Gas Sensor Array Research Articles

Exploiting the Self-Attention Mechanism in Gas Sensor Array (GSA) Data With Neural Networks

Gas sensor array (GSA) data is a sequential series of values that represents the temporal conditions of the existence/absence/mixture of gases and exhibits similarities to the textual stream of natural languages that represents semantic information. We speculate and subsequently prove that there also exist self-attention mechanisms in GSA data that can be exploited for gas classification and recognition. We first convert GSA data into a 1-D token series (called WORDs in this work) through sampling and quantization of the sensor values and then use an enhanced long short-term memory (LSTM) revision network, called LSTM-attention, to extract the self-attention mechanism in the GSA data. We demonstrate that LSTM-attention achieves a much better performance (99.6%) than CNN-based networks as well as other GSA data process techniques on UCI dynamic gases dataset. We also find out that the self-attention mechanism varies with different sampling and quantization levels during data acquisition.

Analyzing Gas Data Using Deep Learning and 2-D Gramian Angular Fields

The notion of employing deep learning (DL) for gas classification has kindled a revolution that has improved both data collection measures and classification performance. Yet, the current literature, with its vast contributions, has the potential in enhancing the current state of the art by employing both DL and novel visualization methods to boost classification performance and speed. Therefore, this article presents a dual classification system for high-performance gas classification: on 1-D time series data and 2-D Gramian Angular Field (GAF) data. For the GAF case study, 1-D data are converted into 2-D counterparts by means of normalization, segmentation, averaging, and color coding. The gas sensor array (GSA) dataset is used for evaluating the implemented AlexNet model for classifying 2-D GAF data and an improved version of GasNet for 1-D time-based data. Using a cloud-based architecture, the two models are evaluated and benchmarked with the state of the art. Evaluation results of the modified GasNet model on time series data signify the state-of-the-art accuracy of 96.5%, while AlexNet achieved 81.0% test accuracy of GAF classification with near real-time performance on edge computing platforms.

A CNN-Based E-Nose Using Time Series Features for Food Freshness Classification

In this article, a convolutional neural network (CNN)-based electronic nose system is proposed, which utilizes time series features to classify the freshness of food. The electronic nose is composed of three elements: a highly sensitive miniaturized micro-electromechanical systems (MEMS) metal–oxide (MOX)–semiconductor gas sensor array, a complementary metal–oxide–semiconductor (CMOS) integrated circuit, and a CNN-based classification unit. The gas sensors exhibit exceptional sensitivity, low power consumption, and a small size. The interface circuit for sensor readout is fabricated using a standard 180-nm CMOS process, with an area of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.5\times1.5$ </tex-math></inline-formula> mm. The interface chip mainly consists of an analog-to-digital converter (ADC), an electrically erasable programmable read-only memory (EEPROM), an oscillator (OSC), power management, a heater driver, and a digital control circuit. Fixed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R}{(}{t}{)}$ </tex-math></inline-formula> exposures are taken at specific intervals under gas input and output conditions. Time series features are extracted from a diverse set of sensor signals, which allows the subtle differences in various food odors at different freshness levels to be identified. The incorporation of time series feature extraction extends the data features, resulting in enhanced classification accuracy with a limited number of sensors. An abstract odor map input to the CNN is formed by combining the steady-state and transient information of the gas sensor response. The final freshness classification accuracy of 20 types of foods is 97.3%. The accuracy is improved by 6.5% after the implementation of time series feature extraction, demonstrating the significance of instantaneous fluctuation information for classification accuracy.

Textural landscapes of VOC-sensitive chiral liquid crystal-based materials

Liquid crystal-based materials, in which liquid crystal molecules are confined and ordered in compartments, are dynamic materials yielding a variety of optical textures that can be tuned as a response to physical and chemical stimuli. While nematic and smectic-based gel materials have been reported as dynamic optical sensors to report volatile organic compounds (VOCs), chiral systems are less explored despite having the potential to yield extremely rich optical landscapes. Here, we report for the first time the confinement of chiral liquid crystal formulations by an interface formed by ionic liquid molecules. The resultant self-assembled ionic liquid/liquid crystal droplets are simultaneously immobilized on a gelatin matrix. The droplets feature a rich variety of unique topological states. We explored, by means of polarizing optical microscopy, the various droplet optical textures and categorized them with regard to their relative chirality parameter. We further investigated their optical response in the presence of gas analytes and discussed their potential utilization as dynamic liquid crystal-based optical VOC sensors. The newly generated soft materials with semi-selective VOC sensing capabilities can be further utilized in arrays of liquid crystal-based gas sensors for the analysis of complex gas samples using artificial olfaction approaches.

Multi-feature Optimization of a Gas Sensor Array to Improve Component and Concentration Detection Performance of VOC Mixtures

Many studies focus on feature extraction and selection of gas sensor arrays for gas identification. In this work, we intended to find a feature subset obtained by selecting the most important features for simultaneously improving component and concentration detection performance of a gas sensor array to three harmful VOCs (toluene, methanol, and ethanol) and their mixtures. First, 30 features were extracted from 6 sensors’ responses to form a multi-feature set. Then, two feature selection methods based on Wilks’ Λ-statistic and random forest were employed to obtain the best feature combination. Seven out of 30 features were finally selected to form the optimal feature set. The gas identification accuracy is 94.3%, and the concentration estimation error is 0.79 ppm (RMSE). Through feature selection, not only qualitative and quantitative analyses performance of VOCs mixtures are significantly improved, but also system complexity (6 to 4 sensors) and computation cost (by about 15%) are effectively reduced.

Inkjet-printed ZnO-based MEMS sensor array combined with feature selection algorithm for VOCs gas analysis

An electronic nose system can enhance the selectivity of semiconductor gas-sensor systems but increase system power consumption and signal data processing complexity. Herein, we developed a gas sensor array composed of six ZnO-based micro-electro-mechanical systems (MEMS) sensors prepared by inkjet printing sensing materials on a micro-hotplate. Its power consumption decreased to 36 mW. Furthermore, the particle swarm optimization (PSO) algorithm was used to select the optimal features, decreasing the total of 72 features to 5. Finally, support vector machine and artificial neural network models based on the optimal 5 features were performed to recognize and quantify volatile organic compounds (VOCs), including formaldehyde, ethanol, toluene, and xylene. The identification accuracy and R2 reach 97.9 % and 0.975, respectively. The result demonstrates that the MEMS sensor array, combined with PSO and pattern recognition algorithms, is a promising approach to accurately identifying and quantifying VOCs.

Arrays of Functionalized Graphene Chemiresistors for Selective Sensing of Volatile Organic Compounds

Real-time detection of volatile organic compounds (VOCs) is being spotlighted for its importance in environmental monitoring and disease detection from the breath. However, the selective detection of VOCs in a small form factor gas sensor remains a challenge. Here, we present the VOC sensing performance of phthalocyanine (Pc)-functionalized graphene chemiresistors supported by the theoretical prediction of Pc’s VOC interactions. Pc functionalization, with and without metalation, was studied parametrically with varying deposition parameters to explore impacts on the gas sensing response toward VOCs as a function of substrate coverage. Sensing data were collected to establish fingerprint signatures for a variety of VOCs. Improvement in selectivity is shown by successfully differentiating the VOCs from each other when combined into a gas sensor array tested with acetone, ethanol, formaldehyde, and toluene.

E-Nose System Based on Fourier Series for Gases Identification and Concentration Estimation From Food Spoilage

This work presents an electronic nose (EN)-based gas identification and concentration estimation method for the detection of food spoilage. The response data of sensors were acquired through a commercial gas sensor array and data acquisition circuit board and transformed into pictures with the form of the Fourier series. A convolutional neural network (CNN) model was used to identify the pictures from the conversion of sensor data, thus achieving the purpose of identifying the gases (C2H5OH, NH3, and H2S). In order to solve the problem of sample imbalance and to improve the generalization performance of classification models, the synthetic minority oversampling technique (SMOTE) and dropout technique were employed. Fivefold cross-validation was used to evaluate the performance of the model, of which the gas identification accuracy rate reached 96.67%. Moreover, a gas concentration regression model with the advantages of simplicity and strong interpretability was further proposed to estimate the concentrations of C2H5OH, NH3, and H2S. The mean absolute errors and coefficients of determination for the concentration estimation of C2H5OH, NH3, and H2S are (3.71 ppm, 0.968), (0.50 ppm, 0.968), and (0.13 ppm, 0.99), respectively. Furthermore, we used our model to evaluate the freshness of kiwifruit, pork, and beef, and it showed satisfactory predictive performance. The method proposed in this work realizes high-precision detection of gases from food spoilage and has a good application prospect in the rapid judgment of food freshness on the EN system.

A Hybrid Machine Learning Algorithm for Detection of Simulated Expiratory Markers of Diabetic Patients Based on Gas Sensor Array

The method for breath detection using gas sensor array is gaining popularity. It was found that acetone can be used as breath marker in diabetic patients. In this article, seven metal oxide gas sensors were used to collect acetone and ethanol gas, which are used to simulate the exhaled breath of diabetic patients to obtain multidimensional response data. The kernel principal component analysis (KPCA) algorithm is used to extract the characteristics from the data collected by the sensor array. The kind of gases is qualitatively identified using the adaptive boosting (AdaBoost) algorithm, and the grid search (GS) method is used to automatically optimize the parameters of AdaBoost algorithm. The quantitative analysis of gas concentration is performed using multivariate relevance vector machine (MVRVM) and it is also trained using the gas sensor array drift dataset at different concentrations from the University of California (UCI) database. The experimental results show that the accuracy of the algorithm in the qualitative identification of acetone and ethanol gas reaches 99.722%, and the root-mean-square errors (RMSE) for quantification of acetone and ethanol gases are 0.027 and 0.030, respectively. The algorithm is used for qualitative identification on the gas sensor array drift dataset at different concentrations with an accuracy of 94.55%, and the RMSE for quantification of acetone and ethanol gases are 11.59 and 8.72, respectively.

Fault Diagnosis of SF6-Insulated Equipment by Micro Gas Sensor Array

SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition products could reflect the running status and inner faults of power equipment, and it's expected to realize a timely warning. In this work, six faults including spark and corona discharge were simulated, and SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> decomposition products with various types and contents were obtained as well. Different from previous investigations employing precision instruments, such as gas chromatography and infrared spectroscopy, a micro sensor array loaded with three gas-sensitive nanomaterials was used to discriminate fault characteristic gases, performing obvious advantages in small size, high integration, and rapid detection. Gas chromatography-mass spectrometry (GCMS) indicated that seven analytes had significant differences in types and contents. Meanwhile, the as-prepared micro gas sensor array also outputted significantly various signals for seven analytes, which provided a basis for gas identification. With the assistance of stacked denoising autoencoder (SDAE)-based discrimination algorithms, the recognition model between the response signals of the array and the discharge faults in power equipment could be established. In comparison with KNN (66.67 %), decision tree (70.47 %), and BPNN (73.33 %), SVM has achieved the highest average accuracy of 75.23 %. Totally, this work provides a promising novel method for rapid on-site inspection of SF <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">6</sub> -insulated power equipment.

Recognition Algorithm for Detection of Precursor Chemicals by Semiconductor Gas Sensor Array Under Dynamic Measurement

Narcotic and psychotropic drugs seriously endanger human health and the stable development of society. In order to snipe drugs from the source, it is extremely significant to detect precursor chemicals. In this article, the precursor chemicals and interference samples are measured by the semiconductor gas sensor array under dynamic measurement and static measurement. The principal component analysis (PCA), support vector machine (SVM), and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${K}$ </tex-math></inline-formula> -nearest neighbor (KNN) algorithms are used for qualitative and quantitative analyses of the acquired data. In addition, the qualitative result of two measurement methods using the SVM and KNN algorithms are given. For quantitative analysis, support vector regression (SVR) is used, and the sum of the mean absolute error (MAE) in the dynamic measurement is better than in static measurement. Besides, the stability experiments are conducted about a month later. It is found that the qualitative recognition in dynamic measurement is better than static measurement under interference and no interference samples. It reveals that the dynamic measurement of the sensor array has certain advantages under qualitative analysis and is more suitable for the real-time detection of precursor chemicals.

A Novel Gas Recognition Algorithm for Gas Sensor Array Combining Savitzky–Golay Smooth and Image Conversion Route

In recent years, the application of Deep Neural Networks to gas recognition has been developing. The classification performance of the Deep Neural Network depends on the efficient representation of the input data samples. Therefore, a variety of filtering methods are firstly adopted to smooth filter the gas sensing response data, which can remove redundant information and greatly improve the performance of the classifier. Additionally, the optimization experiment of the Savitzky–Golay filtering algorithm is carried out. After that, we used the Gramian Angular Summation Field (GASF) method to encode the gas sensing response data into two-dimensional sensing images. In addition, data augmentation technology is used to reduce the impact of small sample numbers on the classifier and improve the robustness and generalization ability of the model. Then, combined with fine-tuning of the GoogLeNet neural network, which owns the ability to automatically learn the characteristics of deep samples, the classification of four gases has finally been realized: methane, ethanol, ethylene, and carbon monoxide. Through setting a variety of different comparison experiments, it is known that the Savitzky–Golay smooth filtering pretreatment method effectively improves the recognition accuracy of the classifier, and the gas recognition network adopted is superior to the fine-tuned ResNet50, Alex-Net, and ResNet34 networks in both accuracy and sample processing times. Finally, the highest recognition accuracy of the classification results of our proposed route is 99.9%, which is better than other similar work.

A Pilot Study for Legionella pneumophila Volatilome Characterization Using a Gas Sensor Array and GC/MS Techniques.

Legionellosis is a generic term describing the pneumonic (Legionnaires' disease, LD) and non-pneumonic (Pontiac fever, PF) forms of infection with bacteria belonging to the genus Legionella. Currently, the techniques used to detect Legionella spp. in water samples have certain limitations and drawbacks, and thus, there is a need to identify new tools to carry out low-cost and rapid analysis. In this regard, several studies demonstrated that a volatolomics approach rapidly detects and discriminates different species of microorganisms via their volatile signature. In this paper, the volatile organic compounds (VOCs) pattern emitted in vitro by Legionella pneumophila cultures is characterized and compared to those produced by other Legionella species and by Pseudomonas aeruginosa, using a gas sensor array and gas chromatograph mass spectrometer (GC-MS). Bacterial cultures were measured at the 3rd and 7th day after the incubation. Sensor array data analyzed via the K-nearest neighbours (k-NN) algorithm showed a sensitivity to Legionella pneumophila identification at around 89%. On the other hand, GC-MS identified a bouquet of VOCs, mainly alcohols and ketones, that enable the differentiation of Legionella pneumophila in respect to other waterborne microorganisms.

Metal oxide semiconductor gas sensing materials for early lung cancer diagnosis

The urgency of early lung cancer (LC) diagnosis and treatment has been more and more significant. Exhaled breath analysis using gas sensors is a promising way to find out if someone has LC due to its low-cost, non-invasive, and real-time monitoring compared with traditional invasive diagnostic techniques. Among sensor-based gas detection techniques, metal oxide semiconductor’s gas sensors are one of the most important types. This review presents the-state-of-art in metal oxide gas sensors for the diagnosis of early LC. First, the exhaled breath biomarkers are described with emphasis on the concentration of abnormal volatile organic compounds (VOCs) caused by the metabolic process of LC cells. Then, the research status of metal oxide gas sensors in LC diagnosis is summarized. The sensing performance and enhancement strategy of biomarkers provided by metal oxide semiconductor materials are reviewed. Another effective way to improve VOC detection performance is to build a gas sensor array. At the same time, various gas sensors combined with self-powered techniques are mentioned to display a broad development prospect in breath diagnosis. Finally, metal oxide gas sensor-based LC diagnosis is prospected.

E-nose based on a high-integrated and low-power metal oxide gas sensor array

To improve the poor selectivity of a semiconductor gas-sensor system, an array of sensors can be utilized. However, this increases the system’s size and power consumption. To overcome these limitations, we propose a high-integrated and highly selective electronic nose (E-nose) comprising two independent gas-sensing elements on a low-power microhotplate (MHP). Pd-SnO2 nanoflowers and Pd-WO3 microparticles were prepared and printed on a bridge-structured MHP 20 µm apart and over an area of 110 µm × 45 µm. This was achieved using electrohydrodynamic inkjet printing aided by in-situ infrared laser curing to form an array of sensors. A power of 17 mW was required to increase the temperature of the MHP to 300 °C, which is the optimal operating temperature of the two gas sensors. Thus, a high response and low cross-sensitivity were achieved for hydrogen, ammonia, hydrogen–ammonia, ethanol, acetone, ethanol–acetone, toluene, and formaldehyde. A wavelet transform was used to reduce the noise and dimensionality of the signals from the gas-sensor array. Qualitative identification of the eight gases with an accuracy of 99.86% was achieved using the k-nearest neighbor (kNN) model. A p neighbors back propagation neural network (pN-BPNN) model was established to remove interfering samples to quantitatively estimate the gas concentration. The quantitative identification accuracy of pN-BPNN was higher than that of the standard back propagation neural network (BPNN) model with the average absolute percentage error of hydrogen detection in the range of 15–500 ppm, decreasing from 5.44% to 2.08%.

Application of Machine Learning Methods for an Analysis of E-Nose Multidimensional Signals in Wastewater Treatment

The work represents a successful attempt to combine a gas sensors array with instrumentation (hardware), and machine learning methods as the basis for creating numerical codes (software), together constituting an electronic nose, to correct the classification of the various stages of the wastewater treatment process. To evaluate the multidimensional measurement derived from the gas sensors array, dimensionality reduction was performed using the t-SNE method, which (unlike the commonly used PCA method) preserves the local structure of the data by minimizing the Kullback-Leibler divergence between the two distributions with respect to the location of points on the map. The k-median method was used to evaluate the discretization potential of the collected multidimensional data. It showed that observations from different stages of the wastewater treatment process have varying chemical fingerprints. In the final stage of data analysis, a supervised machine learning method, in the form of a random forest, was used to classify observations based on the measurements from the sensors array. The quality of the resulting model was assessed based on several measures commonly used in classification tasks. All the measures used confirmed that the classification model perfectly assigned classes to the observations from the test set, which also confirmed the absence of model overfitting.

Spatial Upscaling-Based Algorithm for Detection and Estimation of Hazardous Gases

Recently, society/industry is getting smarter and sustainable through artificial intelligence-based solutions. However, this rapid advancement is also polluting our air ambience. Hence real-time detection and estimation of hazardous gases/odors in the air ambiance has become a critical need. In this paper, a convolutional neural network (CNN) based multi-element gas sensor arrays signature response analysis approach has been presented to achieve higher accuracy in detection and estimation of hazardous gases. Accordingly, the real-time gas sensor array responses are spatially upscaled and processed on the edge, using lightweight CNNs. For the verification of our hypothesis, we have used a four-element metal-oxide semi-conductor (MOS)-based thick-film gas sensor array, fabricated by our group, by using SnO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> , ZnO, MoO, CdS materials for detection and estimation of four target hazardous gases, viz., acetone, carbon-tetrachloride, ethyl-methyl-ketone, and xylene. The four-element (2×2) raw sensor responses are first upscaled to 6×6 responses and a lightweight CNN is trained on 42 samples of 6×6 input vectors. The trained system is then tested using 16 unknown (not used during training) test samples of the considered gases/odors. All the 16 test samples are detected correctly. The Mean Squared Error (MSEs) of detection has been 1.42×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-14</sup> while the estimation accuracy of 2.43×10 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-3</sup> were achieved for the considered gases. Our designed system is generic in design and can be extended to other gases/odors of interest.

Feature Ensemble Learning for Sensor Array Data Classification Under Low-Concentration Gas

Gas sensor array (GSA) data usually has high-dimensional features and small sample size. When a classifier is directly used for GSA data classification, it is prone to overfitting and has high time cost. The traditional solution is to perform feature dimensionality reduction before classification. However, selecting a suitable dimensionality reduction method is time-consuming and laborious, and some features useful for classification may be lost after dimensionality reduction, especially for the weak sensor response data to low concentration gas. In this paper, we proposed a feature ensemble based extreme learning machine framework (FE-ELM) for GSA data classification. In FE-ELM, downsampling is first performed on the time series of each sensor, and then the downsampled features of different sensors are combined to obtain fused feature subsets. Next, a base ELM is trained independently on each fused feature subset with all training samples by solving the least squares problem. The final FE-ELM predictions for input samples are obtained by voting the prediction results of all base ELMs. Compared with traditional methods, the proposed method solves the overfitting problem and can be directly used for GSA data classification without prior feature dimension reduction. Furthermore, the ensemble of all base classifiers with little loss of original features enables the proposed FE-ELM to have more efficient and robust classification performance. Experimental results on data from both homemade GSA under low concentration gases (ppb) and publicly available confirm that the proposed FE-ELM exceeds traditional methods and extends the detection limit of the sensor array.

Quantitative prediction of ternary mixed gases based on an SnO2 sensor array and an SSA-BP neural network model.

This paper describes a tin oxide and copper doped tin oxide gas sensing material synthesized by a biological template method and simple hydrothermal reaction, which were used for the preparation of a gas sensor array. The sensor array is combined with the Sparrow Search Algorithm optimized BP neural network algorithm (SSA-BP) to predict and analyze the concentration of indoor toxic gases, including ammonia, xylene, and formaldehyde. Granular SnO2 was prepared by the biological template method and Cu/SnO2 doped with different copper ion concentrations was prepared by the hydrothermal method. The morphology of the synthesized nanomaterials was characterized by SEM, and the elemental composition and chemical state of the main elements were analyzed by XRD and XPS. The PL emission observed in the visible region is attributed to the defect level gap caused by oxygen. The optimal operating temperature, sensitivity, response/recovery time and the long-term stability of the sensor array have been studied. By combining the sensor array with the neural network algorithm in a simulated indoor environment at four humidity levels, the concentration information of the gas mixtures could be well predicted and the predicted concentration error was less than 0.84 ppm. Therefore, the sensor array prepared in this study combined with the SSA-BP algorithm achieved good results in predicting the concentrations of the three toxic mixtures.

Flow dependence of handheld breath analyzer for body fuel utilization monitoring

Home healthcare medical technologies have been gaining popularity and are more affordable in recent years. Exhaled breath analysis has potential in this field. The development of gas sensor technology has enabled us to build a small affable breath analysis device with the electronic nose concept. In this work, a handheld breath analyzer was developed for monitoring body fuel utilization. A hybrid gas sensor array, including electrochemical and photoacoustic gas sensors, was used to accurately measure oxygen and carbon dioxide in exhaled breath. The bypass configuration volume flow measurement method was developed to fit a small portable device. The experiment shows that both oxygen and carbon dioxide sensors are flow-dependent due to the slow response time of each sensor type. The response of the photoacoustic sensor is relatively slower than those of other sensor types. Thus, a mathematical model was developed to correct the individual sensor value to get a more accurate value of body fuel utilization. The comparison protocol of known concentrations of the oxygen and carbon gases with various flow conditions was conducted, and the mathematical model for reconstructing the original gas concentration was proposed. The result shows that the device is able to detect the RER change of humans after having a high carbohydrate content meal and after exercise