A dual drift compensation framework based on subspace learning and cross-domain adaptive extreme learning machine for gas sensors

Se, Haifeng; Song, Kai; Liu, Hui; Zhang, Weiyan; Wang, Xuanhe; Liu, Jijiang

doi:10.1016/j.knosys.2022.110024

Cited by 18 publications

(6 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As such, four combinations could be tested this way, and either the calibration with the best result or an overall average could be used. Additionally, by collecting more samples regarding degradation, drift compensation algorithms could be used to extend the sensors’ life cycle past single use [ 44 , 45 ].…”

Section: Discussionmentioning

confidence: 99%

Graphene Oxide, Carbon Nanotubes, and Polyelectrolytes-Based Impedanciometric E-Tongue for Estrogen Detection in Complex Matrices

Reis,

Fino,

Raposo

2024

Sensors

View full text Add to dashboard Cite

Currently, it is necessary to maintain the quality of aquifers and water bodies, which means the need for sensors that detect molecules as emerging pollutants (EPs) at low concentrations in aqueous complex solutions. In this work, an electronic tongue (e-tongue) prototype was developed to detect 17β-estradiol in tap water. To achieve such a prototype, an array of sensors was prepared. Each sensor consists of a solid support with interdigitated electrodes without or with thin films prepared with graphene oxide, nanotubes, and other polyelectrolytes molecules adsorbed on them. To collect data from each sensor, impedance spectroscopy was used to analyze the electrical characteristics of samples of estrogen solutions with different concentrations. To analyze the collected data from the sensors, principal components analysis (PCA) method was used to create a three-dimensional plane using the calculated principal components, namely PC1 and PC2, and the estrogen concentration values. Then, damped least squares (DLS) was used to find the optimal values for the hyperplane calibration, as the sensitivity of this e-tongue was not represented by a straight line but by a surface. For the collected data, from nanotubes and graphene oxide sensors, a calibration curve for concentration given by the 10PC1×0.492−PC2×0.14–14.5 surface was achieved. This e-tongue presented a detection limit of 10−16 M of 17β-estradiol in tap water.

show abstract

Section: Discussionmentioning

confidence: 99%

Graphene Oxide, Carbon Nanotubes, and Polyelectrolytes-Based Impedanciometric E-Tongue for Estrogen Detection in Complex Matrices

Reis,

Fino,

Raposo

2024

Sensors

View full text Add to dashboard Cite

show abstract

“…As in several technological domains, the recent explosion in the development, use, and perturbation of machine-learning approaches, such techniques can immensely impact the area of gas sensors. Several papers where machine-learning approaches are used to aid gas sensors have been recently published [148,[294][295][296][297].…”

Section: Discussionmentioning

confidence: 99%

Wearable Nano-Based Gas Sensors for Environmental Monitoring and Encountered Challenges in Optimization

Hooshmand,

Kassanos,

Keshavarz

et al. 2023

Sensors

View full text Add to dashboard Cite

With a rising emphasis on public safety and quality of life, there is an urgent need to ensure optimal air quality, both indoors and outdoors. Detecting toxic gaseous compounds plays a pivotal role in shaping our sustainable future. This review aims to elucidate the advancements in smart wearable (nano)sensors for monitoring harmful gaseous pollutants, such as ammonia (NH3), nitric oxide (NO), nitrous oxide (N2O), nitrogen dioxide (NO2), carbon monoxide (CO), carbon dioxide (CO2), hydrogen sulfide (H2S), sulfur dioxide (SO2), ozone (O3), hydrocarbons (CxHy), and hydrogen fluoride (HF). Differentiating this review from its predecessors, we shed light on the challenges faced in enhancing sensor performance and offer a deep dive into the evolution of sensing materials, wearable substrates, electrodes, and types of sensors. Noteworthy materials for robust detection systems encompass 2D nanostructures, carbon nanomaterials, conducting polymers, nanohybrids, and metal oxide semiconductors. A dedicated section dissects the significance of circuit integration, miniaturization, real-time sensing, repeatability, reusability, power efficiency, gas-sensitive material deposition, selectivity, sensitivity, stability, and response/recovery time, pinpointing gaps in the current knowledge and offering avenues for further research. To conclude, we provide insights and suggestions for the prospective trajectory of smart wearable nanosensors in addressing the extant challenges.

show abstract

“…However, when the temperature or humidity changes (i.e., when identifying the wild MT species in the forest), the e-nose response will drift, which may deteriorate the detection performance. Therefore, in future research, machine learning methods can be considered to design drift compensation methods to promote the intelligent process of e-noses [74,75]. By training algorithms with large datasets of smellprints and corresponding wild MT species, it may be possible to develop automated, real-time identification systems that can identify wild MT species accurately, quickly, and at low costs.…”

Section: Fast Cheap and Reliable Methods To Distinguish Wild Mt Speci...mentioning

confidence: 99%

First Steps in Developing a Fast, Cheap, and Reliable Method to Distinguish Wild Mushroom and Truffle Species

Ferreira,

Dias,

Melo

et al. 2023

Resources

View full text Add to dashboard Cite

Wild mushrooms and truffles (MT) are important resources, which can contribute to the socioeconomic sustainability of forestry ecosystems. However, not all wild MT are edible. Fast, cheap, and reliable methods that distinguish wild MT species (including the deadly ones) can contribute to valuing these important forest resources. Here, we tested if wild MT species, and their edibility, could be distinguished based on their aroma profiles (i.e., smellprints). For that, we combined the use of the electronic nose with classification models (linear discriminant analysis (LDA) and partial least squares discriminant analysis (PLS-DA)) to distinguish between 14 wild MT species (including edible and non-edible species) collected in Portugal. The 14 wild MT species could be accurately distinguished using LDA (93% accuracy), while the edible and non-edible species could be accurately distinguished using both LDA and PLS-DA (97% and 99% accuracy, respectively). Keeping in mind that our methodological design’s feasibility was verified using a small sample, the data show the potential of the combined use of the electronic nose with discriminant analysis to distinguish wild MT species and their edibility based on their aromatic profile. Although a larger dataset will be necessary to develop a quick and reliable identification method, it shows potential to be as accurate as the identification performed by mycologists and molecular biology, yet requiring less technical training, and the analyses are cheaper and faster.

show abstract

A dual drift compensation framework based on subspace learning and cross-domain adaptive extreme learning machine for gas sensors

Cited by 18 publications

References 38 publications

Graphene Oxide, Carbon Nanotubes, and Polyelectrolytes-Based Impedanciometric E-Tongue for Estrogen Detection in Complex Matrices

Graphene Oxide, Carbon Nanotubes, and Polyelectrolytes-Based Impedanciometric E-Tongue for Estrogen Detection in Complex Matrices

Wearable Nano-Based Gas Sensors for Environmental Monitoring and Encountered Challenges in Optimization

First Steps in Developing a Fast, Cheap, and Reliable Method to Distinguish Wild Mushroom and Truffle Species

Contact Info

Product

Resources

About