Elsevier

Building and Environment

Volume 104, 1 August 2016, Pages 68-75
Building and Environment

Selection of the most influential flow and thermal parameters for predicting the efficiency of activated carbon filters using neuro-fuzzy technique

https://doi.org/10.1016/j.buildenv.2016.04.031Get rights and content

Highlights

  • Influence of flow-thermal parameters on the performance of the activated carbon filter.

  • Filter-ventilation system was analyzed.

  • Adaptive neuro fuzzy inference system was applied to the data.

  • To select the most influential parameters for the adsorption process in ventilation filtration system.

  • The most influential parameter for prediction of outlet concentrations of acetone is temperature.

Abstract

This paper analyzes the influence of various flow and thermal variables on the performance of the activated carbon filter in the air-conditioning system. The ANFIS (adaptive neuro fuzzy inference system) method was applied to the data obtained from the experimental apparatus in order to select the most influential parameters for assessing the efficiency of the activated carbon filter. Acetone was selected as the target pollutant component. Experiments were performed for different temperature, humidity, and flow rate conditions, as well as for acetone concentrations. A set of four potential inputs was considered: velocities of gas mixture through ventilation duct (flow), temperature, humidity, and concentration of test chemical pollutants ahead of the filter module. The results show that the most influential parameter for predicting outlet concentrations of acetone is temperature.

Introduction

Indoor air purification is one of the main strategies to improve indoor air quality (IAQ) [1]. In the air filtration and purification applications, activated carbon or activated charcoal is widely used due to many potential advantages [2]. The applications of activated carbon can be found in a large number of industries, water filters, respirators, solvent recovery systems, etc. [3], [4]. Activated carbon can be used for filtering most organic pollutants and solvents from air through the adsorption process [5].

The aim of this study was to analyze how different flow and thermal parameters of the air conditioning system (temperature, humidity, and air velocity) influence the efficiency of the activated carbon filter for different concentrations of the test pollutant. We used acetone, a volatile organic compound (VOC) frequently present in occupational and residential environments, as a test gas pollutant. Depending on its concentration in air, it can cause different symptoms, from irritations to headaches and dizziness [6], [7], [8], so effective elimination of this compound from air is mandatory.

Numerous studies regarding VOC adsorption were reported in open literature [9], [10], [11], [12], [13], [14], [15]. Benzene, toluene, MEK and acetone were used as prevailing challenge gases. Most of these studies were focused on activated carbon filters with concentrations ranging from a couple of ppm (residential indoor environments) to a few hundred ppm (industrial environments). Additionally, numerous studies were performed for theoretical modeling of fixed-bed adsorption filters. A comprehensive literature review of existing sorbent-based gas filter models used for predicting filter performance for nonindustrial buildings is provided in Ref. [16]. All of these studies were conducted for dry air, despite the fact that parameters of the two most important models for breakthrough curves, Wheeler–Jonas and Yoon–Nelson, are a function of humidity level. This inadequacy was reported in Ref. [17], where a framework was developed for predicting the breakthrough curve of activated carbon filters at low concentrations and different levels of relative humidity, applying accelerated test data. However, the influence of different temperature levels as well as of the simultaneous effect of temperature and humidity on the breakthrough and efficiency of filters has not been analyzed.

The aim of this study was to analyze the magnitude of the effect of different flow and thermal parameters (velocity of gas mixture through the ventilation duct (flow), temperature, humidity, and upstream concentration of challenge gas (acetone)) on the efficiency of active carbon filters. We used a statistical learning approach, based on artificial neural networks, to develop suitable models, which were subsequently used to assess the importance of different flow and thermal parameters. We also used artificial neural networks (ANN) as a substitute for the analytical approach, as ANN offers advantages such as no required previous knowledge of internal system parameters [18]. In this study, we applied the adaptive neuro-fuzzy inference system (ANFIS), which is a specific type of the ANN family, to select the most influential parameters for predicting downstream concentration of acetone in the airstream [19].

Section snippets

Experimental setup

Elements of a filter-ventilation system used in experiments under the different thermal and flow conditions are shown in Fig. 1. Supervisory control and data acquisition (SCADA) system was developed for examining the characteristics of the filter elements. The measurement system consists of sensors and transmitters of physical and nonelectrical quantities, a personal computer, and a source of direct current as a transmitter of power supply. The positions of measuring points on the examination

Evaluating accuracy indices

Predictive performances of the proposed model were presented by root means square error (RMSE), coefficient of determination (R2), and Pearson coefficient (r). These statistics are defined as follows:

  • 1)

    root-mean-square error (RMSE)

RMSE=i=1n(PiOi)2n
  • 2)

    Pearson correlation coefficient (r)

r=n(i=1nOi·Pi)(i=1nOi)·(i=1nPi)(ni=1nOi2(i=1nOi)2)·(ni=1nPi2(i=1nPi)2)
  • 3)

    coefficient of determination (R2)

R2=[i=1n(OiOi¯)·(PiPi¯)]2i=1n(OiOi¯)·i=1n(PiPi¯),where Pi and Oi are known as the experimental

Conclusion

This study employed a systematic approach to select the most influential parameters for the activated filter outlet concentration prediction by means of the ANFIS methodology. The ANFIS is used to eliminate the vagueness in the information and to produce the best conditions. We used the proposed ANFIS model to convert the complicated multiple performance characteristics into a single multi response performance index. As a result, the prediction methodology developed in this research is useful

References (21)

There are more references available in the full text version of this article.

Cited by (3)

  • Application of hybrid support vector regression artificial bee colony for prediction of MMP in CO<inf>2</inf>-EOR process

    2020, Petroleum
    Citation Excerpt :

    Other correlations such as Alston et al. [15], Sebastian et al. [16], Eakin and Mitch [17], Lee [18], Dong [19], and Yuan et al. [20] have focused directly on specific oil reservoirs, and this cannot satisfy the level of comprehensiveness and generalization required by various other oil reservoirs where the characteristics are different. A good number of studies have been carried out on the application of various artificial intelligence (AI) techniques such as artificial neural network (ANN), gene expression programming (GEP) and support vector regression (SVR) in many fields of science and engineering [21–30]. In petroleum and reservoir engineering, these techniques are becoming more attractive in resolving many problems of practical importance such as well-test analysis [31], PVT parameters estimation [32–34], history matching [35], WAG optimization [36], bottom hole pressure estimation [37], reservoir characterization [38–40] and others.

  • Estimating forest carbon fluxes using four different data-driven techniques based on long-term eddy covariance measurements: Model comparison and evaluation

    2018, Science of the Total Environment
    Citation Excerpt :

    The former is characterized by apparent simplicity of model design, fast learning and convergence speed (Huang et al., 2006), and the latter takes advantage of the strengths of both ANN and fuzzy inference methods (Jang, 1993). To date, the potential of these two advanced approaches for solving forecasting and prediction issues has been demonstrated by numerous recent studies in other different fields, such as water resource management (Kişi et al., 2011; Kisi et al., 2012; Petković et al., 2016; Shiri et al., 2011; Shiri et al., 2016; Yaseen et al., 2016), sustainable energy engineering (Hosseini Nazhad et al., 2017; Mohammadi et al., 2015; Shamshirband et al., 2015), thermal engineering (Jović et al., 2016; Kariminia et al., 2016a; Kariminia et al., 2016b; Kariminia et al., 2016c; Shamshirband et al., 2016) and environmental science (Kisi et al., 2013a; Kisi and Shiri, 2014; Mohammadi et al., 2016; Raos et al., 2016). To the best of our knowledge, however, no study has been reported that probes into the validity of these two innovative approaches for estimating terrestrial carbon fluxes.

View full text