近年來國內外有關食品安全疑慮之重大事件接二連三爆發,從2008年之三 聚氰胺毒奶粉事件、2011年之塑化劑風暴、2012年的美國牛肉瘦肉精容許殘量之爭議,乃至今2013年順丁烯二酸的毒澱粉事件等,無不造成人心惶惶。而上游所產生的漏洞仍以工業原料未管控流向,中游漏洞則以食品添加物廠商登記不力,且知道有問題卻未通報衛生機關和下游,而政府主管機關單位於事後也無從查起原料、半成品的流向,只能採消極的方式來查驗,凸顯台灣對食品生產鏈追溯與登載產生漏洞百出的現象,連事後的管控也沒有辦法查出原料的流向。因此,產品履歷與食品生產鏈追溯則成為值得深入探討的問題。 本研究以知識本體發展產品履歷模型應用於近期食品生產鏈所發生的「順丁烯二酸」工業用添加物事件為實驗個案,對此食品業在生產鏈關係的分類,由下游至上游劃分為四層,建置公司資料、商品資料及追蹤方式的知識本體,前二者定義知識源之間的邏輯關係,後者為追蹤查詢應用而設計。透過知識塑模與表達,發展語意規則建立公司、商品及追蹤方式三者之間的關係,將追蹤方式推論結果,成就最佳「源頭管理」理念。 利用推論機制執行規則,核對各項屬性內容值,測試結果均與人工模擬作業的結果一致。另驗證系統的二項效能指標─擴充性及可信賴度,則由推論機制重新推導知識結構,經核對均正確無誤,並分別更動數項實例的內容值,觀察推論機制是否重建整個知識結構,可以產生連鎖的知識更新。
Concerning about food safety problems in Taiwan and overseas in recent years, the significant event continues to happen one by one. From the 2008's melamine poison milk event, in 2011 the plasticizer storm and government allows U.S. beef ractopamine residue amounts of controversy in 2012. Since 2013’s poisonous starch events, etc., those events are all caused people panic and worry. The problem for upstream is that they do not control the flow of industrial materials. Midstream problems are food additives firms registered ineffective, but Midstream has not informed those problems to downstream and health authorities. The units of government authorities have no way to check up on afterwards raw materials, semi-finished products flow and they can only take a negative way to check. Highlighting the big problems in control of raw materials in Taiwan with the phenomenon of posting loopholes, and even afterwards there is no way to control the flow of raw materials identified. Therefore, production resumes and supply chain become worthy of further discussion. In this study, the ontology model of the development of production resumes should resume with recent food safety incidents of illegal use of industrial additives "maleic acid" as experimental cases. This relationship between the food industry in the production and marketing chain classification from downstream to upstream is divided into four layers, build company information, product information and tracking method ontology. The former two define the logical relationship between knowledge sources; The latter are designed to track queries applications. Through knowledge modeling and expression, semantic rules are developed and establish the relationship between companies, products, and tracking methods. The result of tracking methods corollary to achieve the best "source management" philosophy. The use of inference mechanisms enforcing rules, check the value of the attribute content. The test results are the same with artificial simulations consistent. Moreover, after checking, it is correct to re-derive knowledge structures by verification system another two performance index ─expansion and trustworthiness. And we also change the content of a number of instances to observe whether the inference mechanism to rebuild the entire structure of knowledge which can generate the knowledge chain updates.