动态混合气体数据集下的气体传感器阵列

Data Set Information: 该数据集包含从16个暴露于不同浓度水平的气体混合物中的化学传感器获取的时间序列。特别是，我们产生了两种气体混合物：空气中的乙烯和甲烷，以及空气中的乙烯和一氧化碳。每次测量均通过连续采集16个传感器阵列信号进行，持续时间约为12小时，无中断。 该数据集是在加利福尼亚大学圣地亚哥分校的BioCiCuiTM研究所化学化学实验室的气体输送平台设备中收集的。测量系统平台提供多功能性，以高精度和高度可重复的方式获得所需的感兴趣化学物质浓度。 传感器阵列包括4种不同类型的16个化学传感器（美国费加罗公司）：TGS-2600、TGS-2602、TGS-2610、TGS-2620（每种类型4个单元）。传感器集成了定制的信号调节和控制电子设备。控制传感器的传感器工作电压A€？在整个实验过程中，工作温度保持恒定在5 V。传感器A€？以100 Hz的采样频率连续采集电导率。将传感器阵列放置在60 ml测量室中，在该测量室中以300 ml/min的恒定流量注入气体样品。 在浓度水平随机变化时，通过连续采集16个传感器阵列信号进行每次测量。对于每次测量（每种气体混合物），连续采集信号约12小时，无中断。 在随机时间（间隔80-120秒）和随机浓度水平设置浓度转换。数据集的构建应确保存在所有可能的转换：增加、减少一种挥发物的浓度或将其设置为零，而另一种挥发物的浓度保持恒定（在固定或零浓度水平）。在开始、结束和大约每10000秒，我们插入额外的预定义纯气体混合物浓度模式。 选择乙烯、甲烷和CO的浓度范围，使传感器响应的感应幅度相似。此外，对于气体混合物，较低的浓度水平是有利的。因此，传感器对所述刺激集的多变量响应是具有挑战性的，因为没有任何配置（单个 气体或混合物的呈现）可以很容易地从传感器A€？响应。尤其是乙烯浓度范围为0-20 ppm；CO为0-600ppm；甲烷含量为0-300 ppm。 使这些数据集可以在线自由访问的主要目的是向传感器和人工智能研究社区提供从化学传感器获取的广泛和连续的时间序列，以开发和测试解决各种任务的策略。特别是，该数据集可用于开发用于连续监测的算法或改进感觉系统的响应时间。此外，在阵列中重复使用相同类型的传感器将允许进一步研究传感器的可变性（相同类型传感器的再现性）。其他有趣的主题可能包括传感器故障（当传感器开始故障时，系统预测降低的程度）或校准传输（一个传感器的模型是否可以扩展到其他传感器）。 有关生成数据集的更多信息，请参见Fonollosa等人的《储层计算补偿连续监测中暴露在快速变化气体浓度下的化学传感器阵列的缓慢响应》；传感器和执行器B，2015年。 Attribute Information: The data is presented in two different files: Each file contains the data from one mixture. The file ethylene_CO.txt contains the recordings from the sensors when exposed to mixtures of Ethylene and CO in air. The file ethylene_methane.txt contains the acquired time series induced by the mixture of Methane and Ethylene in air. The structure of the files is the same: Data is distributed in 19 columns. First column represents time (in seconds), second column represents Methane (or CO) concentration set point (in ppm), third column details Ethylene concentration set point (in ppm), and the following 16 columns show the recordings of the sensor array. Files include a header (one line) with the information of each column: Time (seconds), Methane conc (ppm), Ethylene conc (ppm), sensor readings (16 channels) The order of the sensors in the files is as follows: TGS2602; TGS2602; TGS2600; TGS2600; TGS2610; TGS2610; TGS2620; TGS2620; TGS2602; TGS2602; TGS2600; TGS2600; TGS2610; TGS2610; TGS2620; TGS2620 Sensors' readings can be converted to KOhms by 40.000/S_i, where S_i is the value provided in the text files. Citation Request: Citation of Fonollosa et al. 'Reservoir Computing compensates slow response of chemosensor arrays exposed to fast varying gas concentrations in continuous monitoring'; Sensors and Actuators B, 2015 is required. Creators: Jordi Fonollosa (fonollosa '@'ucsd.edu) BioCircutis Institute University of California San Diego San Diego, California, USA Donors of the Dataset: Jordi Fonollosa (fonollosa '@'ucsd.edu) Ramon Huerta (rhuerta '@' ucsd.edu)

Data Set Information: This dataset contains time series acquired from 16 chemical sensors exposed to gas mixtures at various concentration levels. Specifically, we generated two types of gas mixtures: ethylene and methane in air, and ethylene and carbon monoxide (CO) in air. Each measurement was performed by continuously collecting signals from the 16-sensor array for approximately 12 hours without interruption. This dataset was collected using a gas delivery platform setup in the chemistry laboratory of the BioCircutis Institute at the University of California San Diego. The measurement system platform offers versatility to obtain the desired concentrations of target chemicals with high accuracy and high repeatability. The sensor array consists of 16 chemical sensors from four different types (manufactured by Figaro Engineering Inc., USA): TGS-2600, TGS-2602, TGS-2610, and TGS2620 (4 units of each type). The sensors are integrated with customized signal conditioning and control electronics. The operating voltage of the sensors was maintained at 5 V throughout the entire experiment, and the working temperature was kept constant. The electrical conductivity of the sensors was continuously sampled at 100 Hz. The sensor array was placed in a 60 ml measurement chamber, where gas samples were injected at a constant flow rate of 300 ml/min. Each measurement was performed by continuously collecting signals from the 16-sensor array while the concentration levels were randomly varied. For each measurement (for each gas mixture), signals were collected continuously for approximately 12 hours without interruption. Concentration transitions were set at random times (with intervals of 80–120 seconds) and at random concentration levels. The dataset was constructed to ensure that all possible transitions are included: increasing or decreasing the concentration of one volatile compound, or setting it to zero, while the concentration of the other volatile compound remains constant (at a fixed or zero concentration level). Additional predefined pure gas mixture concentration patterns were inserted at the start, end, and approximately every 10,000 seconds. The concentration ranges of ethylene, methane, and CO were selected such that the induced response magnitudes of the sensors are similar. Additionally, lower concentration levels were preferred for the gas mixtures. Thus, the multivariate response of the sensors to the presented stimulus set is challenging, as no single configuration (presentation of a single gas or mixture) can be easily distinguished from the sensor responses. Specifically, the concentration ranges are 0–20 ppm for ethylene, 0–600 ppm for CO, and 0–300 ppm for methane. The primary purpose of making these datasets freely accessible online is to provide the sensor and artificial intelligence research communities with extensive and continuous time series acquired from chemical sensors, to develop and test strategies for solving various tasks. In particular, this dataset can be used to develop algorithms for continuous monitoring or to improve the response time of sensing systems. Furthermore, the reuse of the same sensor types in the array allows further research into sensor variability (reproducibility of sensors of the same type). Other interesting topics may include sensor fault detection (the degree to which the system predicts degradation when a sensor begins to fail) or calibration transfer (whether a model trained on one sensor can be extended to other sensors). For more information on the generation of this dataset, please refer to Fonollosa et al., "Reservoir Computing Compensates Slow Response of Chemosensor Arrays Exposed to Fast-Varying Gas Concentrations in Continuous Monitoring"; Sensors and Actuators B, 2015. Attribute Information: The data is presented in two different files: Each file contains the data from one mixture. The file ethylene_CO.txt contains the recordings from the sensors when exposed to mixtures of Ethylene and CO in air. The file ethylene_methane.txt contains the acquired time series induced by the mixture of Methane and Ethylene in air. The structure of the files is the same: Data is distributed in 19 columns. First column represents time (in seconds), second column represents Methane (or CO) concentration set point (in ppm), third column details Ethylene concentration set point (in ppm), and the following 16 columns show the recordings of the sensor array. Files include a header (one line) with the information of each column: Time (seconds), Methane conc (ppm), Ethylene conc (ppm), sensor readings (16 channels) The order of the sensors in the files is as follows: TGS2602; TGS2602; TGS2600; TGS2600; TGS2610; TGS2610; TGS2620; TGS2620; TGS2602; TGS2602; TGS2600; TGS2600; TGS2610; TGS2610; TGS2620; TGS2620 Sensors' readings can be converted to KOhms by 40.000/S_i, where S_i is the value provided in the text files. Citation Request: Citation of Fonollosa et al. 'Reservoir Computing compensates slow response of chemosensor arrays exposed to fast varying gas concentrations in continuous monitoring'; Sensors and Actuators B, 2015 is required. Creators: Jordi Fonollosa (fonollosa '@'ucsd.edu) BioCircutis Institute University of California San Diego San Diego, California, USA Donors of the Dataset: Jordi Fonollosa (fonollosa '@'ucsd.edu) Ramon Huerta (rhuerta '@' ucsd.edu)

搜集汇总

背景与挑战

背景概述

该数据集包含16个化学传感器在乙烯与甲烷、乙烯与一氧化碳两种气体混合物中的时间序列数据，采样频率为100 Hz，每次测量持续约12小时。数据集旨在支持传感器和人工智能算法的开发，特别是连续监测和传感器变异性研究。