Abstract:In this work, a virtual sensor for PM10 concentration monitoring is presented. The sensor is based on wavenet models and uses daily mean NO2 concentration and meteorological variables (wind speed and rainfall) as input. The methodology has been applied to the reconstruction of PM10 levels measured from 14 monitoring stations in Lombardy region (Italy). This region, usually affected by high levels of PM10, is a challenging benchmarking area for the implemented sensors. Neverthless, the performances are good wit… Show more
“…for Equations ( 4) and (8). The models were constructed by varying both the autoregressive and exogenous components, denoted as n a and n b , spanning orders from 1 to 4, and the best model for each sensor was selected based on the MAE.…”
Section: S5 S6unclassified
“…These data are then utilized to establish a mathematical approximation of the relationship between the measured variables and the sensors’ output [ 2 , 3 , 4 ]. Machine learning is used in data science to facilitate the identification of patterns and automate the process of data analysis, offering a compelling approach to tackling virtual sensing challenges by leveraging historical data to predict and estimate unmeasured variables due to its capacity to discern complex patterns and relationships within data [ 5 , 6 , 7 , 8 ]. Through various algorithms like neural networks, support vector machines, and ensemble methods, machine learning effectively reconstructs and forecasts missing or inaccessible data points [ 9 ].…”
In recent years, the application of machine learning for virtual sensing has revolutionized the monitoring and management of information. In particular, electrochemical sensors generate large amounts of data, allowing the application of complex machine learning/AI models able to (1) reproduce the measured data and (2) predict and manage faults in the measuring sensor. In this work, data-driven models based on an autoregressive model and an artificial neural network have been identified and used to (i) evaluate sensor redundancy and (ii) predict and manage faults in the context of electrochemical sensors for the measurement of ethanol. The approach shows encouraging results in terms of both performance and sensitivity analyses, allowing for the reconstruction of the values measured by two sensors in a series of six sensors with different dopant levels and to reproduce their values after a fault.
“…for Equations ( 4) and (8). The models were constructed by varying both the autoregressive and exogenous components, denoted as n a and n b , spanning orders from 1 to 4, and the best model for each sensor was selected based on the MAE.…”
Section: S5 S6unclassified
“…These data are then utilized to establish a mathematical approximation of the relationship between the measured variables and the sensors’ output [ 2 , 3 , 4 ]. Machine learning is used in data science to facilitate the identification of patterns and automate the process of data analysis, offering a compelling approach to tackling virtual sensing challenges by leveraging historical data to predict and estimate unmeasured variables due to its capacity to discern complex patterns and relationships within data [ 5 , 6 , 7 , 8 ]. Through various algorithms like neural networks, support vector machines, and ensemble methods, machine learning effectively reconstructs and forecasts missing or inaccessible data points [ 9 ].…”
In recent years, the application of machine learning for virtual sensing has revolutionized the monitoring and management of information. In particular, electrochemical sensors generate large amounts of data, allowing the application of complex machine learning/AI models able to (1) reproduce the measured data and (2) predict and manage faults in the measuring sensor. In this work, data-driven models based on an autoregressive model and an artificial neural network have been identified and used to (i) evaluate sensor redundancy and (ii) predict and manage faults in the context of electrochemical sensors for the measurement of ethanol. The approach shows encouraging results in terms of both performance and sensitivity analyses, allowing for the reconstruction of the values measured by two sensors in a series of six sensors with different dopant levels and to reproduce their values after a fault.
“…Together with other sensors, they can be components of portable or stationary multisensor platforms for measuring PM and CO, NO 2 , O 3 , SO 2 , CO 2 gases, and volatile hydrocarbons (e.g., benzene) [ 42 ]. It is also possible to measure PM by indirect methods using artificial neural networks based on measurements of NO 2 , temperature, humidity, and wind speed [ 43 ]. Information on PM can also be obtained through satellite measurements [ 44 ] or by using a fusion of data from LCSs placed on cars and stationary stations.…”
Particulate matter (PM) suspended in the air significantly impacts human health. Those of anthropogenic origin are particularly hazardous. Poland is one of the countries where the air quality during the heating season is the worst in Europe. Air quality in small towns and villages far from state monitoring stations is often much worse than in larger cities where they are located. Their residents inhale the air containing smoke produced mainly by coal-fired stoves. In the frame of this project, an air quality monitoring network was built. It comprises low-cost PMS7003 PM sensors and ESP8266 microcontrollers with integrated Wi-Fi communication modules. This article presents research results on the influence of the PM sensor location on their indications. It has been shown that the indications from sensors several dozen meters away from each other can differ by up to tenfold, depending on weather conditions and the source of smoke. Therefore, measurements performed by a network of sensors, even of worse quality, are much more representative than those conducted in one spot. The results also indicated the method of detecting a sudden increase in air pollutants. In the case of smokiness, the difference between the mean and median indications of the PM sensor increases even up to 400 µg/m3 over a 5 min time window. Information from this comparison suggests a sudden deterioration in air quality and can allow for quick intervention to protect people’s health. This method can be used in protection systems where fast detection of anomalies is necessary.
“…People with low resistance, such as the elderly and children, are particularly vulnerable to the effects of PM 2.5 [9,10]. Therefore, effective monitoring and control of PM 2.5 concentration has a positive role in improving human health [11][12][13]. Moreover, the picture-based PM 2.5 concentration estimation methods can also be applied to other visual tasks, such as PM 2.5 concentration, which can be used to assess the credibility of person re-identification results in hazy weather [14][15][16].…”
PM2.5 in the atmosphere causes severe air pollution and dramatically affects the normal production and lives of residents. The real-time monitoring of PM2.5 concentrations has important practical significance for the construction of ecological civilization. The mainstream PM2.5 concentration prediction algorithms based on electrochemical sensors have some disadvantages, such as high economic cost, high labor cost, time delay, and more. To this end, we propose a simple and effective PM2.5 concentration prediction algorithm based on image perception. Specifically, the proposed method develops a natural scene statistical prior to estimating the saturation loss caused by the ’haze’ formed by PM2.5. After extracting the prior features, this paper uses the feedforward neural network to achieve the mapping function from the proposed prior features to the PM2.5 concentration values. Experiments constructed on the public Air Quality Image Dataset (AQID) show the superiority of our proposed PM2.5 concentration measurement method compared to state-of-the-art related PM2.5 concentration monitoring methods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.