產品品質之預測與改善對策—以不鏽鋼線材製造為例

在產品製造時，企業對於生產的成本與產出的效益都十分的重視。在鋼鐵工業的產品加工上，對於產品製造的效率與效能都會有嚴格的管控，因其對於企業本體來說，有良好的獲益績效才是企業得以在市場中生存的基礎。因此提升產品品質，盡可能減少因表面缺陷而降成次級品或廢品的數量便成為最重要的議題之一。而不鏽鋼線材表面刮傷是目前一直無法有效改善的缺陷之一。因此有必要藉由智慧化工廠概念的導入，以自動化方式來辨識產品良率，排除使用大量的人工模式，獲得高品質的產品輸出並獲得多方廠商的青睞。本研究以智慧化工廠為核心將軋延參數與要因分析導入製程，並以振動預測應用於刮傷分析。為因應未來產量的提升，並維持產品品質，需優化相關製程，以達到生產製程平穩順暢性以及降低生產線異常事故發生。從根源找出其中影響盤元刮傷品質要因，建立可控關鍵要因及最適製程參數表，由製程最佳參數表來測試並進行作業。本研究結果顯示，經由最佳化數據及現場產線熱軋，盤元刮傷率由D8.0~12.0mm S316Li/LG、S31630、S3043X、S310X 刮傷NG 率由2020 年9 月的29.0%，最佳化參數上線後降至7.0%計算至2021 年4 月，大幅下降了22.0%。藉由降低盤元刮傷率並建立刮傷診斷系統，以經驗法則或數據分析提供相應的參數調整方法，讓原本人工品質監測的「事後分析改善」，進一步到「事中監控改善」，並由振動訊號診斷品質減少人工的干預獲得智能化的提昇。通過綜合時間因素和產品製造參數的收集，以及監控時間序列中異常事件的特徵模式和趨勢，可以提前預測生產異常是否以及何時發生。藉此系統性的導入分析模式能夠具有更高的靈活性、準確性和更少的計算時間，可以處理多源數據，分析並動態調整製造過程。

關鍵字

刮傷；智慧化工廠；振動；經驗法則；診斷系統

並列摘要

The enterprises attach great importance to the cost of production and the benefits of output when manufacturing products. In the product processing of the iron and steel industry, the efficiency and effectiveness of product manufacturing will be strictly controlled, because for the enterprise itself, good profit performance is the basis for the enterprise to survive in the market. Therefore, improving product quality and minimizing the number of inferior or scrap products due to surface defects have become one of the most important issues in the production line. However, scratching on the surface of stainless steel wire is one of the defects that cannot be effectively improved. It is necessary to introduce the concept of smart factory, identify product yield in an automated way, and eliminate the use of a large number of manual modes, so as to obtain high-quality product output and gain the favor of many downstream manufacturers. In this study, the smart factory as the core to introduce the factor analysis of rolling parameters into the process, and applies vibration prediction to the analysis of scratches. In order to cope with the increase in production in the future and maintain product quality, it is necessary to optimize the relevant processes to achieve smoothness of the production process and reduce the occurrence of abnormal production line accidents. In addition, we also try to find out the factors that affect the scratch quality of the disk element from the root cause, establish the controllable key factors, and establish the optimal process parameter table. The optimal process parameter table is used to test and operate. The results of this study show that through optimized data and on-site production line hot rolling, the scratch rate of S316Li/LG, S31630, S3043X, and S310X has increased from D8.0~12.0mm, and the scratch rate of NG has increased from 29.0% in September 2020. , the optimized parameters dropped to 7.0% after going online and calculated to April 2021, a sharp drop of 22.0%. By reducing the scratch rate of disk elements and establishing a scratch diagnosis system, and providing corresponding parameter adjustment methods through empirical rules or data analysis, the original "postanalysis improvement" of manual quality monitoring can be further improved to "in-process monitoring and improvement". And the quality of vibration signal diagnosis reduces manual intervention to obtain intelligent improvement. By integrating the collection of time factors and product manufacturing parameters, and by monitoring the characteristic patterns and trends of abnormal events in the time series, it is possible to predict in advance whether and when production abnormalities will occur. With higher flexibility, accuracy and less computation time, the systematic import of analysis modes can process multi-source data, analyze and dynamically adjust the manufacturing process.