logo

SCIENTIA SINICA Informationis, Volume 48, Issue 7: 856-870(2018) https://doi.org/10.1360/N112017-00300

Identification method for furnace flame based on adaptive color model

More info
  • ReceivedFeb 27, 2018
  • AcceptedMar 18, 2018
  • PublishedJul 16, 2018

Abstract

A kiln and boiler combustion system is a complex nonlinear system. The method of an adaptive color model for furnace flame recognition is proposed to improve real-time flame detection and the recognition effect of combustion flames.First, a variable adaptive color model was designed by using the combustion characteristics of furnace flames. The expression of the threshold in the adaptive color model was deduced by the method of maximum classes square error. Based on this, the human-learning optimization algorithm was used to solve the optimal segmentation threshold. Then, fast and accurate identification of the combustion conditions of furnace flames was achieved. Simulation results are presented to verify the feasibility and effectiveness of the proposed results.


Funded by

国家自然科学基金(61633016)

上海市科委国际科技合作项目(15220710400)

上海市科委重点项目(16010500300)


References

[1] Zhou J L, Yue H, Zhang J F. Iterative learning double closed-loop structure for modeling and controller design of output stochastic distribution control systems. IEEE Trans Control Syst Technol, 2014, 22: 2261-2276 CrossRef Google Scholar

[2] Chen H, Zhang X G, Hong P Y. Recognition of the temperature condition of a rotary kiln using dynamic features of a series of blurry flame images. IEEE Trans Ind Inf, 2016, 12: 147-157 CrossRef Google Scholar

[3] Marjanovic A, Krstic M, Durovic Z. Control of thermal power plant combustion distribution using extremum seeking. ieee trans contr syst technol, 2017, 25: 1670-1682 CrossRef Google Scholar

[4] Liu J X, Yang B, Cai X S. Research of combustion diagnostics and testing in boiler based on radiation spectroscopy. J Eng Thermophys, 2015, 36: 2056--2060. Google Scholar

[5] Liu Q, Wang S A, Zhang X H, et al. Flame recognition algorithm research under complex background. In: Proceedings of the 10th IEEE International Conference on Computer and Information Technology, Bradford, 2010. 503--510. Google Scholar

[6] Liu H, Wu Q S, Wang B, et al. BOF steelmaking endpoint real-time recognition based on flame multi-scale color difference histogram features weighted fusion method. In: Proceedings of the 35th Chinese Control Conference, Chengdu, 2016. 3659--3663. Google Scholar

[7] Wu Y Q, Zhu L, Zhou H C. State recognition of flame images based on krawtchouk moment and support vector machine. Proc CSEE, 2014, 34: 734--740. Google Scholar

[8] An J Y, Ma X M. Simulation of flame video image combustion stability monitoring in thermal power plant. Comput Simul, 2017, 34: 373--377. Google Scholar

[9] Xu L, Tan C, Li X. Fuel-type identification using joint probability density arbiter and soft-computing techniques. IEEE Trans Instrum Meas, 2012, 61: 286-296 CrossRef Google Scholar

[10] Cheng Y, Sheng Y X, Chai L, et al. Burning state recognition using CW-SSIM index evaluation of color flame images. Control Decis Conf, 2015, 45: 3609--3614. Google Scholar

[11] Wang Z C, Liu M, Dong M Y. Riemannian alternative matrix completion for image-based flame recognition. IEEE Trans Circ Syst Video Technol, 2017, 27: 2490-2503 CrossRef Google Scholar

[12] Han T, Lin B W. Research on flame image recognition method in natural scene. In: Proceedings of the IEEE International Conference on Information and Automation, Ningbo, 2016. 1776--1780. Google Scholar

[13] Du D J, Qi B, Fei M R. Quantized control of distributed event-triggered networked control systems with hybrid wired-wireless networks communication constraints. Inf Sci, 2017, 380: 74-91 CrossRef Google Scholar

[14] Du D J, Chen R, Fei M R. A novel networked online recursive identification method for multivariable systems with incomplete measurement information. IEEE Trans Signal Inf Proc Netw, 2017, 3: 744-759 CrossRef Google Scholar

[15] Korhonen I, Ahola J. Microwave spectra inside the combustion chamber of a kraft recovery boiler-effects on communications. IET Sci Meas Technol, 2017, 11: 740-745 CrossRef Google Scholar

[16] Yin F, Luo Z H, Sun J D, et al. Online identification method of coal species in power plant boiler furnace based on local flame spectral characteristics. Proc CSEE, 2016, 36: 5530--5539. Google Scholar

[17] Liu P Y, Zhang D L, Zhang H, et al. Experimental and numerical studies on combustion characteristics of a 600MW supercritical arch-fired boiler equipped with slot burners before and after retrofit. Proc CSEE, 2017, 37: 606--614. Google Scholar

[18] Geng Q T, Yu F H, Zhao H W, et al. New algorithm of flame detection based on color features. J Jilin Univ (Eng Technol Edit), 2014, 44: 1787--1892. Google Scholar

[19] Yin F, Luo Z H, Li Y. Coal type identification based on the emission spectra of a furnace flame. J Zhejiang Univ Sci A, 2017, 18: 113-123 CrossRef Google Scholar

[20] Li Q, Tang H, Chi J N, et al. Gesture segmentation with improved maximum between-cluster variance algorithm. Acta Autom Sin, 2017, 43: 528--537. Google Scholar

[21] Wang L, Pei J, Menhas M I. A hybrid-coded human learning optimization for mixed-variable optimization problems. Knowl-Based Syst, 2017, 127: 114-125 CrossRef Google Scholar

[22] Wang L, Ni H Q, Yang R X. An adaptive simplified human learning optimization algorithm. Inf Sci, 2015, 320: 126-139 CrossRef Google Scholar

[23] Pardalos P M. Preface. J Glob Optim, 2017, 67: 1-1 CrossRef Google Scholar

[24] Wang L, Yang R X, Pardalos P M. An adaptive fuzzy controller based on harmony search and its application to power plant control. Int J Electr Power Energy Syst, 2013, 53: 272-278 CrossRef Google Scholar

[25] Wang L, Yang R X, Ni H Q. A human learning optimization algorithm and its application to multi-dimensional knapsack problems. Appl Soft Comput, 2015, 34: 736-743 CrossRef Google Scholar

[26] Wang L, Ni H Q, Zhou W. MBPOA-based LQR controller and its application to the double-parallel inverted pendulum system. Eng Appl Artif Intel, 2014, 36: 262-268 CrossRef Google Scholar

[27] David O E, van den Herik H J, Koppel M. Genetic algorithms for evolving computer chess programs. IEEE Trans Evol Comput, 2014, 18: 779-789 CrossRef Google Scholar

[28] Zeng N Y, Zhang H B, Liu W. A switching delayed PSO optimized extreme learning machine for short-term load forecasting. Neurocomputing, 2017, 240: 175-182 CrossRef Google Scholar

  • Figure 1

    (Color online) The effect diagram of the combustion flame in different scenes. Flame image (a) I, (b) II, (c) III, (d) IV

  • Figure 2

    (Color online) The 3D histograms of color components of combustion flame image I. (a) Combustion flame image I; (b) red component; (c) green component; (d) blue component

  • Figure 3

    (Color online) The 3D histograms of color components of combustion flame image II. (a) Combustion flame image II; (b) red component; (c) green component; (d) blue component

  • Figure 4

    (Color online) The 3D histograms of color components of combustion flame image III. (a) Combustion flame image III; (b) red component; (c) green component; (d) blue component

  • Figure 5

    (Color online) The 3D histograms of color components of combustion flame image IV. (a) Combustion flame image IV; (b) red component; (c) green component; (d) blue component

  • Figure 6

    The flow diagram of the flame recognition algorithm base on adaptive color model

  • Figure 7

    (Color online) Ten flame images of furnace combustion. Original image (a) I, (b) II, (c) III, (d) IV, (e) V,protect łinebreak (f) VI, (g) VII, (h) VIII, (i) IX, (j) X

  • Figure 8

    (Color online) Comparison of fitness value in original image I. (a) Genetic algorithm; (b) particle swarm optimization; (c) human learning optimization

  • Figure 9

    (Color online) Comparison of fitness value in original image II. (a) Genetic algorithm; (b) particle swarm optimization; (c) human learning optimization

  • Figure 10

    (Color online) Comparison result of flame identification in original image I. (a) Original image I; (b) Ref. [8]; (c) Ref. [11]; (d) proposed method

  • Figure 11

    (Color online) Comparison result of flame identification in original image II. (a) Original image II; (b) Ref. [8]; (c) Ref. [11]; (d) proposed method

  • 1   Table 1Comparison of flame identification of different algorithms
    Algorithms Threshold $t_1$ Threshold $t_2$ Fitness value Identification time (s)
    Original image I GA+proposed model 185 101 9.2733E+03 0.3748
    PSO+proposed model 178 97 9.2862E+03 0.5097
    HLO+proposed model 178 97 9.2862E+03 0.0935
    Original image II GA+proposed model 151 124 1.4230E+04 0.4757
    PSO+proposed model 160 129 1.4246E+04 0.6128
    HLO+proposed model 160 129 1.4246E+04 0.1626
  • 2   Table 2Comparison of flame identification in different scenes
    Method Correct pixels Total pixels Identification accuracy (%) Identification time (s)
    Original image I Ref. [8] 33222 43621 76.16 0.2747
    Ref. [11] 42651 43621 97.78 0.1702
    Proposed method 42823 43621 98.17 0.0935
    Original image II Ref. [8] 266452 284200 93.76 0.4251
    Ref. [11] 276770 284200 97.39 0.2675
    Proposed method 280521 284200 98.71 0.1626
    Original image III Ref. [8] 587651 917631 64.04 0.7643
    Ref. [11] 835341 917631 91.03 0.6568
    Proposed method 849676 917631 92.59 0.4736
    Original image IV Ref. [8] 81957 108747 75.36 0.2065
    Ref. [11] 81413 108747 74.86 0.1950
    Proposed method 101054 108747 92.93 0.0971
    Original image V Ref. [8] 670752 972873 68.95 0.9168
    Ref. [11] 963243 972873 97.77 0.7279
    Proposed method 961911 972873 98.87 0.5164
    Original image VI Ref. 8 474027 782595 60.57 0.7238
    Ref. 11 759863 782595 97.10 0.5636
    Proposed method 774106 782595 98.92 0.3556
    Original image VII Ref. [8] 23141 28400 81.48 0.1513
    Ref. [11] 20451 28400 72.01 0.1595
    Proposed method 27549 28400 97.00 0.0813
    Original image VIII Ref. [8] 92974 97660 95.20 0.1765
    Ref. [11] 89838 97660 91.99 0.1810
    Proposed method 94374 97660 96.64 0.0977
    Original image IX Ref. [8] 539152 735232 73.33 0.6835
    Ref. [11] 393618 735232 94.34 0.5329
    Proposed method 701257 735232 95.38 0.3324
    Original image X Ref. [8] 22704 24928 91.08 0.1305
    Ref. [11] 19890 24928 79.79 0.1596
    Proposed method 23973 24928 96.17 0.0796

Copyright 2019 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1