智慧科技優化直讀儀器偵測結果應用技術探討

A Study on Smart Technology for Optimizing the Application of Direct-Reading Instrument Detection Results

本研究探討智慧科技應用暴露監測的可行性，改善傳統採樣，受限於監測頻率、難以即時掌握污染物濃度變化等缺點。本研究透過智慧科技結合人工智慧（Artificial Intelligence，AI）濃度預測，透過感測器網絡架設、數據分析並進行實場驗證，探討感測器設置與應用於輔助作業環境監測之可行性。本研究選定臺南某校教室進行測試，場域長18.7 m、寬9.1 m、高3 m，並設置16臺感測器進行實驗（改變污染源位置、通風量、污染源生成率，共8組實驗），提供AI進行學習與驗證。AI濃度預測結果顯示，最佳的直讀式儀器設置為將16臺直讀式儀器減少至最少僅需3臺即可推估實場濃度，此時AI驗證每個時間序列中的平均相對誤差範圍小於10％，而隨著使用的感測器的增加，誤差也隨之越低。對於現場感測器的設置原則與布點優化之可行性，AI濃度預測方可以幫助作業環境濃度監測。

This study explores the feasibility of using smart technology for exposure monitoring, aiming to address the limitations of traditional sampling methods, such as restricted monitoring frequency and difficulty in capturing real-time changes in pollutant concentrations. By integrating smart technology with AI concentration prediction, this research investigates the setup of sensor networks, data analysis, and field verification to assess the feasibility of sensor placement for supporting environmental monitoring in workplaces. The study delopyed 16 sensors in a classroom located in Tainan, with a size of 9.1 m wide, 3 m high, and 18.7 m in length. We used eight different setups, each with various parameters, including pollutant source locations, ventilation rates, and pollutant generation rates, which were adjusted, providing data for AI learning and verification. The AI concentration prediction results indicated that the optimal setup for direct-reading instruments was reducing the 16 sensors to 3, which was sufficient for estimating real-time concentration. The AI verification showed that the average relative error in each time series was less than 10%. As the number of sensors increased, the error decreased. This suggests that the AI concentration prediction method can assist in optimizing sensor placement and points of installation for environmental monitoring in real-world work environments.