200830131 九、發明說明： 【發明所屬之技術領域】 本發明所屬技術有：（1)首先採用CATIA程式做為手錶造型之3D圖形 表現基礎设计軟體’（2)次將統什學之問卷调查與分析結果，以感性工學手 法、融入徑向基類神經網路理論，（3)以數學軟體Matlab執行運算，（4)將消 費者意向與設計元素之數學函數關係建立，（5)同時利用CATIA内建之％ 〇 Editor撰寫出簡單且直覺化操作之人機介面應用系統，以便於使用者輪入資 料，（6)最後以Access建立並連結資料庫，而完成此省時簡單之設計專家系 統。 【先前技術】 一、徑向基類神經網路200830131 IX. Description of the invention: [Technical field to which the invention pertains] The technology to which the present invention pertains is as follows: (1) Firstly, a CATIA program is used as a basic design software for 3D graphics performance of a watch model' (2). The results of the analysis are based on the sensible engineering method, incorporating the radial basis neural network theory, (3) performing the operation with the mathematical software Matlab, and (4) establishing the mathematical function relationship between the consumer intention and the design element, (5) simultaneously utilizing CATIA built-in% 〇Editor writes a simple and intuitive human-machine interface application system, so that users can turn in data, (6) finally establish and link the database with Access, and complete this time-saving simple design expert system. [Prior Art] 1. Radial Basis Neural Network
Basis Function Neural Network, RBFNN)，其特質在與觀大腦皮質層軸突的局部調整功能，具備相當良好 的映射能力；其架構與多層感知機相同，具有輸人層、輸出層、一層隱藏 層與輸出層，如圖表-所示，是屬於基本前饋式類神經網路的組合，其優 點在於可大量減少學習時間；但不同於倒傳遞轉經網路，丽順是以函 數逼近(CurveFitting)的方式來建構網路；其訓練流程如圖表二。 當資料輸人網路後，直接由輸人層將輪人向量傳給隱藏層巾的每個徑 向基函數，也就是計异輸入向畺與隱藏層各神經元中心點的距離後 ，經函 200830131 元的輪出如下式: 數轉換獲得隱藏層各神經 式中R(·)表徑向基函數， &間之歐氏距離 出值如(2)式： zj(x) = RQx-Cj (I) (1) ||^〜°/||表示\與 即可求得網路輪 表隱藏層第j個神經元中心點 將L藏層的輪出直經加權傳至輸出層， 、 y ^ Σ w j(x) 式甲y為輸出層的輪出值^ (2) ，為fe藏層弟j個神經元至輪出屉 隱藏層第j個神經元的輪# ^重值，&為 輯出值。將⑴式代入(2)式可寫成(3)式·· ^twjR(\\x^cj ll)^w〇 在RBFNN情人層是將I資難晴連結的介面層，再 非線性活化函數轉換到障菇 貝千4、、工化 至丨__ 就是將輸入空間w)進行非線性映射 到隱藏層空間，，认，t 輪出層則是將隱藏層的輪出進行線性組線 性映射)獲得輸出值。 感性工學 長久、來❸询麵行設肛作時，總是將其杨滩的經驗作為設 計新產品的要素，但隨著產品的形式越趨於錄化，單純從機能以及品質 方面已不再月匕滿足消費者對於產品的要求，如何將消費者的感性應用於 產品的設計當中，才是設計師所需著重的問題。「感性工學(K_i Engineering)」的提出，為感性研究開啟了一個新的方向。 _感性工學之由來 「感性工學」這個名詞是由日本馬自達(Mazda)汽車公司的山本建一 200830131 _年在世界汽車技術會議、美國汽車產業經營者研討會之演 =之=_纖物細’蝴咖「―須能夠對文 興之氣有所貢獻」為重點，展開乘車文化論。並且提案運用「感性工 子」之手法進行乘域與汽車繼計，使符合驗者料、雜要求。 口设計師進行設計時’通常依其過往之經驗以黑箱作業的手法進行產 品屬性之轉。但是领著時代㈣雜、舒耻，對這 也不再是單純只限於㈣魏賴品餅，騎要求紅桃:足我們 感:層面上的需求。換言之，在重視生活的感覺與符合人們感受的時代 潮概中’為了使人們能夠接受某產品，基本必須探究的問題就是「人們 的感性疋什麼？」。但是在這簡單問題的背後’所包含的問題卻有許多， 例如感性是什麼？組成要素是什麼？如何測定人的感性？相關的實 驗、調查步驟？如何進行統計分析？如何將感性轉化為設計要素？等基 本問題’以至於實祕方_紐式料_發程料皆是值得深入^ 时的問題。 _感性工學之操作方法 感性工學之型式共有下列四種： (1) 第一型感性分類。 將消費者之感性需求連結至產品性質，以樹狀結構表現之。 (2) 第二型感性工學系統。 目的乃找尋感性語彙、產品形象、設計。 (3) 第三型混合感性工學系統。 200830131 可將設計產品導入推理引擎以獲得對應之感性語彙。 (4)虛擬感性工學系統。 整合虛擬實境與資料收集系統以展現實際產品。 本發明採用弟二型之雙向混合感性工學系統，其操作程序可以圖表三表 示之。 【發明内容】 本發明係採用Dassault Systems公司出產之CATIA程式之VB Editor開 發工具作為設計軟體，且利用Matlab軟體來進行人工智慧相關理論之模 擬，再將結果建立預測資料庫，其中人工智慧技術為徑向基類神經網路理 論相關之數學模型開發成系統，d以達成功能完善、可做資料輸入、感性 語彙預測及產品進階設計之目的。 首先，本發明擁有感性語彙預測產品造型與經由產品造型預測感性語 彙兩大部分。首先在感性語彙預測產品造型部分，此部份藉由使用者輸入 感性語彙分數，可經由系統與資料庫連結預測出與語彙相對應產品之造型 進而進一步設計產品。另一部份是產品造型預測感性語彙，此部份是經由 設計者輸入產品造型，可經由系統預測出消費者對於產品造型之感性語彙 分數，進而修改產品設計。因此，本發明之首要目的為提供一種功能完善、 可由造型語彙預測產品造型與產品造型預測出感性語彙之雙向預測軟體資 訊系統。本發明之次要目的乃提供可降低程式操作複雜度，提昇分析效能 之人機介面，應用此簡單且直覺化操作之人機介面，可便利使用者做資料 200830131 輸入與後續產品設計之工作。本發明將感性工學理論與徑向基類神經網路 結合，並利用CATIA之VB Editor程式設計整合撰寫人機介面，具有效降 低程式操作複雜度及預測之效能，且能以友善人機介面方式呈現。 本發明之整體程式操作流程、感性語彙預測造型參數操作流程及造型 參數預測感性語彙操作流程等分別顯示於圖表四、五、六；為達成上述目 的以及所採用之相關技術、手段與其他功效，茲列舉一較佳可實施例並配 〃 合圖示加以詳細說明之後，相信本案發明之目的、特徵及其他優點，當可 V」 .· 由之得一深入而具體之瞭解。 【實施方式】 本系統分為感性語彙分數預測造型參數及造塑參數預測感性語彙分數 兩大部分。 一、造型參數預測感性語彙分數 步驟一進入系統首頁，點選“進入系統”，如圖表七。 步驟二觀看說明晝面，如圖表八，可點選“詳細說明’’進一步觀看詳細 手錶之造型參數，如圖表九；點選“進入系統”可進入程式主晝 面’如圖表十。 步驟三首先先點選“造型參數預測感性語彙，’，即可在Formparameters 内輸入造型參數，輸入完畢點選“確定’’，會同時出現預測之感 性語彙及已輸入參數所對應之手錶造梨。 11 200830131 二、感性語彙分數預測造型參數 步驟一進入系統首頁，點選“進入系統，’，如圖表七。 步驟二觀看說明畫面，如圖表八，可點選“詳細說明，，進一步觀看詳細 手錶之造型參數，如圖表九；點選“進入系統，，可進入程式主畫 面，如圖表十。 步驟三首先先點選“感性語彙預測造型參數，，，即可在Kansei words内 輸入感性語彙分數，輸入完畢點選“確定，，，會出現預測之手錶 造型及已輸入之語彙分數，如圖表十；接著可點選“CATIA進階 設計”進入CATIA做進一步的造型修改，如圖表十一。 【圖式簡單說明】 圖表一、徑向基類神經網路架構圖 圖表二、徑向基類神經網路訓練流程圖 圖表三、感性工學之操作程序 (、 圖表四、整體程式操作流程圖 圖表五、感性語彙預測造型參數操作流程圖 .圖表六、造型參數預測感性語彙操作流程圖 圖表七、手錶造型設計專家系統主晝面 圖表八、專家系統說明晝面 圖表九、詳細造型解構說明晝面 圖表十、手錶造型設計專家系統使用者輸入晝面 圖表十一、CATIA進階造型設計晝面 12Basis Function Neural Network (RBFNN), which has a good local mapping function with the axon of the cerebral cortex, has a good mapping ability; its architecture is the same as that of the multilayer perceptron, with input layer, output layer, and a hidden layer. The output layer, as shown in the graph--is a combination of basic feedforward-like neural networks, which has the advantage of greatly reducing the learning time; but unlike the reverse-transfer-transferred network, Lishun is a function approaching (CurveFitting) The way to build the network; its training process is shown in Figure 2. After the data is input into the network, the wheel-person vector is directly transmitted to each radial basis function of the hidden layer towel by the input layer, that is, after the distance of the input input to the center point of each neuron of the hidden layer is The round of the letter 200830131 is as follows: The number conversion obtains the radial basis function of the R(·) table in each neuron of the hidden layer, and the Euclidean distance between the & is the value of (2): zj(x) = RQx- Cj (I) (1) ||^~°/|| indicates that \ and can be obtained from the hidden layer of the network wheel table. The j-th neuron center point passes the L-layer's round-out weight to the output layer, y ^ Σ wj(x) Formula A is the output value of the output layer ^ (2), which is the round of the j-th neuron of the hidden layer of the j-th neuron to the wheel of the drawer. #^重值, &; for the value of the compilation. Substituting (1) into (2) can be written as (3) ··^twjR(\\x^cj ll)^w〇 In the RBFNN lover layer is the interface layer that connects I 难 晴 晴, and then the nonlinear activation function conversion To the typhoon shell, 4, and to __, the input space w) is nonlinearly mapped to the hidden layer space, and the t-round layer is the linear group linear mapping of the hidden layer. Get the output value. Sensible engineering for a long time, when you come to consult the face to set up an anus, always take the experience of Yangtan as an element of designing new products, but as the form of the product becomes more and more recorded, it is no longer in terms of function and quality. It is the problem that designers need to focus on to meet the requirements of consumers for products. How to apply the sensibility of consumers to the design of products is the problem that designers need to focus on. The introduction of "K_i Engineering" has opened a new direction for perceptual research. _The origin of sensible engineering The term "sensible engineering" is the result of the Japanese Mazda Motor Company's Yamamoto Kenichi 200830131 _ at the World Automotive Technology Conference, the American Automotive Industry Operators Seminar = _ The fine 'frozen coffee' - the need to be able to contribute to the spirit of Wen Xing" focuses on the theory of ride culture. In addition, the proposal uses the method of “sense workers” to carry out the multiplication and vehicle relay, so that it meets the requirements of the examiner and the miscellaneous requirements. When designing a design by a mouth designer, it is usually based on the experience of the past to carry out the product attributes in a black box operation. But leading the times (four) miscellaneous, shameful, this is no longer simply limited to (four) Wei Lai pie, riding a request for the red peach: enough of our sense: the level of demand. In other words, in the era of emphasizing the feeling of life and the feelings of people's feelings, the basic question that must be explored in order to enable people to accept a product is "what is the sensibility of people?" But behind the simple question, there are many problems involved, such as what is sensibility? What are the components? How to determine the sensibility of people? Related experiments and investigation steps? How to conduct statistical analysis? How to transform sensibility into design elements? The basic question is so that the secrets of the _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _Inductive engineering methods There are four types of perceptual engineering: (1) The first type of emotional classification. Linking the emotional needs of consumers to the nature of the product, expressed in a tree structure. (2) The second type of inductive engineering system. The purpose is to find sensuous vocabulary, product image, and design. (3) The third type of mixed-inductive engineering system. 200830131 The design product can be imported into the inference engine to obtain the corresponding perceptual vocabulary. (4) Virtual sensory engineering system. Integrate virtual reality and data collection systems to showcase real products. The invention adopts the two-way hybrid inductive engineering system of the second type, and the operation procedure can be expressed in the third chart. SUMMARY OF THE INVENTION The present invention uses the VB Editor development tool of the CATIA program produced by Dassault Systems as a design software, and uses the Matlab software to simulate the artificial intelligence related theory, and then builds a prediction database, wherein the artificial intelligence technology is The mathematical model related to the radial basis-based neural network theory is developed into a system, which achieves the functions of perfect function, data input, perceptual vocabulary prediction and advanced product design. First, the present invention has a sensible vocabulary to predict product styling and predictive vocabulary through product modeling. First, in the perceptual vocabulary, the product styling part is predicted. This part can further design the product by inputting the perceptual vocabulary scores through the system and the database to predict the styling of the product corresponding to the vocabulary. The other part is the product sensation predictive vocabulary. This part is based on the designer's input product model, which can predict the consumer's sensible vocabulary scores on the product model and modify the product design. Therefore, the primary object of the present invention is to provide a two-way predictive software information system that is fully functional and can predict sensation vocabulary from product styling and product styling. The secondary object of the present invention is to provide a human-machine interface that can reduce the complexity of program operation and improve the analysis efficiency. The human-machine interface of this simple and intuitive operation can facilitate the user to do the work of 200830131 input and subsequent product design. The invention combines the theory of sensible engineering with the radial base type neural network, and uses CATIA's VB Editor program to design and integrate the human-machine interface, which has the effect of reducing the complexity of the program operation and the prediction performance, and can be friendly human-machine interface. Way to present. The overall program operation flow, the perceptual vocabulary prediction modeling parameter operation flow and the modeling parameter prediction perceptual vocabulary operation flow of the present invention are respectively shown in the figures 4, 5 and 6 respectively; in order to achieve the above objectives and related technologies, means and other effects, Having described the preferred embodiments and the detailed description, it is believed that the objects, features and other advantages of the present invention can be obtained from the detailed description. [Embodiment] The system is divided into two parts: the perceptual vocabulary score prediction modeling parameter and the plasticity parameter prediction perceptual vocabulary score. First, the modeling parameters predict the sensibility vocabulary score Step 1 Enter the system home page, click "Enter System", as shown in Figure 7. Step 2 View the description, as shown in Figure 8, you can click on “Detailed Description” to further view the detailed modeling parameters of the watch, as shown in Figure IX; click “Enter System” to enter the main surface of the program as shown in Figure 10. Step 3 First, select “Model parameters to predict perceptual vocabulary,” and you can enter the modeling parameters in Formparameters. When you click “OK”, you will see the predicted sentiment vocabulary and the watch pears corresponding to the input parameters. 200830131 Second, the sensibility vocabulary score prediction modeling parameters Step 1 Enter the system home page, click "Enter system," as shown in Figure 7. Step 2 View the explanation screen. As shown in Figure 8, you can click “Detailed Description, and further view the detailed modeling parameters of the watch, as shown in Figure IX. Click “Enter System” to enter the main screen of the program, as shown in Figure 10. Step 3 First, select “Sensor vocabulary prediction modeling parameters, and you can input the sensibility vocabulary score in Kansei words. After inputting, click “OK,”, the predicted watch shape and the input vocabulary score, such as chart Ten; then click on "CATIA Advanced Design" to enter CATIA for further modeling changes, as shown in Figure 11. [Simple diagram of the diagram] Chart 1, radial base class neural network architecture diagram 2, radial base class neural network training flow chart diagram 3, perceptual engineering operation procedures (, chart 4, overall program operation flow chart Chart 5, perceptual vocabulary prediction modeling parameter operation flow chart. Chart 6, modeling parameters prediction sensibility vocabulary operation flow chart chart VII, watch styling design expert system main 图表 surface chart eight, expert system description 图表 surface chart IX, detailed shape deconstruction description 昼Face chart ten, watch design expert system user input face chart eleven, CATIA advanced design design face 12