Development of Neuro-Intelligent Gas Sensing System for Water Environmental Monitering

建立了水环境监测系统。该系统由传感器阵列、测量系统和计算机组成，并采用人工神经网络算法进行气体识别。该系统用于测量河流中溶解的水中的氧、氨和硫化氢。传感器的制作和选择：制作了薄膜氧化锡半导体气体传感器。为了提高传感器的选择性，还掺杂了氧化钙和铝。研究发现，传感器对酒精、铵和乙醛有反应。薄膜式和块状溶解氧传感器也得到了发展。发现薄膜型适合一次性使用，散装型适合长期测量。商用半导体气体传感器（Figaro _<800, bbbb813, bbbb824, bbbb823）也被用于测试系统。结果表明，它们适用于氨和硫化氢的测定。测量系统：浸渍式和流量式…更多的是通过传感器组合而成的通孔式测量系统。分别用于测定固相和气相样品。前者对氨和硫化氢的检出限为2 ppm，响应时间为3分钟，后者为100 ppm，响应时间为20秒。识别算法：该系统采用了误差反向传播算法的人工神经网络，增强了气体识别能力。用标准气体对网络进行训练后，发现对硫化氢和氨的识别误差分别为2 ppm和2-25 ppm。河水的测量：利用所开发的系统对曼谷Chao-Praya河的样品进行了测量。结果发现，样品中氨和硫化氢的浓度均低于系统的检测限3ppm。综上所述，我们开发了一种采用阵列传感器和人工神经网络相结合的水质监测系统，以提高系统的性能。该系统对溶解氨和硫化氢的检测限为3ppm。我们提出，通过将浸入式传感器与流过电池系统相结合，可以对该系统进行改进，以测量100 ppb范围内的气体浓度。少

Water environmental monitoring system has been developed. The system consists of sensors array, measuring system and a computer with an artificial neural network algorithm for gas indentification. The system is designed for measuring oxygen, ammonia and hydrogen sulfide dissolved in water solution from river.1. Sensors fabrication and selection : Thin film tin oxide semiconductor gas sensors were fabricated. Calcium oxide and aluminum were doped to increase the sensor selectivity. It was found that the sensors responded to alcohol, ammonium and acetaldehyde. Thin film and bulk type dissolved oxygen sensor have been also developed. It was found that the thin film type is suitable for disposable use and the bulk type is suitable for long term measurement. The commercial semiconductor gas sensors (Figaro _<800, >813, _<824, >823) were also used for testing the system. It was found that they are suitable for measuring ammonia and hydrogen sulfide.2. Measurement system : Dip type and flow t … More hrough cell type measuring systems have been developed by combining the sensors. They were used to measure samples in solution phase and gas phase respectively. The former one has a detection limit for ammonia and hydrogen sulfide at 2 ppm with the response time of 3 minutes, while they were 100 ppm and 20 seconds for the latter.3. Recognition algorithm : The artificial neural network with the error back propagation algorithm has been used with the above system to enhance the gas recognition capability. After the network was trained with standard gas, it has been found that the recognition error for hydrogen sulfide and ammonia were 2 ppm and 2-25 ppm, respectively.4. Measurement of river water : The developed system was used to measure the samples from Chao-Praya river in Bangkok. It was found that the concentration of ammonia and hydrogen sulfide in the samples were lower than 3 ppm, which is the detection limit of the system.In summary, we have been developed a water quality monitoring system using array sensors tpgether with artificial neural network to enhance the system performance. The system has a detection limit of 3 ppm for dissolved ammonia and hydrogen sulfide. We propose that the system can be modified for measuring gas concentration in 100 ppb range by combining the dip type sensors together with flow through cell system. Less