Development of KSVGRNN: A hybrid soft computing technique for estimation of boiler flue gas components
Introduction
Gas sensing applications include analyzing a gas mixture for finding the components concentration from the sensor data using intelligent techniques like Artificial Neural Networks (ANN), Support Vector Machine (SVM), Fuzzy Logic (FL), to name a few. Control of excess air in Boilers improves boiler efficiency and also leads to air pollution control. Developing sensors to detect various gases for subsequent utilization of their properties in gas sensing areas has gained industrial importance due to their vital applications in industrial safety engineering. A lot of experimental works including a fundamental study on multi-mode quartz crystal gas sensors [1], chemically selective sorbent coating on QCR [2], and detection of Hydrogen in ambient air using a coated piezoelectric crystal [3] has been reported in the literature. In the past decade, a wide variety of pattern recognition techniques has been designed for developing gas recognition systems. Several computational models based on heuristics, Statistical parameters based designs, Optimization procedures have been developed for processing the digital data acquired from sensor responses for decision-making purposes leading to optimum resource utility [4,5].
Gas mixture composition finding consists of a few tool based computational stages such as identification, classification, prediction, estimation and optimization. This process requires periodical data observed from sensors. A good data acquisition system needs to be designed for processing the sizeable data acquired from sensors to classify them into appropriate classes. Once the classification task is over, a good estimator needs to be constructed so as to predict the gas components for quantification of each component with a practical level of accuracy.
Manual analysis of the data is time-consuming, error-prone and inconsistent. Moreover, the large sized data acquired from the gas sensors need a computational model which can capture the nonlinear relationships between sensor reading and the actual composition of gas mixtures. In this context, noteworthy contributions are found in the literature [6], [7], [8], [9], [10], [11], [12] which use Machine learning techniques [31] to address various issues such as feature selection, extraction, clustering classification prediction and estimation. Typically Neural Networks (NN) is employed to capture the nonlinearity in the prediction level. Various types of NN are used for applications ranging from classification to optimization. Since BFA involves prediction and estimation of gas components Generalized Regression Neural Network (GRNN) is found to be an appropriate tool.
The most important issue in machine learning with respect to processing sensor data is over-fitting which reduces the generalization capability of learned model. Since issues pertaining to generalization capability differ from classifier to classifier, designing an appropriate preprocessing technique as a classifier which augurs well with the GRNN is an important aspect of hybridization.
In this work, a hybrid technique has been developed by integrating the following: (i) K-Means clustering technique, (ii) Support Vector Machines based classifier (for multi- classification) and, (iii) a Generalized Regression Neural Network (GRNN). In the initial stage, the digital data acquired from the sensor responses are preprocessed for appropriate scaling of the attributes and similar data are grouped together using k-means clustering algorithm. In the next stage, an SVM with Radial Basis Function Kernel is implemented for classification. The support vectors, so generated in the higher dimensional space of the SVM are used to train the proposed GRNN. This technique is validated with volatile organic compounds data [33]. The quantitative discriminating capability of proposed technique has been analyzed in turn with data acquired from KANE [36] and ORSAT [34] analyzers, whose descriptions and specifications are furnished in [35].
Section snippets
Data and methods
This section describes the fundamental concepts related to Boiler flue gas analysis and a few Machine learning techniques pertaining to gas analysis sensors and applications.
Proposed KSVGRNN hybrid system
Support vector machine has been used as both binary and multi-class classifier, especially for classification purposes and regression techniques are used with SVM for further estimation purposes. Though use of K-means based pre-processing algorithm for conditioning the data improves the performance of the KSVMR [27], the generalization capability is not found to be satisfactory for large sized data sets. Moreover, the individual use of GRNN also poses considerable difficulties in controlling
Validation of proposed SVM-GRNN hybrid model
In order to study the performance of the proposed hybrid scheme, it was first applied to predict the composition of three gases Acetone, Ethanol and Trichloroethylene from Quartz crystal microbalance (QCM) sensor data [33]. The QCM sensor data are in the form of ordered triplet vectors.
Sixty-nine support vectors obtained from the SVM classifier were used to train GRNN to build the prediction model. Then it has been used for the quantitative identification and subsequent individual gas
Experiments, results, observations and analysis in the estimation of boiler flue gas components, volatile organic components of SnO2 sensor and QCM sensor
After ascertaining the efficiency of the proposed hybrid in testing QCM data, it has been applied for BFG data also. The data acquisition procedure of BFG has been explained in Section 2.2. In this experiment, out of 250 data acquired from the sensors, 80% of data have been used for training the hybrid and 20% for testing its efficacy. As in the previous case, grid search has been used to optimize the model parameters of the hybrid. A GRNN model has also been built up separately for comparative
Conclusion
A combinatorial structure of KSVGRNN been designed as a hybrid technique to process the data obtained from gas sensors to perform both classification and estimation and it has been found that this hybrid works very well adjusting itself to various divergent issues such as "Stability – Plasticity dilemma", Curse of dimensionality, Conflicting Meta-parameters during hybridization and it has been found that SVM augurs well with GRNN for this application problem which necessitated both
Acknowledgment
The authors of this paper wish to acknowledge and thank the Professors A.Ozmen, Feyzullah Temurtaz, Ali Gulbag and Robi Polikar for their timely support and motivation by providing their data and sharing their knowledge. The authors also thank Dr. R. John Bosco Balaguru, Associate Dean (Research), School of Electrical and Electronics, SASTRA University, Thanjavur, who provided SnO2 sensor data and shared his knowledge towards the improvement of this paper. The authors also thank the reviewers
Anantharaman Sivakumar is presently pursuing his research leading to Ph.D., degree in SASTRA University, Thanjavur, India. He obtained his Bachelor's degree in Electrical and Electronics Engineering from Bharadhidasan University, India in 1994 and Master's degree in Power System from Regional Engineering College, Trichirappalli, India in 1996. He was with Madras Fertilizers Limited, India for 13 years as Manager-Installation and Maintenance. His specific areas of research interest include
References (37)
- et al.
Solubility interaction and the design of chemically selective sorbent coating for Chemical sensors and arrays
Sens. Actuators B
(1991) Detection of vaporous and odours from a multisensory array using pattern recognition. Part 1. Principal component and cluster analysis
Sens. Actuators B
(1991)Cryptand/metal ion coated piezoelectric quartz crystal sensors with artificial back propagation neural network analysis for nitrogen dioxide and carbon monoxide
Sens. Actuators B
(2005)- et al.
Study of odor recorder for dynamical change of odor using QCM sensors and neural network
Sens. Actuators B
(2002) Pattern recognition in gas sensing: well-stated techniques and advances
Sens. Actuators B
(1995)Wavelet transform and fuzzy ARTMAP based pattern recognition for fast gas identification using a micro-hotplate gas sensor
Sens. Actuators B
(2002)- et al.
Support Vector Machines for olfactory signals recognition
Sens. Actuators B
(2003) - et al.
A novel feature selection techniques for number classification problem using PNN
Sens. Actuators B
(2009) Probabilistic Neural Networks
Neural Netw.
(1990)Finding the composition of gas mixtures by a phthalocyanine-coated QCM sensor array and an artificial neural network
Sensors and Actuators B
(2006)
Fundamental study on Multi-mode Quartz crystal gas sensors
Detection of hydrogen in ambient air using a coated piezoelectric crystal
Chem. Lett.
Support Vector Machine applications in Bio- informatics
Appl. Bio Inf.
Identification of CO and NO2 using a thermally resistive micro sensor and Support Vector Machine
IEE Proc. Meas. Technol.
Multi-component gas mixture analysis using a single tin oxide sensor and dynamic pattern recognition
IEEE Sens. J.
Pattern Classification
Continuous flue gas flow calculation in the new standard EN ISO 16911 for volume flow rate in ducts
Cited by (0)
Anantharaman Sivakumar is presently pursuing his research leading to Ph.D., degree in SASTRA University, Thanjavur, India. He obtained his Bachelor's degree in Electrical and Electronics Engineering from Bharadhidasan University, India in 1994 and Master's degree in Power System from Regional Engineering College, Trichirappalli, India in 1996. He was with Madras Fertilizers Limited, India for 13 years as Manager-Installation and Maintenance. His specific areas of research interest include Artificial Neural Network, Control Engineering and Diagnosis of Power Apparatus.
Ramakalyan Ayyagarri is currently with the department of Instrumentations & Control Engineering, National Institute of Technology, Tiruchirappalli, a premier institution of the Govt. Of India, serving in various capacities. He holds a Ph.D. from IIT Delhi where he worked on dynamic non-cooperative games and robust control for a class of nonlinear systems. He has about 20 years of research experience in the field of Control Systems and Engineering. He was a recipient of DST's (Govt. Of India) Young Scientist award during 2002, and visiting Associate Professor at the Institute of Mathematical Sciences, Chennai (India) during 2001–2004. During 2007–2011 he worked on the British Council funded UKIERI project on the development of Smart Unmanned Air Vehicles (UAVs) in close collaboration with the University of Leicester, UK, Indian Institute of Science Bangalore, National Aerospace Laboratories (A CSIR, Govt. of India Laboratory) Bangalore, and IIT Bombay. He has published numerous research articles in prestigious peer-reviewed journals and international conferences. He has also visited the Electrical Engineering Department of Texas A&M University, USA, Yantai University in China and Politecnico di Milano in Italy, upon invitation. He is a senior member of IEEE and member of Society for Industrial and Applied Mathematics (SIAM, USA). He is also the founder and currently general secretary of Automatic Control and Dynamic Optimization Society (ACDOS) of India which represents the country at the International Federation of Automatic Control (IFAC). He is deeply interested in looking into computational problems that arise out of algebra and graphs in control theory. He has several significant papers in international conferences and journals.
Aravindan Chandrabose is an Assistant Director of SSN School of Advanced Software Engineering, Chennai, India. He received his B.E. degree with honors in Computer Science and Engineering from NIT, Trichy, India in 1986 and M.E. and Ph.D. degrees in Computer Science from Asian Institute of Technology, Bangkok, Thailand, in 1990 and 1995, respectively. His post-doctoral research was with the University of Koblenz, Germany, during 1995–1997, where he worked on a funded project in the area of disjunctive logic programming. He has been in academia as a professor of Computer Science since 1998 and has completed four research projects. Aravindan has published a number of papers on logic programming and non-monotonic reasoning, soft computing, and image processing. Aravindan is a member of ACM, Association for Logic Programming, IEEE, IEEE Computer Society, IEEE Computational Intelligence Society, and Indian association of research in computer science.
Kannan Krithivasan is a professor of Mathematics at SASTRA University, Thanjavur, India. He obtained his Bachelor's and Master's degrees in Mathematics from University of Madras, India, in 1980 and 1982, respectively. He also received his Bachelor's and Master's degrees of Education from Madurai Kamaraj University, India, in 1984 and 1986, respectively. He obtained his M.Phil degree from NIT, Trichy, India in 1988. He was conferred Ph.D. in Mathematics in the area of computational fluid dynamics by Alagappa University, Karaikudi, India, in 2000. Kannan has been in academia for the past 25 years. His specific areas of research interest include combinatorial optimization, artificial neural networks, Wavelet Transforms and hypergraph based image processing.
Swaminathan Venkataraman, working as Asst. Professor in the Department of Mathematics, Srinivasa Ramanujan Centre, SASTRA University, Kumbakonam, India. He is in the academia for the past 15 years and the area of interest are categories of modules, lattices, triple systems, hypergraph and its applications and soft computing.
Sarala Durai, working as Asst. Professor in the Department of Mathematics, SASTRA University, Thanjavur, India. She is in the academia for the past 20 years and her area of interest includes graph theory, differential equations and soft computing.