2021臺灣塑橡膠機械設備產值出口額超過10億美金。面對近年因國際競爭、缺工問題、商品少量多樣與減碳節能的挑戰,臺灣對設備智慧化技術的需求亦與日俱增。智慧化的基礎技術包含感測器與數據的收集、通訊與資訊模型、智慧數據分析算法等。製程參數地圖結合異常檢測技術、度量學習與降維技術、機器學習方法、結合現場專家知識,透過製程參數地圖可提供直觀的肇因分析與即時警示工具,推估原因並避免不良品持續生產造成的浪費。整個塑橡膠射出成型的過程包括提取關鍵特徵、由高維數據的距離度量進行資料清理、降維度提供直觀圖像、與行業專業人員協作標記地圖等等。期望以此技術對未來設備智慧化中的問題定義、技術組合方式、提供累積的經驗與更多啟發。
In 2021, the export value of Taiwan's plastic and rubber machinery and equipment has exceeded US$1 billion. Due to international competition, shortage of workers, small quantity and variety of products, and international requirements for carbon reduction and energy saving, Taiwan's demand for intelligent equipment technology is increasing. The basic technologies of intelligence include sensors, data collection, communication, information models, and intelligent data analysis algorithms. The parameter map combines anomaly detection technology, metric learning and dimensionality reduction technology, machine learning methods, and domain knowledge to provide intuitive cause analysis and instant warning tools to analyze possible causes and avoid waste caused by continuous production of defective products. The entire process of plastic injection molding is featured by: feature extraction, data cleaning performed through distance metric for high-dimensional data, intuitive pictures provided by dimensionality reduction, and interactive map labeling by industry professionals. It is expected that this technology will provide accumulated experience and more inspirations for the problem definition and technology combination strategies for intelligent equipment technology in the future.