TW201118766A - Adaptive computation method and framework thereof - Google Patents
Adaptive computation method and framework thereof Download PDFInfo
- Publication number
- TW201118766A TW201118766A TW98139641A TW98139641A TW201118766A TW 201118766 A TW201118766 A TW 201118766A TW 98139641 A TW98139641 A TW 98139641A TW 98139641 A TW98139641 A TW 98139641A TW 201118766 A TW201118766 A TW 201118766A
- Authority
- TW
- Taiwan
- Prior art keywords
- equations
- equation
- variable
- computing
- architecture
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 96
- 230000003044 adaptive effect Effects 0.000 title abstract description 9
- 238000013473 artificial intelligence Methods 0.000 claims abstract description 19
- 238000004364 calculation method Methods 0.000 claims description 62
- 238000010586 diagram Methods 0.000 claims description 20
- 238000004458 analytical method Methods 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 3
- 235000011389 fruit/vegetable juice Nutrition 0.000 claims description 2
- 230000008094 contradictory effect Effects 0.000 claims 1
- 238000004088 simulation Methods 0.000 description 23
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 17
- 230000006870 function Effects 0.000 description 15
- 230000008569 process Effects 0.000 description 15
- 230000008859 change Effects 0.000 description 14
- 230000018109 developmental process Effects 0.000 description 13
- 210000003128 head Anatomy 0.000 description 13
- 238000011161 development Methods 0.000 description 12
- 239000003673 groundwater Substances 0.000 description 12
- 239000011159 matrix material Substances 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000006399 behavior Effects 0.000 description 5
- 230000010354 integration Effects 0.000 description 5
- 239000000463 material Substances 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000005086 pumping Methods 0.000 description 5
- 239000002689 soil Substances 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 230000033772 system development Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000005520 cutting process Methods 0.000 description 3
- 239000006185 dispersion Substances 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012827 research and development Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 241000272525 Anas platyrhynchos Species 0.000 description 1
- 241000239290 Araneae Species 0.000 description 1
- 235000011511 Diospyros Nutrition 0.000 description 1
- 244000236655 Diospyros kaki Species 0.000 description 1
- 208000001613 Gambling Diseases 0.000 description 1
- 241001547860 Gaya Species 0.000 description 1
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 241000269435 Rana <genus> Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000009435 building construction Methods 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011438 discrete method Methods 0.000 description 1
- 210000000744 eyelid Anatomy 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010197 meta-analysis Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000007639 printing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
- 230000001603 reducing effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 210000002784 stomach Anatomy 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
201118766 六、發明說明： 【發明所屬之技術領域】 本發明係相關於一種計算方法及其架構。尤指一種可適性 計算方法與架構。 【先前技術】 弟一圖為傳統數值模式之開發流程’一般而言可分為4個 階段：201118766 VI. Description of the Invention: [Technical Field to Which the Invention Is Ascribed] The present invention relates to a calculation method and an architecture thereof. Especially a method and architecture for adaptability. [Prior Art] A picture of the development of the traditional numerical model is generally divided into four stages:
1. 概念模式階段(Conceptualmodel):對於所探討問題、自然現 象或是系統，定義其所涵蓋範圍（如模擬區域，邊界條件， 初始條件等），以及對發生於其中的基本行為進行描述。 2. 數學定義階段(Mathematical formulation):在數學定義上， 通常包含兩個步驟： a. 對於問題中之各種物理或化學基本行為以一個或多個 .數學量化的法則（變數及方程式）描述之。 b. 將前述多個變數及方程式以數學技巧推導整合為數目1. Conceptual model: Define the scope (such as the simulated region, boundary conditions, initial conditions, etc.) and the basic behaviors that occur in the problem, natural phenomenon, or system in question. 2. Mathematical formulation: In mathematical definition, it usually consists of two steps: a. Describe the various physical or chemical basic behaviors in the problem by one or more mathematically quantified rules (variables and equations) . b. Integrate the aforementioned variables and equations into mathematical techniques to derive the number
較少之變數及控制方程式(通常為偏微分方程^(ρ 的型態）。 3..數值離散階段(]^11111此&1(^1^2油〇11):定義空間格點， 原時空上連續之解函數以時空上不連續“數 似’再選用適當的數值方法，如有限差分法（f職），有 有限體積法(FVM) ’邊界元素法_ 及有限解析法（讓）等，將前述之控制方程 空間及時間上之數值積分推導，而將原來之控制方 表達成空間格點上之離散型代數方程式。 工''' 4 201118766 ^式^皆段㈣顧也油啊亦前述離散型變數 吊以《程式計算，因此需進行相_式_發， 大部份；先將所有空·點上之離散型代數方程式，电 陣方程式，此矩陣方程式描述了所有空間格點上離散 欠數彼此間之時空_。前述之矩陣方程式，仍需再以適當 之矩陣解法求解各變數之數值。 田 第-圖之傳統數鋪式開發流程乃環環相扣，惟若欲加入 考量之變量或現象，則須加人描述此行為之規則 (絲式）’減_鱗推導並修正朝題讀制方程式、 數值解及轉絲式，侧電齡切需要重新修改，修改工 作往往相當繁複，也因此關了顧式擴充祕正的彈性。 傳統數值模式由於缺乏彈性而無法與人工智慧方法緊密 的結^。因此，傳統數值模式，只能模擬其控制方程式所描述 之現象’無自我學料而擴充本身之模擬魏，亦無依環境的 不同而自行改變模擬内容等類「智慧」的能力。Less variables and governing equations (usually partial differential equations ^ (types of ρ). 3. Numerical discrete stages (]^11111 this &1(^1^2 oil 〇11): define spatial lattice points, The continuous solution function in the original space-time is discontinuous in the space-time "number-like" and then select the appropriate numerical method, such as the finite difference method (f job), there is finite volume method (FVM) 'boundary element method _ and finite analytical method (let And so on, the above-mentioned control equation space and time numerical integration is derived, and the original control party is expressed as a discrete algebraic equation on the spatial lattice point. Gong ''' 4 201118766 ^式^段段(四)顾也油Ah, the above-mentioned discrete variables are also calculated by the program, so it is necessary to carry out the phase_form, most of them; firstly, all the discrete algebraic equations on the space and the point, the electric matrix equation, this matrix equation describes all the space grids. The space-time between the discrete and the negative numbers is _. The matrix equations mentioned above still need to solve the values of the variables by the appropriate matrix solution. The traditional development process of Tiandi-Map is interlocking, but if Add a variable or phenomenon to be considered The rules describing this behavior (silk type) 'decrease _ scale derivation and correct the equations, numerical solutions and wire-turning equations, the side-age age cuts need to be re-modified, the modification work is often quite complicated, and therefore the expansion is The elasticity of the secret. The traditional numerical model cannot be closely related to the artificial intelligence method because of its lack of flexibility. Therefore, the traditional numerical model can only simulate the phenomenon described by its governing equation, and expand its own simulation Wei without self-study. There is also no ability to change the "smart" of analog content, etc., depending on the environment.
近年來，細胞自動機（Cellular Automata)的架構雖可對 於問題進餘當之化簡，__㈣⑽_Aut_ta本身 造成相當之限制，例如應用於三維未餘和地下水模式時，若需 使用綱題巾某區域之總進出流量，則CdIularA咖她本身 之架構無法提供此種資訊，且CeMar Automata之各個細胞僅 能依，t時刻的狀態進行w時刻之計算(亦即顯示法）。 、’'T、上所述，可知習知技術具有下述缺點：（1)開發計算模 擬軟體費時f力的問題；⑺難以更新、擴充及累積計算功能;、 以及(3)，χ整合各種人m法’ _自鮮胃與演進。 )職疋之故’申請人鑑於習知技術巾所產生之缺失，經過悉 ^忒驗/、研九，並一本鍥而不捨之精神，終構思出本案「可適 八 ϊ S1 5 201118766 性計算方法絲構」，關克服上述缺點， 說明 以下為本案之簡要 【發明内容】 建模射適崎算方法與轉，以解決傳統數值 料發之計算模擬軟體，難以更新、擴充及累 並將數值模式進一步發展成具備自我學習 慧型計算模擬系統。應用本發明的方法，模式可分 =:=具有階段性的效益’第-階段的完成，研究 模式，彈性的藉由增減方程&· 式所的問題，而不須如傳統數值模式般，進 =璃且複雜之程式修改及再開發流程。此將幫助研究人員有 題本質面的探討，而不需如過去般將大量時 間耗費在開發祕正程式以求解欲探討的問題。第二階段可進 羽步t度結合人工智慧方法，使模式進一步發展成具備自我學 白及演進的智慧型計算模擬系統，可藉由「群體智慧」效應， 使所發展的模賊式驗體展現A f人的功能與彈性。〜 、根據本發明的第一構想，提出一種可適性計算方法，包含 以下步驟.(a)提供一方程式集合具有至少一方程式與至少一變 數’（b)對該方程式集合進行--致性分析，以檢驗該至少一 方程式與該至少—變數之一致性，並得一求解順序，且定義多 個空間網格與多個節點，並依該等節點將該方程式集合與該至 从—變數，離散化成為一離散方程組；以及(c)依據該求解順 序，求解該離散方程紐。 ' 較佳地，本發明所提供的方法，其中步驟(b)包括一步驟： (bl)使用人工智慧方法補齊該方程式集合中所缺乏的一或多 201118766 個未知方程式’以完成該一致性分析，其中該或該等未知方程 式對應於該方程式集合所缺乏的一或多個應變數。 較佳地’本發明所提供的方法，其中步驟(M)包括一步 驟’（bll)建立一類神經網路，以找出該或該等未知方程式。 較佳地’本發明所提供的方法，其中步驟(bll)包括一步 驟·（bill)以一主成份分析找出與該或該等應變數相關的一或 多個自變數，以建立該類神經網路。 較佳地’本發明所提供的方法，其中步驟(b)包括下列步 （b2)於該離散方程紕所欲探討的變數所在的一空間劃分該 ，節點’其中該等節點係可依凡諾依圖定義；以及(b3)使用一 簡單差分績該方程絲合離散化成為_散方程組。 較佳地，本發明所提供的方法，其中步驟(c)包括一步驟使 用一$代法來求解該離散方程組的一待解變數在該等節點中 的一任意節點的值。 較隹地，本發明所提供的方法，其中該疊代法包括一内疊 代與一外疊代。In recent years, the Cellular Automata architecture can be used to simplify the problem. __(4)(10)_Aut_ta itself imposes considerable restrictions. For example, when applied to three-dimensional and groundwater modes, if you want to use a certain area of the outline towel The total flow of incoming and outgoing traffic, CdIularA coffee's own structure can not provide such information, and each cell of CeMar Automata can only calculate according to the state of time t (that is, display method). , ''T, above, it can be seen that the prior art has the following disadvantages: (1) the problem of developing time-consuming f-forces for computing software; (7) difficulty in updating, expanding, and accumulating computing functions; and (3) Human m method ' _ fresh stomach and evolution. In the light of the omission of the professional technology, the applicants have come up with the spirit of perseverance, research and development, and the spirit of perseverance, and finally conceived the case "S8 5 201118766 "Wire structure", to overcome the above shortcomings, the following is a brief summary of the case [invention content] Modeling the method of shooting and calculation, to solve the traditional numerical simulation of the simulation software, it is difficult to update, expand and accumulate numerical mode Further developed into a self-learning smart type computing simulation system. Applying the method of the present invention, the mode can be divided =: = has a phased benefit 'the completion of the first stage, the research mode, the elasticity by increasing or decreasing the equation & · the formula, without having to be like the traditional numerical model , into the glass and complex program modification and redevelopment process. This will help the researcher to explore the nature of the problem without having to spend a lot of time developing the secret program to solve the problem to be explored. In the second stage, the tune degree can be combined with the artificial intelligence method to further develop the model into a self-learning and evolving intelligent computing simulation system, which can develop the model thief-like body by the "group wisdom" effect. Show the function and flexibility of Af people. According to the first concept of the present invention, a method for calculating suitability is provided, which comprises the following steps: (a) providing a set of programs having at least one program and at least one variable '(b) performing a pairwise analysis on the set of equations , to verify the consistency of the at least one program with the at least-variable, and to obtain a solution order, and define a plurality of spatial grids and a plurality of nodes, and according to the nodes, the equations are combined with the slave-variables, Discretization becomes a discrete system of equations; and (c) solving the discrete equations according to the solution order. Preferably, the method provided by the present invention, wherein step (b) comprises a step of: (bl) using artificial intelligence to complement one or more of the 201118766 unknown equations in the set of equations to complete the consistency Analysis wherein the or the unknown equation corresponds to one or more strain numbers that are lacking in the set of equations. Preferably, the method provided by the present invention, wherein step (M) comprises a step (bll) establishing a neural network to find the or the unknown equation. Preferably, the method provided by the present invention, wherein the step (b11) comprises a step (bill) to find one or more independent variables associated with the or the number of strains by a principal component analysis to establish the class Neural network. Preferably, the method provided by the present invention, wherein the step (b) comprises the following step (b2) of spatial division of the variable to be discussed in the discrete equation, the node 'where the nodes are According to the definition of the figure; and (b3) using a simple difference score, the equation is discretized into a set of _-discrete equations. Preferably, the method of the present invention, wherein step (c) comprises a step of using a $ generation method to solve the value of an arbitrary node of the discrete equations in the nodes. More generally, the invention provides a method wherein the iterative method comprises an inner iterative and an outer iterative.
較么地’本發明所提供的方法，其中該内疊代係使用一最 佳化方法來求_贿變數_任意節點的值。 •較佳地/本發明所提供的方法，其中該外疊代包括以下步 驟.（cl)方該任意節闕—鄰近節點之該待解變數更新時， 判斷該任意節點是否重啟該内疊代；以及(c2)重複步驟㈣直 到該等節點皆不需重啟該内叠代。 —較佳地，本發明職供的方法，其巾該等麵可 吕實現。 ，本發明所提供的方法，更包括—步驟_整該 方私式w，_述所轉決之_定義朗射基本行為或 201118766 現象。 =本發:第二構想，提出—種可適性計算方法 以下步驟.⑷提供-核式集合；（b)對 括 -致性分析轉得-求解順序，且將該輕料 一離’·以及(推據該求解順序，求解該離散方錬。 根據本發日㈣第三猶，提出―種資购構「可適 架構」以實作第-構想’包含：一計算層’其將描述問題特= 之-方程式集合所欲探討的變數所在的—S間劃分多個節 點，並計算該方程式#合的—待解變數在該等節點上的值。 〜較佳地，本發明所提供的架構，其中該計算層包括多個計 算元’該^計算元㈣-任意計算元對應該等節財一任意節 點，並計算該待解變數在該任意節點上之值，且該等計算元係 獨立計算。 較佳地，本發明所提供的架構，其中該等計算元包括一核 心平台與一應用模組。 較佳地’本發明所提供的架構，其中該等計算元分別執行 一内疊代並以最佳化方法計算該待解變數。 較佳地’本發明所提供的架構，更包括： 一主控協調層，用以執行一外疊代並協調該等計算元間的 資料交換。 較佳地，本發明所提供的架構，其中該主控協調層偵測到 該任意計算元之一鄰近計算元的該待解變數更新時’判斷該任 思計算元是否重啟該内疊代，且持續偵測直到該等計算元皆不 需重啟該内疊代。 較佳地’本發明所提供的架構，其中該計算層與該主控協 調層具有人工智慧演算法。 201118766 較佳地，本發明所提供的架構，其中該等計算元對該方程 式集合進行-致性分析’以驗證該方程式集合有 有解，並得峰絲絲合t各方程絲_序、。盾”疋否 【實施方式】 本案將可由以下的實施例說明而得到充分瞭解，使得熟習 本技藝之人士可崎以完成之’然本案之實施鱗可由下^實 施案例而被關其實施魏。其中相_標號始終代表相同的 組件。 第=圖為本發簡提出之「可適性計算方法」流程示音 圖’第三圖為「可適性計算方法」與傳統數健式建構流程之 差異比較，㈣三圖可看出本發明與傳統方歧異甚大，是整 個開發流程的改變，由第二圖與第三圖可進一步說明各階段= 差異如下： .· · ⑴概念模式(Conceptual model)，如第三圖步驟S31 . 此階段乃在定制題及其t之基行為贿，因此兩者必須 相同，如第二圖之步驟S21。 (2)數學定義(Mathematical f〇rmulati〇n)，如 S32 : 此階段又可分為兩大轉’首先為以各種基本方程式描述 第-階段所定義問題中的各種變量之變化行為，此為正確量化 描述問題所必需’因此本發明與傳統方式皆相同。惟接下來之 步驟本發鴨與魏方法獨，本發赌保留_之基本方程 式組，並蛛做為後續計算縣咖並不如傳統方式般， 將前述所得之多條絲錢行人為的整合料，以盡量減少方 程式及變社數目。傳財法的域為最麵解的方程式與變 201118766 愈有 無法自動化或麟彳卜U_ 導與 a人力為之而 :改前述之基本方程式時==== ==::=式擴充模擬功能的可能。本發 式進杆、Λ彈性’料再對祕歸基本方程 變數，惟Γ城本㈣_將需處雜多財程式與 Ϊ較基本fit程ΐ並未再經人為整合推導，將可維持其原 方程ί數目此Γ像傳統方式般，軸最後面對的 將比較。式已經人終合後所得，其型式 法，=方法常需面對二階以上之微分方程式，惟本發明的方 、，斤处理的方程式絕少高於一階微分方程式。方 度，將直接影響下-步驟數值離散之難易，—階微分町之方 私式可以很鮮的方錢行離散，若是二階微分以上之方程式 則其數值離散之難度將大為提高。除了不進行方程式整合外， t維持將來增減方程式的彈性，這些基本柿式絲亦不採同 ，聯立求解，而是依變數間之相互關係循序逐條求解，如未來 右有方私式的增減’則只需重新定義方程式計算順序即可。因 此本發明在此亦提出雜式—雜的實作方式，以檢驗方 與受數關係的-雜，同時決定雜式之計算順序， 之步驟S22。 — (3)數值離散(Numerical discretizati〇n)，如第三 S33 ： 在此步驟傳統方法由於常需面對較高階的微分方裎式，因 此需要複雜的數值離散方法，如有限元素法(Finite £1伽邮 201118766More preferably, the method provided by the present invention, wherein the inner iteration uses an optimization method to find the value of the arbitrary variable. Preferably/the method provided by the present invention, wherein the outer iteration comprises the following step: (cl) the arbitrary node—when the variable to be solved of the neighboring node is updated, determining whether the arbitrary node restarts the inner iteration And (c2) repeat step (4) until the nodes do not need to restart the inner iteration. Preferably, the method of the present invention, the face of the towel can be realized. The method provided by the present invention further includes a step-by-step _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ = The present invention: The second concept proposes the following steps for the calculation of the suitability: (4) providing a set of nuclei; (b) transferring the order of the syntactic analysis to the order of the solution, and separating the light material from the '· and (According to the solution order, the discrete square is solved. According to the third day of this issue (4), the "property structure" is proposed to implement the "conceptual architecture" to include: a computing layer, which will describe the problem - = - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - An architecture, wherein the computing layer includes a plurality of computing elements 'the computing element (4) - an arbitrary computing element corresponding to an arbitrary node, and calculating a value of the to-be-solved variable on the arbitrary node, and the computing elements are Preferably, the architecture provided by the present invention, wherein the computing elements comprise a core platform and an application module. Preferably, the architecture provided by the present invention, wherein the computing elements respectively perform an inner stack Calculate the variable to be solved by optimization method Preferably, the architecture provided by the present invention further includes: a master coordination layer for performing an iteration and coordinating data exchange between the computing elements. Preferably, the architecture provided by the present invention, Wherein the master coordination layer detects that the one of the arbitrary computing elements is updated adjacent to the computing element, and determines whether the Rensi computing unit restarts the inner iteration, and continues to detect until the computing elements are not The inner iteration is required to be restarted. Preferably, the architecture provided by the present invention, wherein the computing layer and the master coordination layer have an artificial intelligence algorithm. 201118766 Preferably, the architecture provided by the present invention, wherein the computing The meta-analysis of the set of equations is performed to verify that the set of equations has a solution, and the peaks of the equations are obtained. The order of the equations is determined by the following examples. However, it is fully understood that the people who are familiar with the art can complete the implementation of the case. The implementation scale of the case can be implemented by the following example. The phase_label always represents the same component. The third section of the "Compatibility Calculation Method" of the "Easy Computation Method" is a comparison between the "Compatibility Calculation Method" and the traditional number-building construction process. (4) The three figures show that the present invention is very different from the traditional one. , is the change of the entire development process, from the second and third diagrams can further explain the various stages = the difference is as follows: . . . (1) Conceptual model (Conceptual model), such as the third step S31. This stage is in the custom problem and The basis of t is bribe, so the two must be the same, as in step S21 of the second figure. (2) Mathematical definition (Mathematical f〇rmulati〇n), such as S32: This stage can be divided into two major turns' first The variation behavior of the various variables in the problem defined in the first stage is described in various basic equations, which is necessary for correctly quantifying the problem of description. Thus, the present invention is identical to the conventional method. However, the next step is the development of the duck and the Wei method alone, the gambling reserve _ the basic equation group, and the spider as the follow-up calculation of the county cafe is not as the traditional way, the aforementioned multi-skilled pedestrian-made integration material In order to minimize the number of equations and changes. The domain of the method of wealth transfer is the most solvable equation and change 201118766. The more it is impossible to automate or the U 彳 U 导 与 a a : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Possible. The haircut of the hairpin and the elasticity of the hairpin are then returned to the basic equation variables. However, the city’s (4) _ will need to be complicated and the financial plan and the basic fit process will not be artificially integrated and will maintain its The original equation ί number is like this in the traditional way, the axis will face the final comparison. The formula has been obtained after the finalization, and its type method, = method often needs to face the second-order differential equation, but the equation of the invention is less than the first-order differential equation. The degree will directly affect the difficulty of the dispersion of the lower-step values. The square of the differential order can be very discrete. If the formula is second-order differential or more, the difficulty of numerical dispersion will be greatly improved. In addition to not integrating the equations, t maintains the flexibility of the equations in the future. These basic persimmon filaments are not used in the same way, but are solved in tandem, but are solved sequentially according to the relationship between the variables, such as the future right-handed private The increase or decrease of 'just need to redefine the equation calculation order. Therefore, the present invention also proposes a hetero-hybrid implementation method for verifying the relationship between the square and the acceptor, and determining the order of calculation of the hybrid, step S22. — (3) Numerical discretizati〇n, such as the third S33: In this step, traditional methods often need to face higher-order differential squares, so complex numerical discrete methods, such as finite element method (Finite), are needed. £1 伽邮201118766
Method，FEM) ’有限差分法(Finite以迁⑽⑽Meth〇d，⑽等 皆是。本發明由於絕少面對一階以上之微分式，因此只需簡單 的H分’即可對原方程組進行離散化4外，為維持增減 方程j的彈性，本發明為了更易於進行離散化，並建議了^佳 以凡諾依多邊形進行空間格綱之定義，如第二圖之步驟切。 ，（4)程式開發（Pr〇gram deve丨嘴㈣，如第三圖步驟Method, FEM) 'Finite difference method (Finite is moved (10) (10) Meth〇d, (10), etc. Since the invention rarely faces the differential equation of more than one order, the simple equation H can be used to perform the original equations. In addition to discretization 4, in order to maintain the elasticity of the increase and decrease equation j, the present invention is more easy to discretize, and it is suggested that the space lattice is defined by the Vanoyoi polygon, as shown in the second figure. 4) Program development (Pr〇gram deve 丨 mouth (four), as shown in the third step
延續前述數值離散，傳統方式為照顧各節闕之互 響，必需將所有節點上之離散方程式與變數，再組合成矩財 程式’接著發展各種不同矩陣解法之計算程式，對所組成的矩 ^方程，，計算其數值解。本發明將不如傳統數值方法般，組 占所有_上㈣數喊_方程式。減的 節點上之離散方程式為主進行計算，再以各計算節點ΐ =換貝誠合g代計算方式，求得整侧題所有節點 如第二圖之步驟S24。 由剛侧細第二圖關發之數賴式具有可彈性辦 g)此應前述開發流程，經適#規#彳可實作^ 大幅超出傳統數值建模方法之處。如第四圖所示，數值 ί 可舰巍計算平台41與絲式擴充介面 ^ X可因應所欲解決之問題而雜織與擴充。 以下說明本發明之一系統架構： (1)系統架構·· 201118766 ==數=話說，各格點的意義主要是空間上的座標 及其對應的變數而已。而本發明主要的計算皆在格點上 可將格點代表的耗#充如第五圖所示，各格點23不僅具 空間座標及變數，而可視為—個具有獨立運算能力的個體^ 發明將其暫名為「計算元」。而完整的_則以此為基礎於充 至二個層次’即分別如第五圖+所示的計算層21 層22,其進-步說明如下： /、控協調 a.計算元211r211n ··本發明計算層21由計算元2ιι】·2η 組成，計算元211r211n與其鄰近計算元的相鄰關係，亦如以° ，由網格^定義，計算元211^211^；^為最絲的數值計 算’其計算方式’發展之初由開發者依問題而定義，惟若能結 合專豕系統與機惟若能結合專家系統與機器學習等人工智筹、 =法25,各計算元將可容易的以人工辅助#方式增加或改變計 算功能，或是經由自我學習演化增加本身的計瞀To continue the above-mentioned numerical dispersion, the traditional way is to take care of the reciprocal sounds of each thrift. It is necessary to combine the discrete equations and variables on all nodes into a mathematical program. Then develop the calculation formulas of various matrix solutions, and form the moments ^ Equation, calculate its numerical solution. The present invention will not occupy all of the _up (four) number shouting equations as in the conventional numerical method. The discrete equations on the reduced nodes are mainly calculated, and then the calculation nodes ΐ = 贝 诚 合 g g g 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 From the second side of the second picture, the number of types can be flexibly handled. g) This should be the development process mentioned above, and the actual method can greatly exceed the traditional numerical modeling method. As shown in the fourth figure, the numerical value of the ship's computing platform 41 and the silk expansion interface ^ X can be woven and expanded in response to the problem to be solved. The following describes a system architecture of the present invention: (1) System Architecture·· 201118766 ==Number= Talk, the meaning of each grid point is mainly the coordinates of space and its corresponding variables. The main calculations of the present invention can be used to represent the consumption of the grid points on the grid points as shown in the fifth figure. Each grid point 23 not only has space coordinates and variables, but can be regarded as an individual with independent computing power^ The invention temporarily named it "computation element". The complete _ is based on this, which is based on the two levels of 'computing layer 21 layer 22 as shown in the fifth figure +, respectively. The further steps are as follows: /, control coordination a. Computational element 211r211n ·· The computing layer 21 of the present invention is composed of a computing element 2ιι]·2η, and the adjacent relationship between the computing element 211r211n and its neighboring computing elements is also defined by the grid ^, and the computing element 211^211^; ^ is the silkier value. The calculation of the 'calculation method' is defined by the developer at the beginning of the development. However, if the combination of the system and the machine can be combined with expert systems and machine learning, such as artificial intelligence, = method 25, each calculation element will be easy. Add or change computing functions in a manual way, or increase your own calculations through self-learning
—㈣麵^輸叫及涵蓋的 犯圍為本身及與其緊鄰的計算元，惟難以避免的，某些計算與 $量必須就部份或全體計算元整體考量，因此必須在Z別 舁層之上設置主控協調層22，可依需求調整計算元的計算行 為。主控協調層22亦可結合專家系統與機器學習等人工智鼙 方法25增強其功能。 " (2)系統開發原則 •計算元獨立運算 為使系統的運算能力可彈性的變化及擴充，而具有與人工 智慧方法25進行深度結合的可能，計算元211]-21込必需可獨 立運异’而不可如以往數值模式一般，將所有格點23上的變 數以矩陣方程式一起求解。因此’為維持計算元211]-211n獨 12 201118766 立計算的亦兼顧計算元211「211n與其舰計算元的互 動關係三某種形式的疊代運算將是不可避計算元211Γ211η 獨立運开的特性’使發展的計算軟體，可容㈣進行大量平行 化計算或分散式計算。- (4) face ^ lose and cover the crime is itself and the calculation element immediately adjacent to it, but it is difficult to avoid, some calculations and the amount must be considered in part or in all calculations, so it must be in the Z The master coordination layer 22 is set up, and the calculation behavior of the calculation unit can be adjusted according to requirements. The master coordination layer 22 can also enhance its functionality in conjunction with expert systems such as expert systems and machine learning. " (2) System development principles • Computational element independent operations In order to make the system's computing power elastically changeable and expandable, and with the possibility of deep integration with the artificial intelligence method 25, the computing elements 211]-21 must be independently transportable It is not possible to solve the variables on all lattice points 23 in a matrix equation as in the previous numerical model. Therefore, 'for the maintenance of computational unit 211]-211n alone 12 201118766 The calculation of the calculation also takes into account the interaction of the computational element 211 "211n with its ship computing elements." Some form of iterative operation will be an inevitable computational element 211 Γ 211 η independent of the characteristics of the ' To enable the development of computational software, it is possible to carry out a large number of parallelization calculations or decentralized calculations.
7單一類型(均質架構)或多麵(非均質架構)的計算元 <第五圖所不為系統主要計算架構示意圖，惟系統實作上， 计^·元211! 211η可依需求或底詹平台而有不同的型態如：（1) 早-類，計算元(均質架構）：此即每—計算元皆具有強大而相 同的6十=功能’各計算元皆可處理所有的底層計算需求。⑺多 類塑計算元(非均質架構）：計算元依計算功能分成幾種類型， 每一類型的計算元具有部份底層運算力，不_型計算元 間必須交換資訊，以完成所有的底層計算需求。 (3)系統發展步驟 如第五圖所示為系統架構，就系統開發的步驟而言，可以 兩步大階·方式進行發展m⑽效益能及^現，降 低整體開發離，第六圖所示即為本個_㈣關發兩大 階段，其中第—階段乃是衫内嵌人工智慧方法，惟仍需遵守 前述開發原則，維持系統擴充功能的彈性，並為未來與人工智 慧方法的結合預留空間，此階段完成的系統雖未具自我學習^ 擴充功能等特質’惟已是個全新而良好的新型驗值模式，可 ^新及累__擬能力’前述第—階段本發_為可適性計 算(Adaptive Computation，AC) 〇 第二階段則再整合及建置適當的人工智慧方法進一步月 加模式的功能’甚至具有自我學習演化的能力，當資料或模揭 之案例增加後可料各項錄準確度、增加_纽與變數或 甚至增加制等。此人工智慧部份可包括如_經網路 13 201118766 專豕系統（Expert System )或其他機器學習（Machine e^ig、方法等。前述整個兩階段系統的概念，本發明 9 Computation, IAC) 〇 以士主要說明在「可適性計算」階段各計算節點之實作方 各略點間品再以豐代方式求解整個問題，而不以求解全域 矩f方程式之方絲解，依此再適當之程式語言進行程式 寫二即可完成依「可適性計算」概念開發之模擬模式，其 β，程式集合一致性分析演算法」、變數空間上之離散化^ ί求解ί細部實作方式，將錢續之實作案例進—步說明。 刚述各#㈣若再結合人工智慧方法，則升級為「工作元 此時整個^統將晉升至「智慧型可適性計算架構」之層次。 以下#明本案一較佳實施例，其系統架構示意圖為如前述 Ϊί圖所示之可適性計算_，其包含1作元2llr2lln、協調 22卜主控者222與背後的人工智慧系統&而如說明 二苐六_不在尚未整合人功慧前為可適性計算⑽咖 omp咖on，AC)，以下將㈣下水模擬的具體例子，說明至 白㈣^具體實作方式之―，此例將說明各卫作元如何進 行數值計算，及工作元與卫作元之間的溝通方法。此實作案例 稱為「新型態地下水模擬模式」。 ^ 新型態地下水模擬模式之說明： 1·空間網格與節點定義： 本案例採用凡諾依圖(Voronoi Diagram)作為空間網格盥節 點=定義，因此可以配置規則或不規則分佈之運算節點，透過 =讀圖空間分割’可定義運算節闕之相鄰聽。本發明提 出之可適性計難Q之制並秘制哺之定義方式，惟採用 2011187667 The single type (homogeneous architecture) or multi-faceted (non-homogeneous architecture) computing unit < fifth figure is not the main calculation architecture diagram of the system, but the system implementation, counting ^ 211! 211η can be based on demand or bottom There are different types of Zhan platform, such as: (1) Early-class, computing element (homogeneous architecture): This means that each-computation element has a powerful and identical 6=functions. Each computational element can handle all the underlying layers. Calculate the demand. (7) Multi-class plastic computing elements (non-homogeneous architecture): Computational elements are divided into several types according to the calculation function. Each type of computational element has a part of the underlying computing power, and the non-type computing elements must exchange information to complete all the underlying layers. Calculate the demand. (3) System development steps As shown in the fifth figure, the system architecture, in terms of system development steps, can be developed in two steps and large ways. m(10) Benefits and benefits, reducing overall development, as shown in Figure 6. This is the two major stages of _(four), the first stage is the artificial intelligence method embedded in the shirt, but still need to abide by the aforementioned development principles, maintain the flexibility of the system expansion function, and pre-combined with the future and artificial intelligence methods. Space left, although the system completed at this stage does not have the characteristics of self-learning ^ expansion function, but it is a new and good new value-added mode, which can be new and tired __like ability 'the above-mentioned stage--this is _ Adaptive Computation (AC) 〇 In the second stage, the integration and establishment of appropriate artificial intelligence methods to further the function of the monthly add-on model even has the ability to evolve self-learning. Accuracy of the project, increase of _News and variables or even increase system. This artificial intelligence part may include, for example, the network 13 201118766 Expert System (Expert System) or other machine learning (Machine e^ig, method, etc. The concept of the entire two-stage system described above, 9 Computation, IAC of the present invention) The taxi mainly explains that in the "fitability calculation" stage, the implementation of each calculation node will solve the whole problem in abundance mode, instead of solving the square solution of the global moment f equation, and then appropriate The programming language can be written in two to complete the simulation mode developed according to the concept of "adaptivity calculation", the β, program set consistency analysis algorithm, and the discretization in the variable space ^ ί solution 细 detailed implementation method, the money Continued implementation of the case into the step-by-step instructions. Just say each #(四) If you combine the artificial intelligence method, upgrade to "Working Unit. At this point, the whole system will be promoted to the level of "Smart Adaptive Computing Architecture". In the following, a preferred embodiment of the present invention, the system architecture diagram is as shown in the foregoing figure, which includes the 1 element 2llr2lln, the coordination 22b master 222 and the artificial intelligence system behind the & For example, the description of 苐 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Explain how each of the guards performs numerical calculations and how to communicate between the work elements and the guards. This implementation case is called the “new state groundwater simulation model”. ^ Description of the new state groundwater simulation model: 1. Space grid and node definition: In this case, the Voronoi Diagram is used as the spatial grid node = definition, so you can configure the rule node with irregular or irregular distribution. , through the = reading space segmentation 'can define the adjacent operation of the operation thrift. The applicability of the present invention is based on the definition of the system and the definition of the secret feeding method, but the adoption of 201118766
凡諾依圖網格較有彈性。 2.地下水方程式集合定義： 以下1出地下水模擬模式所需之所有方程式及其說明。 ;丨!*cs P/Sdf.d(n2rea) 上式為最重要喊本之f量守財程式，M孔隙 數，穿越面積(“)與控制體綱麟空間切 d方式損，ϋ此在實際顯運算時，水文地脖_空間切 割均已經訂縣已知參數。流雜度(如、、餘和度 與水一速⑺’為式问之變數。而此三個變數需藉由後 、、負之方程式進步疋義，&、度(")可由密度變化方程式式⑼ 推估，此方程式為Rana a Fine等人所建議(丨973)，可定義不 同溫度與不同壓力下的地下水流密度；飽和度(¾)則需透過特 ^生曲線(式c與d)定義，在此採用van Genuchten經驗式(1980)， 其為土壤中不同壓力狀態的水分含量關係；地下水流速(K)可 由達西㈣式(e)定A，其乃湘水力麟來估算水流流速。/B + AlP + A2P2)jΆ(Ρ)(η — θ“ . b. pf n < 0 Θ, for /z > 0 e. iarcy = nV = -K{P)~s ds 上式中尸為壓力水頭、了為溫度，5、4、冷與k〇均為溫度 f S； 1 15 201118766 (Π之函數，6>r為殘餘含水量，為有效含水量，0為質量流 率，匕mD；為達西流速，声為實際流速，尤(户)為未飽和水力傳 導係數，/ζ為總水頭’ ：?代表流線方向，α、广、γ為van Genuchten 經驗式之相關參數，《為土壤孔隙率。 式e為達西公式’其中水力傳導係數([(尸))與總水頭(办）， 則是新出現的未知變數，後續將由式f與g進一步定義之。未 飽和層中，水力傳導係數(夂(P))隨壓力變化而變化，在此亦採. 用vanGenuchten經驗式(式f)計算。另外，總水頭⑻則定義為 壓力水頭與位置水頭之和(式g)，式h為控制體積内部蓄水質 量式，以該時刻之孔隙率、水流密度與飽和度求得該時刻之蓄 水質量。 另外，根據地下水理論，拘限含水層之水量進出與壓力變 化關係，係受到水的壓縮性與土的壓縮性所造成，亦即土壤孔 隙率會隨動變化。式i為土壤絲率隨壓力變化方程式。 (K(P) = KsKr(P) f. < where Kr = θ^5 [1-(1- θ\/γ γ ]2 g h-P + z h. ^ ^(pfnS^Vol .ί« = «ο(1 + ^Ρ) for Ρ>〇 1 η = η0 for Ρ < 〇 其中（為德和水力傳導係數、心⑻為相關水力傳導係數，其 數值隨壓力變化而在0至1之間變化，2為高程，y代表控制 體積内於時刻< 的蓄水質量。 工 至此，在定溫情形下，由質量守恆方程式(式收其他八 個方程式(式b〜1)，合計共9個方程式定義了本問題；其變數為 201118766 P、θε、Pf、Sd、fdarcy ' 、h、n、9 x 數。若在變溫問題中，溫度(r)為變數，^ 9個未知變 流相關方程式。 ^4項引入另外之熱 3.方程式集合一致性分析： 方程式集合-致性分析是以演算法批 充分可解，印是確認方程式_目必_ 程式^合是否 一步訂定方程式集合的求解順序，後^目一致’並進 順序進行求解。 、、了依據此訂定的求解 若下式(1)為方程式集合.· 伽Ά'2χ2，〜3Χ3……Ci Χη) = 〇 /== 甘士广 K3...W⑴ 八中，力代表第i個方程式，丑有 + 集合中，X.彳〇^ 個方程式存在方程式 '代表此墩題的第j個變數， 議題内、為方程式變數係數，其為^十財n個變數纽 表式1並不存在變數j，其方程式變的布林值’若方 卜接著將上述之地下水方程式隹人數^數應為0 ;反之，則為 口理為變數係數第一表， --S. οΤΤΤΓΓΓΊ ο _0^ _0^Τ 〇_ Τ _ο^Τ ο 0 0 2 0 0 0 0 0 2 0 "〇~ 4 2_ 本方法是^ 法疊代求解，即表示會 “’透過最佳化的計算力 值，此問題中，待解變數==的待解變數給予一起烟 欠數為壓力柄(仏配合前述之疊代冰 ί ?； 1 17 9 201118766 Γϊ始相水糊—瓣倾歧為曝改後 ί t 二表可得知更新後式b，f、g、i此四條方程式 有一個未知變數尤(尸）、Λ、以與”，其從 2得。因此可再將第二表上各_此四個變 改=，似細㈣，細㈣上，式d、e== 個未知變數g，％，其亦可由此兩式求得，再 攔位的數值為0 _:後如第四表），由第四表可得以式= &進行麵，執行上铜樣的步驟，可在紅表得The Van Nuo map grid is more flexible. 2. Definition of the set of groundwater equations: The following are all the equations required for the groundwater simulation model and their descriptions. ;丨!*cs P/Sdf.d(n2rea) The above formula is the most important shouting of the f-keeping program, M pore number, cross-sectional area (") and control body outline space cut d mode loss, this is In the actual explicit operation, the hydrological neck _ space cutting has already set the known parameters of the county. The flow noise (such as, the sum and the water and the speed of the water (7)' are the variables of the formula. And the three variables need to be followed by , negative equations, ambiguity, & degree (") can be estimated by the density variation equation (9), which is recommended by Rana a Fine et al. (丨973), can be defined at different temperatures and different pressures Groundwater flow density; saturation (3⁄4) is defined by the special curve (forms c and d), where van Genuchten empirical formula (1980) is used, which is the relationship between the water content of different pressure states in the soil; groundwater flow rate ( K) can be determined by Darcy (4) (e), which is the water flow rate estimated by Xiang Lili. /B + AlP + A2P2)jΆ(Ρ)(η — θ“ . b. pf n < 0 Θ, for /z > 0 e. iarcy = nV = -K{P)~s ds The above formula is the pressure head, the temperature, 5, 4, cold and k〇 are the temperature f S; 1 15 201118766 (Π Function, 6>r is the residual water content, which is the effective water content, 0 is the mass flow rate, 匕mD; for the Darcy flow rate, the sound is the actual flow rate, especially (the household) is the unsaturated hydraulic conductivity coefficient, /ζ is the total head': ? represents the direction of the streamline, α, 广, γ are the relevant parameters of the van Genuchten empirical formula, "for the soil porosity. Equation e is the Darcy formula" where the hydraulic conductivity ([(尸)) and the total head (office), It is a new unknown variable, which will be further defined by the formulas f and g. In the unsaturated layer, the hydraulic conductivity (夂(P)) varies with pressure, and is also used here. Using the vanGenuchten empirical formula (form f In addition, the total head (8) is defined as the sum of the pressure head and the position head (formula g), where h is the internal mass of the control volume, and the moment is obtained by the porosity, water flow density and saturation at that moment. In addition, according to the groundwater theory, the relationship between the water inlet and outlet of the aquifer and the pressure change is caused by the compressibility of the water and the compressibility of the soil, that is, the soil porosity will follow the change. Soil silk rate as a function of pressure change equation (K(P) = KsKr(P) f. < where Kr = θ^5 [1-(1- θ\/γ γ ]2 g hP + z h. ^ ^(pfnS^Vol .ί« = «ο(1 + ^ Ρ) for Ρ>〇1 η = η0 for Ρ < 〇 where (for the German and hydraulic conductivity, the heart (8) is the relevant hydraulic conductivity, the value varies from 0 to 1 with the pressure change, 2 is the elevation, y represents the water storage quality at the time < within the control volume. At this point, in the case of constant temperature, the problem is defined by the mass conservation equation (the other eight equations (formula b~1), a total of nine equations; the variables are 201118766 P, θε, Pf, Sd, fdarcy ', h, n, 9 x number. If the temperature (r) is a variable in the temperature change problem, ^ 9 unknown equations for the unknown flow. ^4 introduces additional heat 3. Equation set consistency analysis: Equation set - The causal analysis is fully solvable by the algorithm batch, and the printing is to confirm the equation _ 目必 _ the program ^ is a step to determine the solution order of the set of equations, the latter is consistent 'in parallel order to solve. ·, according to this set Solve as follows: (1) is a set of equations. · Gaya '2χ2, ~3Χ3...Ci Χη) = 〇/== Ganshiguang K3...W(1) Eight, force represents the i-th equation, ugly + set In the equation, the existence equation of X.彳〇^ represents the j-th variable of the question, and within the subject, it is the coefficient of the equation variable, which is the variable of the ten variables of the formula, and there is no variable j, the equation The changed Boolean value, if the square is followed by the above groundwater equation 隹The number ^ should be 0; otherwise, the first table is the variable coefficient, --S. οΤΤΤΓΓΓΊ ο _0^ _0^Τ 〇_ Τ _ο^Τ ο 0 0 2 0 0 0 0 0 2 0 " 〇~ 4 2_ This method is ^ iterative solution, which means that it will "pass the optimal calculated force value. In this problem, the variable to be solved == the variable to be solved is given the same as the pressure handle (仏In conjunction with the aforementioned iterative ice ; 1 1 9 9 201118766 Γϊ 相 水 — 瓣 瓣 瓣 瓣 瓣 ί ί ί ί ί ί ί ί ί ί ί ί 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二 二You (the corpse), Λ, and ", which are derived from 2. So you can change the _ these four on the second table =, like fine (four), fine (four), formula d, e = = unknown variables g,%, which can also be obtained by the two formulas, and the value of the block is 0 _: after the fourth table), the fourth table can be used to perform the method of performing the copper-like step. In the red table
^進行求解，此時已僅剩式a，即守怪方程式，即為此: 隶後一供需要求解巧方程式。_L 内 列裊(步驟1)^ Solve, at this time only the formula a, that is, the singular equation, that is, for this: the latter is needed to solve the clever equation. _L inner column (step 1)
18 20111876618 201118766
19 201118766 表各階段可求解之方與已知變數 處理階段 第一階段 可求解之方程式 式 b、f、g、i 未知轉為已知之變數 K{P) ' h ·> η > pf19 201118766 Table can be solved at each stage and known variables Processing stage First stage Solvable equations b, f, g, i Unknown to convert to known variables K{P) ' h ·> η > pf
第七表内表示的處理階段即代表了此數值方法内方程3 的處理順序，各階段⑽絲式處糊序並無規定，而階段座 階段間的處理顺序則必須一定。以第二階段與第三階段舉例 f二階段_式d與e兩方程式的順序沒有固定，任何-條p 先處理都是可行的’但是在進行第二 v、s 條方程式的求解。灿丁第—队之削，必須完成此择 =程式-致性分析的過程中，錄後可求到攔位全為⑽ =二不此方程組是可以解的，並簡時也可定出方程_ 通過方程式—致性分析的方程組，即可作 奴，各郎點内之内疊代的運算目標。 ^ 1計异兀（節點）運算：由上述—致性分析結果可知 之變數數目紗方程式求解順序，因此變數數夕 =式往往是最後求解，在此财最後求解的柄 _式有所= 求得，故此方程變;皆已t:序較前之方程式 此方程式評估等號左右兩邊之數數’而在於利用 驗所有的變數解是否給可滿 二！（於相近’進而檢 外了滿足所有方程式，若等號兩邊數值之 20 201118766 =為賴存在’差，再經由第七圖中之疊代計算可將此誤 〃取小化’而得所有變數之解，本發明稱此一步驟為 「内疊代j 、七圖）。此内豐代計算旨在獨立計算每個節點之變數值，底 下將針對「内疊代」做進一步說明。 内疊代計算方法 每個喊點内豐代計算時鄰近節點之待解變數乃取其當下 在此條件下各節點之待解變數值則可透過如最陡坡降法 專最佳=方法進行求解’如第七圖。第七圖為以最陡坡降法為 Φ 狀1邊代汁异流程圖’在前述之地下水流問題中，壓力水頭 $待解&數由於最陡坡降法需要給予初始解並以其開始搜 尋’在此必須給予初始壓力水頭，代入方程式集合中 ，以鄰近 結，之壓力水頭進—步計算出代表變量(在此·為地下水) 之穿越流1’代入連續方程式中，控制表面的總穿越量應與控 糖軸的變化I相等，若非如此則為摊誤差。透過差分方 式可以求碰力水頭值對守恆誤差之微分近似值，應用此微分 近似值’可料步歸鱗收斂。當守㈣差趨近於零時，代 紐制表面的總穿越量與控制體積内的變化量相等，意即已求 得結點本身之壓力水頭。本輪卜步代」，此「内 4代」計算即為第五圖中計算元叫风之計算工作。 5.外疊代處理方法 述内疊代計算’每個節料獨立計算其變數值，計算時 ?近，，僅考慮其當下之值，惟鄰近節點之值可能因其本身之 ^計算而’因_有—上位驗序重覆啟動各節點之計 异直至收斂為止，本案稱此為「外疊代」。此乃不同於傳統方 法透過矩陣解法同時求解整體區域所有節點之值，以維持各 點計算之獨立性與隨之而來的彈性。兹說明如下： 21 201118766 鄰節點資:點ΓΓί f内叠代計算時，部分方程式需運用到相 點所取狀相鄰節點各f，為獨，算，各節 能因各相鄰節點本权^代下之貧訊’而其可 和代汁异而更新，因此每次各節點之 後’則尚需確認其相鄰節點是否有更新資訊， 是轉要再啟軸疊代計算，其判斷標 斜，二s Γ即點有貝料更新之相鄰節點數及各相鄰節點之資 ’、，度’決定是賴再重新啟動該節點之内疊代計算，若 ，節點皆重新進行内疊代計算，則整個計算可收斂結 上述判斷各即點是否須重新啟動内疊代計算以及是否整體 T收歛之齡，本_之為「外疊代」，其如第人圖所示， 外登代結束制進人下-咖之計算。 6整體數值模擬流程 *體數讎擬雜如帛九圖，首先猶模式設定檔，讀入 =間切割，關資訊、水文地質參數、邊界條件與方程式集合等 貝訊’接著則依據模擬之模擬型態、起始_、結束時刻盘模 ，間距’開始進行模擬，流程中當外疊代收斂後，則進行^ 時刻域，，並依_判斷是否結束計算。若為穩態模擬，則 僅執行外豐代流程—u為非穩態模擬’則依據起始時刻、 結束時刻與模擬時刻等資訊進行判斷。整體數值模擬流程之控 制L έ外豐代之计异及時刻之前進等乃屬於第五圖中主控協 調層22負責。 7_智慧型可適性計算方法實作 如述可適性計算方法為本發明之模擬方法，而近一步的結 合人工智慧後，方成為智慧型可適性計算方法(IAC)。目前學 者專家於人工智慧領域的研究發展快速，且種類繁多，諸如類 22 201118766 ^網路(AnnUam_m_rk)、專家系統历_ system)、 m_Fuzzy TheGIy)等等’唯基於各種人工智慧之特性，應 太私j购慧型計算之部分林盡_，町纽例將說明 本^與類神經網路結合之應用。對於許多問題而言，其變化 尚未明白’但可以藉由大量聽之觀測資料，透過 e —)或資料採礦—-侧之人 日=料，建立變數與變數之相對_。對於未知的問 碭’雖有料補已可帛量化讀學办 =尚未能完整描述，因此其變化機制的控制方程= 的方程式，本發明可利用現地朗資_資 =_#人1智慧技術補足缺少財程式。其中本發明所使用 的工具是細類神經網路(A.N.N.) ’步驟如下： I. 以方程式—致性分析檢驗既存方程式集合，如果 =足=的方程絲定義基本鱗應的.守㈣間的 =聽=完備。統計各變數於各方程= =數數夏，如果其數字為〇，則表示該變數無法求解 、即賴出目制繊•巾所有無法求解之應變數。 II. 假設待補齊函數之變財n個自魏與m個 自變數與應變數皆由目前問題範脅已知 -’ 變數即前述之應變數數量為〇者，其數量^決Γ中二個應 數的數量η則尚未決定。因此 個變數，在此藉由外界採集的大量觀測資料=二 數^=自t樹倾崎峨最佳的自魏 路。ΙΠ.依據前述選定之自變數與應變數組合建立類神經網 23 201118766 IV-完成神經網路，即可在IAC中呼叫計算。 8.應用可適性計算架構於地下水模擬模式成果說明·· a本發_㈣料算㈣實作齡應用於不制網格之 暫態飽和舰含水層模擬，底下將以兩個案例說明模擬成果。 此兩案例之模擬區域為n公尺乘上ί3公尺之垂向二維方形薄 板，在邊界條件的設定上，左右邊界之總水頭均設定為80㈣， 上下邊界則没定為無流量邊界⑽F1〇w Β_~)。在材質設 疋方面，孔隙率為0.38、水力傳導係數為〇 〇1(m/day)。在初始 條件上，所有位置之勒始總水頭為如㈣，意即初始水位代 2經抽水時狀態，處於靜水麼分佈。於點位以习處配置抽水 ’並以500(kg/day)之抽水量進行抽水，模擬間距為〇 〇 iju分鐘）數為6個時刻。因此單—時刻之抽水 里為5(kg)。在網格切壯，案例一其網格配置為㈣ :形=，如第十圖所示。案例二僅在抽水井周 ?格=卜圍區域仍以見方之較粗網格為主，以節省計管 ^如第十-圖所示，如此可在—定的精度絲下，維持低^ 得到圖’不同網格尺寸均可 表，兩荦例之相對mi為案例之相對系統守恆誤差 衣兩賴之相對系統守怪誤差均極小 n 量、抽水量與系統蓄水變化量均符合質量守怪‘'# 1流 ，網狀指财，難料料可符合 在水位的呈現上可能會有不同妹 、律仁 化較為顧，傳統模擬技巧多會建議配=水井附近水位變 掌握水位變化，_奴_ =、、=格=精確 案例二所建立之峨格在糊點;^=節 24 201118766 =，計算時間方面則分別為3】5秒 例之計算_ 示，細_域加強配置一定比 〜、’透過Vor〇n(M Dmgram的空間切割方法， /的增加成本τ，有效提升計算精度，以軸兮， 二异誤差僅約0.21米深，計算時_不職例—的十^之The processing stage indicated in the seventh table represents the processing sequence of Equation 3 in this numerical method. The order of the filaments in each stage (10) is not specified, and the processing order between stages is necessarily constant. In the second and third stages, the order of the two stages _ the equations d and e are not fixed, and any - strip p first processing is feasible 'but the second v, s equation is solved. Can Ding Di - team cutting, must complete this choice = program-induced analysis process, after the record can be found to block all (10) = two not this equation is solvable, and can also be determined Equation _ The equations that are analyzed by the equation-scientific analysis can be used as slaves, and the calculation targets of the inner iterations within each lang. ^ 1 兀 兀 (node) operation: From the results of the above-mentioned symmetry analysis, we can know the order of the variable number of yarn equations, so the number of variables 夕 = is often the last solution, and the stalk of the last solution is Yes, so the equation changes; all have t: the earlier equations of this equation to evaluate the number of the left and right sides of the equal sign 'and use the test of all the variables to give the full two! (In the vicinity of the 'outside', all the equations are satisfied. If the values on both sides of the equal sign are 20 201118766 = there is a difference, then the iteration can be minimized by the iterative calculation in the seventh figure. The solution, the present invention refers to this step as "internal iteration j, seven maps." This intra-fertilization calculation is intended to calculate the variable value of each node independently, and the "internal iteration" will be further explained below. The internal iterative calculation method is used to calculate the number of nodes to be solved in each node when the abundance is calculated. The current value of each node under this condition can be solved by the steepest slope method. 'As shown in the seventh picture. The seventh picture shows the flow path of the Φ-shaped 1-side generation with the steepest slope method. In the aforementioned groundwater flow problem, the pressure head is to be solved and the number is required to give the initial solution and start searching with the steepest slope method. 'The initial pressure head must be given here, substituted into the equation set, and the pressure head of the adjacent knot is used to calculate the cross-flow 1' of the representative variable (here, groundwater) into the continuous equation to control the total crossing of the surface. The amount should be equal to the change I of the sugar control axis, and if it is not the case, it is the error. Through the difference method, the differential approximation of the conserved head value to the conservation error can be obtained, and the differential approximation ’ can be used to converge the scale convergence. When the (4) difference approaches zero, the total crossing amount of the surface of the substitution system is equal to the variation within the control volume, which means that the pressure head of the node itself has been obtained. This round of Bu Budai, this "fourth generation" calculation is the calculation of the calculation of the yuan called the wind in the fifth picture. 5. Outer iterative processing method The inner iterative calculation 'Each section is calculated independently for its variable value. When calculating, the value is near, only the current value is considered, but the value of the neighboring node may be calculated by its own ^ Because the _--upper order repeats the calculation of each node until it converges, this case calls this "outer generation". This is different from the traditional method of solving the values of all nodes in the whole region simultaneously through the matrix solution, in order to maintain the independence of the calculation of each point and the consequent elasticity. It is explained as follows: 21 201118766 Neighbor node: Point ΓΓί f In the case of internal iteration calculation, some equations need to be applied to the adjacent nodes of the phase point f, which are independent, count, each energy saving due to the power of each adjacent node ^ Under the generation of the poor news 'and it can be updated with the generation of juice, so each time after each node', it is still necessary to confirm whether its neighboring nodes have updated information, is to re-start the iterative calculation, its judgment The second s Γ Γ 有 有 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新 更新For the calculation, the whole calculation can converge to determine whether the points need to restart the inner iteration calculation and whether the overall T is the age of convergence. This is the "outer generation", which is shown in the figure of the person. End the system into the people - the calculation of the coffee. 6 The overall numerical simulation process * The number of simulations is similar to that of the nine-figure diagram. First, the mode is set in the mode, the reading is in the inter-cut, the information, the hydrogeological parameters, the boundary conditions and the equations are collected, and then the simulation is based on the simulation. Type, start _, end time disc mode, pitch 'start simulation, after the outer iteration converges in the flow, then ^ time domain, and judge whether to end the calculation according to _. In the case of steady-state simulation, only the external abundance process—u is an unsteady simulation” is determined based on information such as the start time, the end time, and the simulation time. The control of the overall numerical simulation process is not the responsibility of the main control coordination layer 22 in the fifth figure. 7_Intelligent adaptive calculation method implementation As described in the adaptive calculation method, the simulation method of the present invention, and the further integration of artificial intelligence, becomes the intelligent adaptive calculation method (IAC). At present, scholars and experts in the field of artificial intelligence research and development is fast, and a wide variety, such as class 22 201118766 ^ network (AnnUam_m_rk), expert system calendar _ system), m_Fuzzy TheGIy), etc. 'On the basis of various artificial intelligence features, should be too The private j-purchasing part of the calculation of the forest type _, the town of the new example will illustrate the application of this combination with the neural network. For many problems, the changes have not yet been understood, but the relatives of variables and variables can be established by a large number of observations, through e-) or by data mining. For the unknown question 砀 'Although it is expected that the Quantification of the School of Reading can not be fully described, so the equation of the change mechanism of the control equation =, the present invention can use the local land capital _ capital = _ #人1智慧技术Make up for the lack of financial programs. The tool used in the present invention is a fine-like neural network (ANN). The steps are as follows: I. Test the existing set of equations by equation-based analysis, if the equation of === defines the basic scales. = Listen = complete. Counting the variables in each equation == counting summer, if the number is 〇, it means that the variable can't be solved, that is, the number of strains that cannot be solved by the eyelids. II. Assume that the variable of the function to be complemented is n self-wei and m self-variables and the number of strains are known from the current problem---the variable is the number of strains mentioned above, and the number is determined by The number η of the number is not yet determined. Therefore, a variable, here a large number of observations collected by the outside world = two numbers ^ = the best self-wei road from the t-tree.建立. Create a neural network based on the combination of the selected self-variable and strain number. 23 201118766 IV-Complete the neural network, you can call the calculation in IAC. 8. Application adaptability calculation architecture in the groundwater simulation model results description·· a hair _ (four) material calculation (four) the actual age applied to the simulation of the transient saturated ship aquifer without grid, the simulation results will be explained in two cases . The simulated area of the two cases is n meters multiplied by ί3 meters of vertical two-dimensional square thin plate. In the boundary condition setting, the total head of the left and right boundaries is set to 80 (four), and the upper and lower boundaries are not defined as no flow boundary (10) F1 〇w Β_~). In terms of material design, the porosity is 0.38 and the hydraulic conductivity is 〇 〇 1 (m/day). In the initial conditions, the total head of all positions is as shown in (4), meaning that the initial water level is 2 when pumped, and it is in a static water distribution. At the point of location, the pumping is performed ‘and the pumping amount is 500 (kg/day), and the simulated spacing is 〇 〇 iju minutes). Therefore, in the single-time pumping, it is 5 (kg). In the case of the mesh, the case of the grid is configured as (4): Shape =, as shown in the tenth figure. Case 2 is only in the pumping well week? The grid area is still dominated by the coarser grid of the square, so as to save the metering ^ as shown in the tenth-picture, so that it can be kept low under the precision of the wire ^ Obtaining the graph 'different grid sizes can be listed, the relative mi of the two examples is the relative system conservative error of the case. The relative system error of the two systems is extremely small, the amount of pumping and the amount of water storage in the system are in line with the quality. Strange ''#1 stream, mesh refers to wealth, it is difficult to predict that there may be different sisters in the presentation of the water level, and the legal skills will be more concerned. The traditional simulation techniques will be recommended to match the water level near the well to master the water level change. _ slave _ =, , = 格 = accurate case 2 established by the grid in the past; ^ = section 24 201118766 =, calculation time is 3] 5 seconds of the calculation _ show, fine _ domain strengthen configuration must Than ~, 'through Vor〇n (M Dmgram's space cutting method, / increase the cost τ, effectively improve the calculation accuracy, to the axis, the difference is only about 0.21 meters deep, the calculation time _ inoperative - ten ^ It
而在第二_開發流程中，可套•第十三圖的架構 式’包含核心平台m與制漁112，其幡辞台m包 含基本必要功能’如Vo嶋iDigmm空_散模組、通用數值 模組與方程式-雜分龍組，其分顧於齡娜、求解方 程式、以及純練式的合雜與求_序。本镇技術人士 可以理解的疋，上賴組可由能夠達到相同效果的演算法模植 所取代，而不陵於上述舉出的模組。而應用模組112包含地下 25 201118766 水流模組、熱流傳輸模組與 題所定義的控财料輕雜應其他研究問 模組可依研究需求增減或替換，H未顯示於圖中），該等 後可具槪的·，而_=^圖繼建構完成 性計=發流程可擴充如第十四圖，先建構「可適 如步驟S51 構過程可由其架構開發者完成’ 含如前述内疊代、外疊代與劃分網格 荨基本必須功能。而步驟松與SS3建構延伸㈣及驗 類似，唯其僅需依照所欲解決之問題建構其： 二複建構内4代與外疊代等基本必須功能。 因而保留大1彈性並節省開發成本。 —而。’本案實為—難得—見，值得珍惜的難得發明， 上所粒’僅為本翻之最佳實施綱已，當不能以之限 疋發月所貝知之範圍。即大凡依本發明申請專利範圍所作之 ί等變化與修飾，㈣仍屬於本發明專儀蓋之麵内，謹請 貝審查委員明鑑，並祈惠准，是所至禱。 【圖式簡單說明】 ，一圖為習知技術的傳統數值模式之開發流程。 第二圖為本發明之流程示意圖。 5二圖為本發明與傳統數值方法建構流程之差異。 ，四圖為***型數值模式建構系統之示意圖。 第五圖為本發明之架構示意圖。 第六圖為本發明之系統開發階段示意圖 第七圖為内疊代流程。 第八圖為外疊代流程0 第九圖為本發明模擬流程圖。 26 201118766 第十圖為本發明案例一的網格配置示意圖。 第十一圖為本發明案例二的網格配置示意圖。 第十二圖為本明邊界流量變化圖。 第十三圖為本發明之一架構示意圖。 第十四圖為本發明之一流程圖。 【主要元件符號說明】 21計算層 211r211n計算元 22 主控協調層 23 格點 24 網格 25 人工智慧 S21-S24 步驟 S31-S34 步驟 41 可適性數值計算平台 42 方程式擴充介面 111核心平台 112應用模組 S51-S53 步驟 Γ Λ 1 27In the second development process, the architecture of the thirteenth diagram includes the core platform m and the fishing system 112, and the slogan m contains basic necessary functions such as Vo嶋iDigmm empty_scatter module, universal The numerical module and the equation-hybrid dragon group are divided into the age, the solution equation, and the pure combination of the order and the order. The technical person in the town can understand that the upper group can be replaced by an algorithm that can achieve the same effect, instead of the above-mentioned modules. The application module 112 includes the underground 25 201118766 water flow module, the heat flow transmission module and the control material defined by the problem. The other research module can be increased or decreased according to the research requirement, and H is not shown in the figure) After that, the _=^ map can be constructed. The completion process can be expanded as shown in the fourteenth figure. The first construction can be completed as follows: Step S51 can be completed by its architecture developers. The inner iteration, the outer iteration and the meshing are basically necessary functions, while the step loose is similar to the SS3 construction extension (4) and the test, but it only needs to be constructed according to the problem to be solved: 4 generations and outer layers in the second reconstruction Substitute basic functions must be preserved. Therefore, it retains the large elasticity and saves the development cost. - And. 'This case is really rare. See, it is rare to invent the rare invention, and the above is only the best implementation of this turn. It is not allowed to limit the scope of the moon's knowledge. That is to say, the changes and modifications made by Dafan according to the scope of patent application of the present invention, (4) still belong to the face of the special instrument cover of this invention, and I would like to invite the members of the review committee to understand and pray Precise, is the prayer to be. Single explanation], one picture is the development process of the traditional numerical model of the conventional technology. The second figure is the flow chart of the invention. The second figure is the difference between the construction process of the invention and the traditional numerical method. The four figures are the inserted numerical values. Schematic diagram of the model construction system. The fifth diagram is a schematic diagram of the architecture of the present invention. The sixth diagram is a schematic diagram of the system development stage of the present invention. The seventh diagram is the inner iteration process. The eighth diagram is the outer generation process 0. Inventive simulation flow chart. 26 201118766 The tenth figure is a schematic diagram of the grid configuration of Case 1 of the present invention. The eleventh figure is a schematic diagram of the grid configuration of Case 2 of the present invention. The twelfth figure is a diagram showing the change of the boundary flow rate. The three diagrams are schematic diagrams of one of the structures of the present invention. The fourteenth diagram is a flowchart of the present invention. [Description of main component symbols] 21 computation layer 211r211n computational element 22 master coordination layer 23 grid point 24 grid 25 artificial intelligence S21- S24 Steps S31-S34 Step 41 Adaptability Numerical Calculation Platform 42 Equation Expansion Interface 111 Core Platform 112 Application Module S51-S53 Step Γ Λ 1 27
Claims (1)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW098139641A TWI489400B (en) | 2009-11-20 | 2009-11-20 | Adaptive computation method and framework thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW098139641A TWI489400B (en) | 2009-11-20 | 2009-11-20 | Adaptive computation method and framework thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| TW201118766A true TW201118766A (en) | 2011-06-01 |
| TWI489400B TWI489400B (en) | 2015-06-21 |
Family
ID=44935845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| TW098139641A TWI489400B (en) | 2009-11-20 | 2009-11-20 | Adaptive computation method and framework thereof |
Country Status (1)
| Country | Link |
|---|---|
| TW (1) | TWI489400B (en) |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI318185B (en) * | 2007-07-31 | 2009-12-11 | Univ Nat Taiwan Science Tech | Online monitoring method of driver state and system thereof |
-
2009
- 2009-11-20 TW TW098139641A patent/TWI489400B/en not_active IP Right Cessation
Also Published As
| Publication number | Publication date |
|---|---|
| TWI489400B (en) | 2015-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | From channelization to restoration: Sociohydrologic modeling with changing community preferences in the K issimmee R iver B asin, F lorida | |
| Gao et al. | A hierarchy-based solution to calculate the configurational entropy of landscape gradients | |
| CN109887282A (en) | A kind of road network traffic flow prediction technique based on level timing diagram convolutional network | |
| CN107133427A (en) | A kind of construction method of the flood risk analysis model based on 2DGIS platforms | |
| Saadatpour et al. | Meta-model assisted 2D hydrodynamic and thermal simulation model (CE-QUAL-W2) in deriving optimal reservoir operational strategy in selective withdrawal scheme | |
| Xing et al. | Graph neural networks for state estimation in water distribution systems: Application of supervised and semisupervised learning | |
| Zheng et al. | Parallel computational building-chain model for rapid urban-scale energy simulation | |
| Cui et al. | Improved Thomas–Fiering and wavelet neural network models for cumulative errors reduction in reservoir inflow forecast | |
| CN109448124A (en) | Simulation of water quality method and apparatus | |
| CN106649968A (en) | Visualization design system and method for village water supply pipeline network | |
| Wu et al. | Local and global Bayesian network based model for flood prediction | |
| Davidsen et al. | The cost of ending groundwater overdraft on the North China Plain | |
| Gu et al. | Optimizing operation rules of sluices in river networks based on knowledge-driven and data-driven mechanism | |
| CN108694232A (en) | A kind of socialization recommendation method based on trusting relationship feature learning | |
| Hu et al. | Adaptive multi-scale population spatialization model constrained by multiple factors: A case study of Russia | |
| Li et al. | Analysis of the challenges and solutions of building a smart city | |
| TW201118766A (en) | Adaptive computation method and framework thereof | |
| CN201570064U (en) | City pipe network analog result data integration display system | |
| Pan et al. | APPLICATION OF A WEB-BASED DECISION SUPPORT SYSTEM FOR WATER SUPPLY NETWORKS. | |
| Ying et al. | Research on town ecological landscape planning and governance based on fuzzy optimization method of Internet of things technology | |
| Guo et al. | Research on assimilation simulation of chlorophyll a concentrations in a virtual geographic environment | |
| Schmitt | Spatial modeling issues in future smart cities | |
| Kotze | Towards total water awareness: a technology framework | |
| Daki et al. | Big-data architecture for electrical consumption forecasting in educational institutions buildings | |
| Chairez et al. | DNN-state identification of 2D distributed parameter systems |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MM4A | Annulment or lapse of patent due to non-payment of fees |