- 學位論文
使用紅外光譜儀進行動態吸附實驗並以類神經網路建立突破曲線預測模式
Dynamic adsorption study by a gas phase FTIR and prediction of breakthrough curves using the neural
摘要
在動態吸附實驗中，以傅立業紅外線光譜儀取代傳統中使用氣相層析儀為氣體濃度分析儀器，更能夠快速且有效率的偵測濃度的變化。在實驗中可以發現，突破曲線受到氣體濃度、氣體流速、吸附材重量甚至水氣的影響，突破時間因氣體濃度和氣體流速的增加而提前，而吸附材的重量增加會使突破時間延後。水氣也是影響到突破時間的重要因素，水氣的濃度越大，突破時間也會越早而所造成的roll-over的現象也就愈明顯。半經驗式Yoon and Nelson的模式所迴歸本系統中所探討的突破曲線在實驗值與回歸值間有良好的一致性。 此外類神經網路結構則提供了一個良好模式，除了能夠模擬Yoon and Nelson模式中的κ’、τ值，更能夠預測到不同情況下的κ’、τ值。另外類神經網路結構成功的模擬了在不同條件下的突破曲線，對於不同水氣影響下的競爭吸附，也能夠提供良好的模擬結果，並且提供了良好的預測結果。
並列摘要
Instead of the traditional gas chromatography(GC), a gas-phase Fourier Transformation-IR（FT-IR）spectrometer was used to measure the evolution of concentrations of these volatile organics due to its quick response to the instantaneous changes of the gas concentration. The results show that breakthrough curves were effected by the initial concentration, the flow rate, the adsorbent weight and the relative humidity. And the breakthrough time decreases with the increase of initial concentration, and flow rate and the decrease of adsorbent weight. The relative humidity also influence the breakthrough time and cause phenomenon of roll over. The higher relative humidity the higher roll over is. An empirical model proposed by Yoon and Nelson was applied to fit the experimental data. The results showed that the model can fit the experimental data well under various operating conditions and the standard deviation values are small. Neural network provided a good model to simulation the Yoon and Nelson’s κ’ and τ values and it also can predict the κ’ and τ value at different conditions well. In addition to, the neural network can simulate and predict the breakthrough curves well. At the competition adsorption of VOCs and water vapor, the neural network technology provides good predictions.