Recognition Algorithm for an Infrared Flame Detector Based on an Improved Takagi-Sugeno Fuzzy Radial Basis Function Neural Network
-
摘要: 针对三波段红外火焰探测器中可能出现的单一非火焰波段通道的数据丢失、失真、饱和3种对火焰特征数据的强干扰情况,本文提出了一种改进型T-S(Takagi-Sugeno,高木-关野)模型RBF(Radial Basis Function,径向基函数)神经网络的火焰识别的鲁棒性融合算法。该算法通过聚类算法确定模型需要的模糊规则数,在模糊后件多项式中加入特征分量隶属度生成节点输出,同时定义了加权模糊节点激活度和特征表征系数代替了原先模型的马氏距离(模糊规则适用度)。通过设计三波段火焰探测器并进行了常规及鲁棒性实验,实验数据证实,改进型模型在隐含层所需节点数、收敛速度、精度、泛化能力、鲁棒性上较传统T-S模型的RBF神经网络模型、GA(Genetic Algorithm,遗传算法)-BP(Back Propagation,反向传播)模型都有明显的提升。Abstract: To address the data loss, distortion, and saturation of a single non-flame channel that may occur in a three-band infrared flame detector, a robust fusion algorithm for flame recognition based on a radial basis function (RBF) neural network entailing an improved Takagi-Sugeno (T-S) model is proposed in this paper. In this algorithm, the number of fuzzy rules required by the model is determined by a clustering algorithm. The membership degree of the feature component is added to the subsequent fuzzy polynomial to generate node output, and the weighted fuzzy node activation degree and feature characterization coefficient are defined to replace the Markov distance (fuzzy rule applicability) of the original model. Through the design of a three-band flame detector and routine and robustness experiments, it is shown that the proposed model significantly improves the number of nodes, convergence speed, accuracy, generalization ability, and robustness as compared with those of the traditional T-S model RBF neural network and genetic algorithm-back propagation models.
-
Key words:
- infrared flame detector /
- improved T-S /
- RBF neural network /
- recognition algorithm
-
表 1 部分样本示意表
Table 1. Schematic table of some samples
Sample No. x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 1 2.58 2.51 2.46 1.02 0.97 16 0.840 0.032 0.023 0.027 1.6875 0.207 2 2.39 2.71 2.43 0.88 0.89 12 0.408 0.032 0.020 0.023 2.25 0.284 3 2.07 2.32 2.10 0.89 0.90 40 0.237 0.017 0.020 0.012 1.6875 0.061 4 2.15 2.24 2.29 0.95 1.02 4 0.306 0.0261 0.023 0.019 0.5625 0.062 表 2 网络效果比较
Table 2. Comparison of network effects
Network type RSME training Training accuracy RSME Test Test accuracy Node Improved T-S-RBF 0.007 100% 0.006 100% 15 Traditional T-S-RBF 0.025 100% 0.071 97.5% 50 GA-BP 0.035 100% 0.074 98.7% 17 表 3 数据丢失模型效果比价
Table 3. Comparison of data loss models
Network type RSME of 3.8 μm Test accuracy RSME of 5.0 μm Test accuracy Improved T-S-RBF 0.0226 100% 0.0051 100% Traditional T-S-RBF 0.6185 90% 0.9356 77% GA-BP 1.3362 53% 0.2379 99% 表 4 数据失真模型效果比价
Table 4. Comparison of data distortion models
Network type RSME of 3.8 μm Test accuracy RSME of 5.0 μm Test accuracy Improved T-S-RBF 0.0093 100% 0.0109 100% Traditional T-S-RBF 0.386 96% 0.6542 86% GA-BP 0.3769 94% 0.0605 100% 表 5 数据饱和模型效果比价
Table 5. Comparison of data saturation models
Network type RSME of 3.8 μm Test accuracy RSME of 5.0 μm Test accuracy Improved T-S-RBF 0.0227 100% 0.0475 100% Traditional T-S-RBF 1.4715 45% 1.2798 51% GA-BP 0.6012 89% 0.6364 86% -
[1] 魏崇毓, 王馨民.基于红外、紫外双波段探测的智能水炮系统设计[J].红外技术, 2016, 38(10): 877-883. doi: 10.11846/j.issn.1001_8891.201610012WEI Chongyu, WANG Xinmin. Design of Intelligent Water Cannon System Based on PIRS and UV Detection[J]. Infrared Technology, 2016, 38(10): 877-883. doi: 10.11846/j.issn.1001_8891.201610012 [2] 邓理文, 刘晓军.基于模糊神经网络的智能火灾探测方法研究[J]. 消防科学与技术, 2019, 38(4): 522-525. doi: 10.3969/j.issn.1009-0029.2019.04.021DENG Liwen, LIU Xiaojun. Intelligent fire detection method based on fuzzy neural network[J]. Fire Science and Technology, 2019, 38(4): 522-525. doi: 10.3969/j.issn.1009-0029.2019.04.021 [3] YAO Lina, ZHANG Yanna. Fault diagnosis and model predictive tolerant control for non-Gaussian stochastic distribution control systems based on T-S fuzzy model[J]. International Journal of Control Automation & Systems, 2017, 15(1): 2921-2929. [4] 程玉虎, 王雪松, 孙伟.自适应T-S型模糊径向基函数网络[J]. 系统仿真学报, 2007(19): 4440-4444. doi: 10.3969/j.issn.1004-731X.2007.19.021CHENG Yuhu, WANG Xuesong, SUN Wei. Proposal of Adaptive T-S Fuzzy Radial Basis Function Network[J]. Journal of System Simulation, 2007(19): 4440-4444. doi: 10.3969/j.issn.1004-731X.2007.19.021 [5] WU Qinghui, WANG Xinjun, SHEN Qinghuan. Research on dynamic modeling and simulation of axial-flow pumping system based on RBF neural network[J]. Neurocomputing, 2016, 186: 200-206. doi: 10.1016/j.neucom.2015.12.064 [6] 张谦, 王双红, 陈海峰.基于RBF模糊神经网络的垂直轴风力机设计[J]. 计算机测量与控制, 2014, 22(7): 2237-2239, 2243. doi: 10.3969/j.issn.1671-4598.2014.07.072ZHANG Qian, WANG Shuanghong, CHEN Haifeng. Research on vertical axis wind turbine control algorithm based on RBF fuzzy neural network[J]. Computer Measurement and Control, 2014, 22(7): 2237-2239, 2243. doi: 10.3969/j.issn.1671-4598.2014.07.072 [7] JUANG Chiafeng, CHIU Shihhsuan, CHANG Shuwew. A Self- Organizing TS-Type Fuzzy Network With Support Vector Learning and its Application to Classification Problems[J]. IEEE Transactions on Fuzzy Systems, 2007, 15(5): 998-1008. doi: 10.1109/TFUZZ.2007.894980 [8] 张颖.改进的T-S模糊神经网络在化工软测量中的应用[J]. 电子测量与仪器学报, 2010, 24(6): 585-589. https://www.cnki.com.cn/Article/CJFDTOTAL-DZIY201006017.htmZHANG Ying. Improved T-S fuzzy neural network applied in soft sensing of chemical industry[J]. Journal of Electronic Measurement and Instrument, 2010, 24(6): 585-589. https://www.cnki.com.cn/Article/CJFDTOTAL-DZIY201006017.htm [9] 阮慧, 党德鹏.基于RBF模糊神经网络的信息安全风险评估[J]. 计算机工程与设计, 2011, 32(6): 2113-2115, 2128. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201106060.htmRUAN Hui, DANG Depeng. Risk assessment of information securitybased on RBF fuzzy neural network[J]. Computer Engineering and Design, 2011, 32(6): 2113-2115, 2128. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201106060.htm [10] 袁积德.三波段红外火焰探测器的研究与开发[D].杭州: 浙江大学, 2012.YUAN Jide. Research and Development of Triple Channels Infrared Flame Detector[D]. Hangzhou: Zhejiang University, 2012.