was also successfully verified and that demonstrate the flexibility of IACF to integrate various artificial intelligent methods. The flexibility of modifying an exited function or adding a new function in a developed model by using IACF can increase the model development efficiency and facilitate researchers focus more on the problem mechanism instead of program coding and modification.

並列關鍵字

adaptive computation ； artificial intelligent ； numerical model ； expert system ； anual nerual network

參考文獻

24. 陳韋圻，應用專家系統於地下水模式自動化參數檢定之研究，交通大學碩士學位論文，2008。

48. 張良正、陳宇文、朱宏杰、黃浚瑋，遺傳演算法與類神經網路於地表地下聯合營運之應用，農業工程學報，Vol.54，No.2，81-93，2008.

1. T. Toffoli, Cellular automata as an alternative to (rather than an approximation of) differential equations in modeling physics. Physica D, 10, 117–127, 1984.

3. K. Ioannis. A model for the prediction of oil slick movement and spreading using cellular automata. Environment International, 23, No. 6, 839–850, 1997.

6. J. Machta, Complexity, Parallel Computation and Statistical Physics. J. Machta, Complexity Journal, 11(5), 46-64, 2006.