利用脈衝電源驅動之水溶液電漿建立水質檢測與機器學習電漿光譜平台

Development of Water Monitoring and Machine Learning Spectroscopy Platforms Using Plasma in Solution Driven by Pulsed Power

王靖宇(Ching-Yu Wang)

指導教授 ： 徐振哲

國立臺灣大學/工學院/化學工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201902439

摘要

水溶液電漿由於其廣泛的應用近年來被大家關注，本研究包含三個部分，分別為利用雙極脈衝電源驅動水溶液電漿特徵化電漿行為、利用水溶液電漿放射光譜建立連續微量重金屬分析系統、利用機器學習分析電漿光譜。
本研究第一部分為利用雙極脈衝電源驅動水溶液電漿特徵化電漿行為，我們使用具有不同電壓波形的脈衝電源模式以改變電漿的行為，結果顯示，使用prepulse於高壓脈衝之前，prepulse產生的氣體顯著改變電漿於高壓脈衝中的行為，此外，當使用正極之prepulse時，光譜中Pb放光強度增加10倍，此增加可能歸因於prepusle經電化學反應產生的氧氣提供不同的反應路徑，這些發現使水溶液電漿系統被更深入的理解並且提供一新穎的方法於環境分析應用。
本研究第二部分為利用水溶液電漿放射光譜建立連續微量重金屬分析系統，目標為在不同導電度之水溶液中控制電漿與放光行為，我們利用主動調控式脈衝電源驅動電漿，此電源現地測量導電度並即時根據導電度調控脈衝時間寬度，我們進行連續Pb與Zn偵測，濃度與導電度範圍分別為0.5-250 ppm與300-1200 μS/cm，結果顯示多種重金屬能在不同導電度的水溶液中被獨立且即時的偵測，我們認為應用此主動調控式脈衝電源於連續式重金屬偵測提供環境分析重大的影響。
本研究第三部分為利用機器學習分析電漿光譜，我們建立高效率取得水溶液電漿光譜的平台並使用演算法分析電漿，光譜激發於具有特定導電度與pH值範圍2.2-5.2水溶液，40k張光譜被使用於主成分分析與人工神經網路以預測水溶液導電度。主成分分析結果顯示，在第一主成分與第二主成分所形成之分數圖中，大部分數據重疊，暗示主成分分析無法分辨導電度的特徵。人工神經網路結果顯示，深層神經網路大幅增加導電度預測之準確性，均方誤差相較使用單一波長之值進步三個數量級。我們將探討使用機器學習電漿光譜特徵化電漿之意義與展望。

關鍵字

水溶液電漿 ； 電漿與水溶液交互作用 ； 放射光譜 ； 重金屬分析 ； 機器學習 ； 主成分分析 ； 人工神經網路

並列摘要

Plasmas in and in contact with liquids are attracting increasing attention for broad applications. In this work, we present the characterization of plasma in aqueos solution driven by bipolar pulsed power source, the development of an online continuous heavy metals monitoring system using optical emission spectroscopy (OES) of plasma in water, and machine learning OES of plasma.
The first part of the work presnts characterization of plasma in aqueos solution driven by bipolar pulsed power source. We applied several pulse modes with different voltage waveforms to change the plasma behavior. The results show that using a prepulse is able to generate gases for plasma ignition in the following high-voltage pulse stage, which greatly change the plasma ignition process. In addition, by using the prepulse with positive polarity, the atomic Pb emission intensity increases significantly about 10 times. This enhancement could be attributed to the application of O2 from the electrolytic reactions in the positive prepulse stage, which provides different routes for Pb emanations. These findings provide a better understanding of plasma-liquid interactions and a novel route for environmental monitoring applications.
The second part of the work presents the development of an online continuous heavy metals monitoring system using OES of plasma in water. The plasmas were driven by actively modulated pulsed power (AMPP) in order to control the plasma and its emission behavior in solutions with a wide range of electrical conductivity (EC). The AMPP quantified in situ the solutions’ EC and modulated in real time the pulse width based on the EC. We demonstrated the online monitoring of the metallic elements Pb and Zn with a concentration from 0.5 to 250 ppm in solutions with EC ranging from 300 to 1200 μS/cm. The results show that multiple metallic elements, namely Pb and Zn, can be independently and simultaneously detected with less than a 10% variation in the corresponding optical emission lines in solutions with a wide range of EC. We believe the system using plasma spectroscopy with AMPP for online monitoring of metals in water will have a significant impact on the fields of environmental monitoring and protection.
The third part of the work presents machine learning OES of plasma. A specially designed platform for efficient acquisition of spectra emanated from plasmas in solutions is developed, and several machine learning algorithms are tested for plasma analysis. We test the OES of plasmas ignited in solutions with designated ECs with pH of 2.2-5.2. A total 40k spectra are collected and tested with principal component analysis (PCA) and artificial neural network (ANN) to predict the solution’s EC. In PCA, the results show that most data points are overlapped in the score plot constructed using principal components 1 and 2, implying that PCA cannot discriminate the EC based on the spectra. In ANN, the results show that the deep ANN significantly improves the accuracy of EC prediction in terms of mean squared error by three orders of magnitude compared with the method of using single emission line.We will discuss the implication and perspectives employing machine learning to plasma spectroscopy as a route for characterizing the plasmas.

並列關鍵字

solution plasma ； plasma-liquid interaction ； optical emission spectroscopy ； heavy metals detection ； machine learning ； principal component analysis ； artificial neural network

參考文獻

1. I. Adamovich, S. D. Baalrud, A. Bogaerts, P. J. Bruggeman, M. Cappelli, V. Colombo, U. Czarnetzki, U. Ebert, J. G. Eden, P. Favia, D. B. Graves, S. Hamaguchi, G. Hieftje, M. Hori, I. D. Kaganovich, U. Kortshagen, M. J. Kushner, N. J. Mason, S. Mazouffre, S. M. Thagard, H. R. Metelmann, A. Mizuno, E. Moreau, A. B. Murphy, B. A. Niemira, G. S. Oehrlein, Z. L. Petrovic, L. C. Pitchford, Y. K. Pu, S. Rauf, O. Sakai, S. Samukawa, S. Starikovskaia, J. Tennyson, K. Terashima, M. M. Turner, M. C. M. van de Sanden, and A. Vardelle, "The 2017 Plasma Roadmap: Low temperature plasma science and technology," J. Phys. D-Appl. Phys., 50 (32), 46 (2017).

2. D. Pappas, "Status and potential of atmospheric plasma processing of materials," J. Vac. Sci. Technol. A, 29 (2), 17 (2011).

3. C. C. Hsu, M. A. Nierode, J. W. Coburn, and D. B. Graves, "Comparison of model and experiment for Ar, Ar/O-2 and Ar/O-2/Cl-2 inductively coupled plasmas," J. Phys. D-Appl. Phys., 39 (15), 3272-3284 (2006).

4. M. A. Lieberman, J. P. Booth, P. Chabert, J. M. Rax, and M. M. Turner, "Standing wave and skin effects in large-area, high-frequency capacitive discharges," Plasma Sources Sci. Technol., 11 (3), 283-293 (2002).

5. M. A. Lieberman, and S. Ashida, "Global models of pulse-power-modulated high-density, low-pressure discharges," Plasma Sources Sci. Technol., 5 (2), 145-158 (1996).