首页> 外文期刊>Adsorption Science & Technology >Towards the Prediction of the Remaining Lifetime of a Carbon Filter Bed on Hydrogen Cyanide Challenge Using In-bed Sensors and a Neural Network: Preliminary Findings
【24h】

Towards the Prediction of the Remaining Lifetime of a Carbon Filter Bed on Hydrogen Cyanide Challenge Using In-bed Sensors and a Neural Network: Preliminary Findings

机译:使用床内传感器和神经网络预测氰化氢挑战下碳滤床的剩余寿命:初步发现

获取原文
获取原文并翻译 | 示例
       

摘要

The residual lifetime of an activated carbon filter bed challenged by hydrogen cyanide, a chemisorbing gas, has been successfully predicted on one occasion using transducers embedded within the filter bed. No thermostatting nor relative humidity control of the bed or influent challenge was imposed. Four thermo-couples and a fibre optic pH probe, sensitised to HCN by the addition of cobalt(II) chloride, were used to measure the progress along the filter bed of the adsorption front resulting from intermittent HCN challenges of 0.9 mg/dm~3. Twenty randomised neural networks for cyanogen breakthrough prediction were created. These were based on a simple system, with the occurrences of maximum temperatures around two thermo-couples at the front of the filter bed providing the input variables and the cyanogen breakthrough time (BT) being the response variable. Temperature maxima were thought to be due to cyanogen hydrolysis. Each network was trained with the data from four completed challenges and then interrogated with the input variables from a fifth. The mean response variable produced was 36.5d (±2.1 d standard deviation). This represented 94% of the actual observed time to breakthrough which occurrred on challenge day 39. Averaging of network responses was required because four training examples only loosely map the possible variable space such that the potential error in any one network prediction is large. This preliminary set of experiments is most encouraging, having successfully predicted the lifetime of the bed whilst it was only half exhausted. The fact that, with no climatic control, simple sensors still permitted this achievement under realistic conditions of intermittent, long-term exposure illustrates the potential of this approach to residual lifetime monitoring. Further studies to repeat this success and extend the work to other challenge gases are now required.
机译:使用嵌入在滤床内的换能器已经成功地预测了受到化学吸附气体氰化氢挑战的活性炭滤床的剩余寿命。无需对床进行恒温或相对湿度控制或进水挑战。通过添加氯化钴(II)对HCN敏感的四个热电偶和一个光纤pH探针用于测量0.9 mg / dm〜3的间歇性HCN激发导致沿吸附前沿滤床的进展。创建了二十个随机的神经网络,用于氰化物的突破预测。这些是基于一个简单的系统,在滤床前部的两个热电偶附近会出现最高温度,从而提供输入变量,而氰化物突破时间(BT)是响应变量。认为最高温度是由于氰的水解。每个网络都接受了来自四个已完成挑战的数据的训练,然后接受了来自第五个挑战的输入变量的询问。产生的平均响应变量为36.5d(标准偏差为±2.1 d)。这代表了在挑战第39天发生的实际突破时间的94%。需要平均网络响应,因为四个训练示例仅宽松地映射了可能的变量空间,因此任何一个网络预测中的潜在误差都很大。初步的实验是最令人鼓舞的,它成功地预测了床的寿命,而床只有一半被用尽。在没有气候控制的情况下,简单的传感器仍然可以在间歇性,长期暴露的现实条件下实现这一目标,这一事实说明了这种方法对剩余寿命监控的潜力。现在需要进一步研究以重复这一成功,并将工作扩展到其他挑战性气体。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号