以類神經網路建構之染缸排程系統規劃

Abstract

國內染整廠多屬中､小型衛星加工廠。由於資本並不雄厚﹐因此投資在 廢水處理設備的費用﹐在無法回收投資報酬情況下﹐投資意願不高。且不 論其製程是否為印花､壓染或浸染﹐一般均未考慮工廠廢水污染程度差異 性﹐而大量地採用生物處理法或化學混凝浮除併用法減少染整廢水之污染 。這些方法雖然治標但仍對環境造成污染。因此本研究針對改善染整場的 廢水污染﹐提出治本地從改變其染色製程以減少染缸排放廢水的頻率方面 著手。其中本研究將運用電腦科技中之人工智慧技術﹐以類神經網路結合 基因演算法嘗試學習老師父之染整研判的經驗﹐將之用於改善染整排缸的 排程﹐以期能減少染劑污水的排放。若因此減少廢水之排放﹐不但對企業 最有利﹐且使污染程度降到最低﹐還可還給台灣美麗的河川。本研究所建 構之類神經網路將融合染整師父對染缸處理之判斷經驗﹐再依此判斷經驗 的掌握﹐結合合理的排程管理規則以建立最適化的染色排程之順序。本研 究經由類神經網路之學習測試結果﹐歸納出影響染缸處理之重要輸入因子 ﹐其影響程度由大至小依序為前後顏色之亮度差､彩度差､色相差與總色 差。而其他輸入因子亦有其影響程度﹐但相較之下較不顯著。最後本研究 藉由類神經網路所學得輸入因子之行為特性﹐對待染布匹與染缸狀況的相 容性分析結果﹐再結合排程規則之設計發現﹐經由此種布匹染色排程方式 的改善可明顯地減少染缸廢水之排放。對染整業而言不但因而節省染劑﹐ 更對環境保護做出重大之貢獻。 In this thesis, a efficient method is proposed to deal with the effluent pollution of dye houses. This method reduces the draining frequency of the bacs by changing the dying process. Artificial intelligence techniques in computer science category are used here. In this thesis, we try to learn the experience of the dyer by using a neural network combined with genetic algorithm, and apply it to improve the scheduling of the bacs as to reduce the effluent of the colorants.The neural network constructed in this thesis contains the experience of the dyer in the management of the bacs. Then a proper bac scheduling is constructed by combining some reasonable management rules.By the learning and testing results of the neural network constructed in the thesis, some important factors that influence the handling of the bacs are generalized. These factors are the difference of the two colors on the lightness, chroma, hue and the total color difference. Other factors also influence the results, but are not significant relatively. The last part of the thesis, a scheduling method is proposed by combining some management rules to the neural network. This scheduling evidently reduce the draining of the dying effluent.In summary, the method proposed here not only reducing the cost of dye works but also have a large contribution to the environment.